Interpreting a Campus Survey
A student club at a NU-affiliated college surveys 200 students about whether the library should extend its hours. Of 200 respondents, 130 say yes. However, the club distributed the survey only through its own Facebook group, whose members are mostly honours students who study late. The college has 2,400 students, including many who commute long distances and cannot stay late. The principal says the survey 'proves' students want longer hours and plans to cut the library's morning shift to fund it. A librarian objects that morning users are mostly first-year students who need help finding materials.
1. Evaluate whether the survey supports the principal's conclusion. Identify the sampling problem and its direction of bias.
Read the worked answer
The survey does not support a general conclusion. The sample is 200 of 2,400 students, drawn only from the club's Facebook group - a self-selected, non-random group skewed toward honours students who study late. This is a convenience sample with selection bias. The bias runs in a predictable direction: late-studying students are more likely to favour extended hours, so the 65% 'yes' figure overstates support among all students. Commuters and first-year students are underrepresented. To justify the decision, the club would need a random sample of the full student body, or at least stratified sampling by year and commuting status, with a reported response rate. The current data can suggest interest among one group but cannot 'prove' campus-wide preference.
2. The principal proposes cutting the morning shift. What evidence would be needed to judge that trade-off fairly?
Read the worked answer
The principal needs data on usage by time of day: how many students use the library in the morning versus the evening, who they are (year, commuting status), and what they use it for. The librarian's claim that morning users are first-years needing help should be tested with records or a short usage survey. Fair judgement requires comparing the number of students harmed by cutting mornings against the number helped by extending evenings, and checking whether both could be served with a smaller change (for example, staggered staffing) rather than a zero-sum cut. Without usage data, the principal is trading a documented service for an inferred demand, which reverses the burden of proof.
3. Design a better survey in three or four sentences, naming the key features.
Read the worked answer
Draw a random sample from the full enrolment list, stratified by year of study and by commuting status, aiming for a few hundred respondents. Ask about preferred library hours, current usage patterns, and constraints such as transport, using neutral wording that does not lead toward 'yes'. Collect responses through multiple channels - email, classroom announcements, and the library desk - not only a club group, and report the response rate. Pre-test the questions on a small group to catch ambiguity. This design reduces selection bias, captures the commuting and first-year voices missing from the original survey, and produces evidence proportionate to a decision that affects the whole college.