Sixty-Eight Returns and a Guide Who Wants One Hundred
A student of a government junior college is doing her Commerce Practical Work project on 'consumer awareness and preference for branded versus unbranded packaged drinking water in the local market'. She distributed 120 questionnaires to households and shopkeepers. Sixty-eight came back complete, twelve came back partly filled, and forty were never returned. Of the 68 complete returns, 22 are from students in her own hostel, all aged 17 to 19. Only 4 of the 25 shopkeepers approached agreed to answer. Her draft questionnaire contains: 'Don't you agree that branded water is safer than unbranded water?'; 'Do you buy branded water because of its taste and its safety?'; and 'How much do you spend on packaged water?' with no time period. The project guide has told her to 'make it 100 respondents by filling the gaps yourself, drop the hostel ones so the sample looks random, and write the conclusion as 80 per cent prefer branded'. The teacher has asked for objectives, methodology, tables and graphs, findings, limitations and a bibliography.
1. How many responses can the student legitimately use, and how should she report the shortfall and the composition of her sample?
Read the worked answer
Usable responses: 68. The twelve partly filled returns can be used only for the questions they actually answer, and that must be stated question by question, not silently pooled. The forty unreturned questionnaires are non-responses and cannot be counted, filled or estimated.
How to report it:
- Response rate: 68 ÷ 120 = 56.7 per cent, with the twelve partial returns counted separately as 10 per cent. A 57 per cent response rate is normal for a student survey and is a finding in itself, not a failure to hide.
- Composition: 22 of 68 (32 per cent) are hostel students aged 17 to 19 - a cluster with similar age, income and access, not a cross-section of households. Shopkeepers number only 4 of 25 approached (16 per cent), so no conclusion about retailers is possible.
- Honest treatment: report the sample as convenience or purposive, not random, and present the hostel group and the non-hostel group as two sub-samples side by side. If the two groups differ sharply, pooling them would average away the most interesting result.
The judgement point is that a small, well-described sample supports cautious, qualified conclusions, while an inflated sample of 100 supports nothing. Dropping the 22 hostel responses to make the sample 'look random' would not make it random; it would merely misdescribe it and throw away a third of the data.
2. Rewrite the three flawed questionnaire items and name two additional data items the student needs to answer her objective.
Read the worked answer
Item 1 - leading question. 'Don't you agree that branded water is safer than unbranded water?' invites only agreement and puts the answer in the respondent's mouth. Rewrite: 'In your opinion, which is safer to drink: branded packaged water, unbranded packaged water, both equally safe, or cannot say?' with a tick-box for each option.
Item 2 - double-barrelled question. 'Do you buy branded water because of its taste and its safety?' asks about two reasons at once; a respondent who buys for safety but not taste cannot answer. Split it: 'Which one factor matters most to you when buying drinking water: taste, safety, price, availability, or brand name?' followed by a separate ranking of reasons.
Item 3 - no time frame and no unit. 'How much do you spend on packaged water?' is unanswerable without a period and a basis (household or person). Rewrite: 'How much does your household spend on packaged drinking water in a typical month? Less than Rs 100 / Rs 100–300 / Rs 301–500 / more than Rs 500.' Fixed brackets give comparable data.
Two additional data items:
- Source of purchase and frequency - shop, home delivery, water can, online; daily, weekly, occasionally. Without this, 'preference' cannot be linked to actual buying behaviour.
- Awareness of the BIS/ISI mark or the licence number on the bottle. The stated objective is consumer awareness, yet nothing in the draft measures it; a yes/no/not-sure item on the mark makes the objective testable.
A household size or income band is also worth adding, since it allows cross-tabulation and turns a list of percentages into an analysis.
3. The guide asks her to invent 32 responses, drop the hostel returns, and write a conclusion of '80 per cent prefer branded'. Advise her, referring to research ethics and to what makes a project report credible.
Read the worked answer
She should decline all three instructions, and she can say why without being rude.
Inventing 32 responses is fabrication. It is not a shortcut; it is the manufacture of evidence. If detected, the project is cancelled and the practical marks are lost, and it teaches a habit that would end a career in accounts or business. Every figure in a report must be traceable to a returned questionnaire kept in an appendix.
Dropping the hostel returns misdescribes the sample. Discarding a third of the data because it does not look random does not make the remainder random; it makes the methodology statement false. The correct move is to disclose the composition and, if she wishes, analyse the two sub-groups separately - which is better analysis, not worse.
Pre-writing the conclusion reverses the method. The conclusion must follow from the data. If 54 per cent of respondents prefer branded water, she writes 54 per cent and discusses why. A hypothesis may be stated in advance and then rejected; a well-argued finding that contradicts the expectation earns credit, while a fabricated 80 per cent is indefensible when the questionnaires are attached.
What a credible report contains: objectives stated before the survey, the instrument in an annexure, the achieved sample and response rate, tables and graphs drawn only from the returns, findings tied to each objective, a limitations section naming non-response bias, the hostel cluster and the near-absence of retailer data, and a bibliography distinguishing primary from secondary sources. A defensible remedy within the deadline is to conduct fifteen short, documented shopkeeper interviews to strengthen the weakest part of the sample - real data collected late is worth far more than invented data collected on time.