Rituparna's Questionnaire on Girls' Dropout in the Char Villages

Flaw 1: the wrong population - she would be studying the survivors. Distributing questionnaires to girls currently in Classes VIII to X samples exactly those who did not drop out. The register's fall from 96 to 41 tells her where the phenomenon is, but her instrument cannot reach it. Her findings would describe the educational experiences of girls who stayed, and she might conclude that early marriage or distance is unimportant simply because the girls sitting in front of her did not leave for those reasons. The sampling frame must be built from past registers across at least three cohorts, supplemented by community listing.

Flaw 2: gatekeeper capture. The headmaster selects respondents, explains the purpose and collects the forms. He is also part of the phenomenon being studied - teacher absence, fees, corporal punishment, a broken toilet and the timetable are all plausible causes of dropout, and none is likely to be reported through him. He has already refused income questions, which shows he is willing to shape the instrument.

Flaw 3: instrument–population mismatch plus a self-administered format. A closed questionnaire assumes literacy and that the real reason is on the list of five. Dropout reasons in char villages are often none of the five: harassment on the road, a younger sibling to care for, a school that floods, a marriage proposal, a shift in the boat timetable. Self-administration with mothers who cannot read also rules out the household as a unit of analysis.

Two practical flaws compound these: July is the worst month for access, and the ₹4,000 budget with three weeks cannot support 200 usable returns plus travel.

Redesign: a comparative case study of two accessible char villages.

1. Build a retrospective cohort. Use three years of school registers to list girls who were enrolled in Class VIII and did not appear in Class X, with name and village. This solves the survivor problem at almost no cost.
2. Household interviews, interviewer-administered. Twenty-five to thirty households, using a short schedule read aloud, with the mother and the girl present, plus open questions. A household roster with land, assets, siblings, and months of male migration replaces the refusal-prone income question.
3. Unstructured interviews with eight to ten girls who left school, contacted through the register list, the ASHA worker and the self-help group - snowball sampling where no list exists.
4. One focus group with mothers, which produces community-level reasoning that individual interviews miss.
5. Observation: walk the route to school and, if permitted, sit in a classroom; record the road, the boat, the toilet and the timetable, since these are part of the explanation.
6. Triangulate register data, household accounts and interviews; report where they disagree.

What it can claim: a credible, mechanism-level account of why girls in these two villages stop after Class VIII, and testable hypotheses for a wider study.

What it cannot claim: a dropout rate for the char areas or for Nagaon district, or any statistical generalisation. That needs a probability sample drawn from a village-level frame, with DISE or district records, and a budget several times her own. She should say so explicitly rather than inflate her findings.

It is nearly useless for three reasons, each of them a general lesson about survey categories in rural Assam.

1. Seasonality. Most fathers migrate for four to six months. Asked in July, a man may be reported as 'labourer'; asked in January, as 'farmer'. The variable measures the date of the interview as much as the household's economy.
2. Household, not individual, production. Char households combine cultivation, wage labour, fishing, livestock and remittances. A single occupational label for one male member discards most of the information relevant to whether a girl stays in school.
3. Who is present. In a migrating household the person with authority over the girl's schooling may be the mother or the grandmother. An indicator built on the absent father misses the decision-maker entirely.

What she should measure instead:

• a household roster with each member's activities and months of absence over the last 12 months;
• land cultivated and owned, plus livestock and assets (a durable-goods count is easier to report honestly than income);
• number and ages of siblings, and birth order of the girl;
• mother's literacy and schooling, which is repeatedly a stronger predictor of a daughter's continuation than father's occupation;
• distance and travel mode to the nearest secondary school, and whether the route is seasonally cut off;
• whether the household has a migrant remittance and its regularity.

The general point: choose variables that match how the household actually earns and decides, rather than variables that fit neatly into a table.