Did you know? ‘Eight in ten students prefer the new library’ sounds persuasive—until you discover that only ten people answered the questionnaire, even though 200 were invited. The numbers alone do not tell you whether the survey describes a class, a school or merely a few enthusiastic respondents. Three phrases—sample size, response rate and representativeness—change how responsibly we read that claim.
The useful answer: sample size tells us how many people or units were included in an analysed sample; response rate tells us what proportion of eligible sampled or invited units responded, under the survey’s stated definition; and representativeness asks how well the participating group reflects the population we want to understand. A large number of replies does not automatically make a sample representative. A high response rate is useful information, but it does not by itself prove that the answers are unbiased.
The reader’s job: find the denominator before believing the headline
Consider a fictional school survey headline: ‘Most students want later library closing hours.’ It may come from a carefully sampled group of different age levels, or from a link shared only in a club chat. Both can have honest responses; they support different conclusions. Vocabulary helps a reader ask, ‘Most of which people?’ That question protects classroom discussions, parent decisions and later professional choices from accidental exaggeration.
The two tempting mistakes are counting only the people who agree and forgetting the people who were never asked or never replied. A percentage is always a part of something. Until the denominator and selection method are clear, ‘80%’ is not a complete story.
Six survey words worth separating
Population: the whole group a claim is about
Suppose the question is whether all students at one school want additional library hours. That school’s enrolled students form the intended population for the question. It is different from ‘students who use the library every day’. A survey of daily visitors may be useful for designing a library service, but it should not quietly become a claim about every student. In ordinary conversation, population often means residents of a country; in statistics it means the complete group of interest.
Sampling frame: the list or method from which people can be selected
A sample might be drawn from class lists, a school email directory or a set of addresses. That source of possible participants is often called a sampling frame. It matters because students not on the list cannot be chosen through it. A survey link placed only in the library club’s online group does not cover every student merely because the link is technically shareable.
Sample size: the number included in the relevant set
If 40 students completed a survey and all 40 answers were used in a particular analysis, a headline might state n = 40. But take care: some questions may have 40 answers, others 37 because some respondents skipped them. The denominator for each item can differ. Sometimes researchers report the number originally sampled, the number eligible and the number completing the questionnaire separately. Do not silently treat these figures as identical.
Response rate: how many eligible sampled units responded
In a very simple school example where all 200 invited students were eligible and 80 returned complete forms, a basic completed-response proportion is 80 ÷ 200 = 40%. But professional response-rate calculation has important rules about eligibility, partial responses, contact outcomes and survey design. The American Association for Public Opinion Research (AAPOR) sets out standardised approaches and warns that response rate alone is not a sufficient measure of survey quality. This article’s arithmetic is a teaching illustration, not a substitute for formal survey-disposition coding.
Nonresponse: missing answers may matter
If a quarter of the invitees reply, what do the other three quarters think? We do not know from that survey alone. It is tempting to assume nonrespondents agree with respondents, but their experiences may differ. Perhaps the busiest pupils are least likely to complete a survey about homework hours. Nonresponse bias describes potential distortion when respondents and nonrespondents differ on the outcomes that matter. A low response rate raises questions, but its relationship with actual bias depends on circumstances; a high rate does not automatically remove bias.
Representative: does the sample reflect the group we wish to describe?
A school-wide estimate needs attention to who was covered and how participants were selected. If responses mainly come from one level, programme or interest group, percentages may not describe the rest fairly. Representative is not a synonym for large, popular or carefully written. It is a claim that must be justified by the way the sample relates to its target population.
Worked example: a library-hours survey
Fictional exercise: A school invites 200 eligible students to answer a short questionnaire. Eighty respond; 50 say they favour later library closing. The simple completed-response proportion is 80 ÷ 200 = 40%. Among the 80 respondents, the proportion in favour is 50 ÷ 80 = 62.5%. These are two separate percentages with two separate denominators.
What sentence is supported? ‘Of the 80 students who answered, 62.5% favoured later library closing’ is accurate for this example. What sentence is not yet supported? ‘62.5% of the school’s students want later hours’ assumes the respondents represent everyone. We would need to know who was invited, how invitations were made, who did not answer, and whether the responses plausibly represent the relevant school population.
Now suppose an online story rounds that 62.5% to ‘nearly two-thirds’. That may be a reasonable approximation to the respondents’ percentage, but if the story drops of respondents, it changes the reach of the claim. Grammar alters what the numbers are said to represent. This is why English and Mathematics must work together in serious reading.
Another worked example: ‘80%’ can describe tiny or substantial groups
Imagine two fictional student polls: Poll A receives 10 responses, of which 8 agree. Poll B receives 1,000 responses, of which 800 agree. Both report 80% agreement among respondents. Poll B has many more responses, but neither figure tells us alone how the participants were selected. If Poll B was a voluntary online click poll open to fans of a particular option, it may still be systematically unrepresentative.
A classroom learning goal is to say what we know without erasing what we do not. Better sentences are ‘8 of 10 respondents to Poll A agreed’ and ‘800 of 1,000 respondents to Poll B agreed’. A stronger report then explains the target population and participation method. There is no need to invent a margin of error for a voluntary-response poll.
Response rate is not agreement rate
Students sometimes calculate 50 ÷ 200 = 25% and then call that the response rate in our library example because 50 pupils agreed. But the 50 figure answers an agreement question, not ‘How many invited students responded?’ The response rate uses responses; the agreement rate among respondents uses favourable answers. Change the noun, change the denominator. The formula is only meaningful when the labels are correctly matched.
Another common mix-up concerns completion rate. An online form might record 100 people who opened it, 70 who started answering, and 60 who submitted every required question. A site might display a ‘60% completion rate’ using openers as a denominator, or 60 ÷ 70 ≈ 85.7% using starters as a denominator. The label alone is insufficient; check the definition. Professional survey response rates may require a different denominator altogether.
Why question wording matters as much as the arithmetic
Compare these two fictional questions: ‘Would you support an exciting improvement that gives students more library time?’ and ‘Would you prefer the library to close at 6 pm rather than 5 pm, if staffing and costs are available?’ The first contains favourable language and vague benefits; the second names a possible decision while adding a condition that should be explained. Neither is perfect by definition, but the difference shows why leading, neutral, balanced and ambiguous are useful vocabulary.
Response options also matter. ‘Always’, ‘often’, ‘sometimes’, ‘rarely’ and ‘never’ measure frequency, not level of agreement. ‘Strongly agree’ and ‘strongly disagree’ describe attitudes. A learner who combines unlike scales into one tidy percentage may produce an impressive-looking but hard-to-interpret result. Ask what each response category meant to the participants.
Collocations that improve survey reading and writing
Useful word partnerships include target population, sample size, survey respondents, response rate, voluntary participation, nonresponse bias, representative sample and question wording. In an English composition, ‘A survey found…’ is usually too thin if the audience needs to judge the result. ‘In a fictional survey of 80 student respondents…’ states the scope immediately.
The terms random and representative should not be substituted. Random selection describes part of a process; representation is a property to assess in relation to the population and potential errors. Significant also has both everyday and technical statistical senses. A ‘significant’ concern in a news headline need not be a statistically significant result. Students can learn to pause when a familiar adjective does unusually heavy work.
A reading routine for Primary and Secondary learners
For younger students, begin with a class choice rather than formal polling theory. Ten fictional classmates answer which fruit they prefer; four choose mango. Ask ‘Four out of how many?’ Next, imagine that half the class was absent. How should the sentence change? ‘Four out of ten respondents’ is more careful than ‘Everyone in class loves mango.’ Children can practise the difference between some, most, all and none without memorising advanced statistical formulas.
For upper-primary and secondary students, provide both numerator and denominator, then add the number of people invited. Ask for two distinct percentages. Finally, ask whether the survey was a complete class census, a sample chosen by a defined method or a voluntary response. Students can present a bar chart—but the chart needs clear labels, because graphical presentation cannot repair an unsupported claim about the whole group.
A practical transfer task is to read two poll reports and write one cautious comparison paragraph. Require students to use three phrases accurately: among respondents, number invited and may not represent. Then ask them to explain why the stronger or weaker conclusion follows from the design. For chart-reading foundations, see eduKateSG’s graph interpretation guide.
Five diagnostic questions with model reasoning
Question 1: what percentage of invited students replied?
Of 200 eligible invitees, 80 respond. The simple reply proportion is 80 ÷ 200 = 40%. Check that these counts are the correct eligible denominator and respondent numerator for the classroom exercise.
Question 2: what percentage of respondents agree?
Of the 80 respondents, 50 agree. That gives 50 ÷ 80 = 62.5%. It is not the same as the response rate.
Question 3: what must the headline say?
A careful headline can say ‘About 63% of surveyed respondents favoured later hours’ and a fuller report gives the count and survey method. Saying ‘63% of all students’ without evidence overreaches.
Question 4: can a large sample still be biased?
Yes. Thousands of volunteers who all come from one group can give a very large yet systematically unbalanced set of answers. Size helps answer some questions, but selection and coverage still matter.
Question 5: does a high response rate prove the survey is accurate?
No. AAPOR explains that survey quality needs attention to multiple sources of error, not simply the participation fraction. Question wording, coverage, measurement, and nonresponse differences can matter.
A three-day retrieval and transfer exercise
Day 1: Write definitions without looking: population, sample, respondent, response rate. Match each to a fictional school questionnaire. Do not move on until the student can explain the denominator out loud.
Day 2: Recalculate the library example after changing the numbers to 120 invitees, 90 respondents and 54 in favour. The simple response proportion is 75%; the proportion in favour among respondents is 60%. Ask why one fraction uses 120 and the other 90. Provide one correct sentence for each.
Day 3: Transfer to a new setting—a neighbourhood park questionnaire or a student council poll. Ask who did not have a chance to answer, which groups may have been missed and what cautious action the organisers can reasonably take. The goal is not cynicism. It is a habit of asking proportionate questions before treating a survey as a verdict.
For parents and teachers: the encouraging question to ask
Instead of replying ‘That percentage is misleading’, ask ‘Can you show me who is included in the percentage?’ The learner may simply have omitted a small phrase such as of respondents. Repairing the sentence together teaches more than rejecting a number. If the student asks whether a small school survey is pointless, explain its proper use: a small, clearly described consultation can still uncover useful suggestions, provided we avoid generalising beyond its evidence.
The idea also bridges Science: repeated readings of the same object need not be independent observations, just as many survey answers from one narrow group need not describe everyone. See repeated measurements versus independent trials for a hands-on analogy. Broader language practice is available through the eduKateSG Vocabulary Learning Hub.
Frequently asked questions
If 100% of respondents agree, does everyone agree?
No. It means all respondents in that stated group agreed. The conclusion about everyone else depends on how the survey was designed and who participated.
Is response rate the same as participation rate?
Not always. They can be calculated differently depending on eligibility, completion, invitations and definitions. Look for the survey’s methodology rather than assuming any two displayed percentages are identical.
Does a low response rate make a study useless?
Not automatically. Low response can increase concern about nonresponse bias, but the amount and direction of actual bias depend on how respondents differ from others. Interpret the study’s purpose, design and limitations together.
Can primary-school children learn these words?
Yes. Start with a bowl of ten paper responses and ask ‘Who answered?’ before formal statistical terminology. Older learners can calculate the proportions and explain sampling limitations.
Fact-checking references: AAPOR: response rates and why they are not a complete quality test; AAPOR: eligibility, contact and response outcomes in survey-rate calculation. All counts and scenarios in this article are fictional teaching examples, not claims about actual students or polls.