Six Weeks, Three Schools, One Confused Design
Rehana, a second-year student, wants to find out why girls in her taluk's government schools drop out between classes 8 and 10. She has six weeks, no budget, one classmate to help, and permission from three schools. She has drafted a 30-item questionnaire in English, to be posted to 200 households, containing items such as: “Don't you agree that early marriage and household work are the main reasons girls drop out?” and “Do your parents neglect your education because they are uneducated?” She also intends to “do a case study” of one dropout, and to sit in a class 9 classroom for a week to observe teaching. Her teacher returns the draft saying the design is confused: three methods, one student, no clarity about what each is meant to answer.
1. Identify the flaws in the questionnaire using methodological terms, and rewrite one item.
Read the worked answer
The flaws, named
- Leading question. “Don't you agree that…” invites agreement and supplies the answer, so the finding becomes an artefact of the wording rather than a fact about the respondent.
- Double-barrelled item. “Early marriage and household work” bundles two distinct causes; a “yes” tells you nothing about which one operates, or in what order.
- Loaded and judgmental wording. “Neglect” and “because they are uneducated” insult the respondent and will produce defensive or socially desirable answers - especially with a stranger at the door.
- Closed form on an open question. A fixed 30-item instrument administered once cannot surface a reason the researcher did not anticipate, which is the main thing she needs.
- Language and literacy mismatch. English items posted to Kannada-speaking, partly non-literate households will depress the response rate and bias the achieved sample towards the literate and better-off - precisely the group she is not studying.
- No pilot and no sampling frame. There is no list from which the 200 were drawn, so the response rate cannot even be computed.
Rewrite
Do not: “Don't you agree that early marriage and household work are the main reasons girls drop out?”
Do (in Kannada, read aloud, one cause per line): “For girls in this village, how important is each of these in stopping school after class 8? (a) marriage (b) work at home (c) cost (d) distance and safety (e) the school itself. Very important / somewhat important / not important.”
Then, unprompted: “In your own words, what happened in your family?”
The rewrite removes the leading frame, separates the two causes, and adds one open item so the data can surprise her.
2. Given her question and her constraints, which methods should she use and which should she drop? Justify each choice.
Read the worked answer
Fit the method to the question, then to the constraints
Her question is descriptive (how many, and why), explanatory (which factors), and sensitive. Six weeks, no budget, one helper, three schools: that rules out a probability sample and any design requiring travel across the taluk.
Survey - keep, but change the instrument. Best for prevalence and comparison across households. A posted questionnaire is the wrong vehicle (response rate, literacy). An oral interview schedule administered to dropouts listed in the three schools' registers is feasible: the sampling frame already exists in the admission and attendance registers, and the population is small - perhaps 40 to 60 girls.
Case study - keep, but narrow. Good for mechanism and meaning: how the decision was actually made in one household. It cannot support a claim about “the main reason”. Use five or six, chosen for contrast - one who returned to school, one married, one in paid work, one from a landless household.
Observation - drop the week. A structural problem, not a scheduling one: the girls who dropped out are absent. She would be observing the survivors and inferring about the leavers - selection on the dependent variable. Defensible only as two days of context on the school environment.
The design
Five weeks: records from three schools; a 10-item oral schedule to every recorded dropout; six contrast case studies; three teacher and three parent interviews; two days of school observation for setting.
That is triangulation: the schedule gives distribution, the cases give process, the observation gives context. Where they disagree, the disagreement is a finding, not a failure of the study.
3. What ethical safeguards must be built into this study, and which one is hardest to honour?
Read the worked answer
Safeguards to build in, not bolt on
- Consent and guardianship. Most respondents will be minors. Obtain guardian consent and the girl's own assent. A signature from a male head of household does not override her refusal.
- Confidentiality and anonymity. No names, no school-identifiable details; codes on schedules; notes kept locked. But state plainly what cannot be promised.
- No harm, no stigma. Never use the word “dropout” in front of neighbours; interview in private; no photographs; no recording without permission.
- Institutional permission. Head teachers and the Block Education Officer in writing; do not remove registers from the school.
- Power and expectations. She is a college student; parents may believe she can produce a scholarship. She must say clearly that she cannot.
- Reciprocity. Leave a short Kannada summary for each school, without inflating hopes.
- Honest reporting. Null or negative findings get reported too, and the three schools should not be named if the findings are critical of them.
The hardest one
Disclosure of child marriage. If a respondent reveals that a girl under 18 was married, the researcher cannot promise absolute secrecy - yet a promise of confidentiality is exactly what makes her talk. She must decide in advance, with her teacher, what she will do: not report an individual, but raise the pattern with the school and the Child Development Project Officer without identifying anyone, unless there is a risk of serious harm.
This must appear in the consent script before fieldwork. Improvising it mid-interview is how a good study turns into a harmed respondent and a failed project.