Q.Explain the difference between sampling error and non-sampling error. State whether each of the following is an example of sampling error or non-sampling error:

The dividing line is whether the error would still occur if a census (no sampling at all) were carried out - if yes, it is non-sampling error; if the error exists only because a sample rather than the whole population was studied, it is sampling error.

Sampling error is the difference between a sample result and the true population value that arises purely because a sample, rather than the whole population, was examined - it exists even under perfect survey execution, and it falls as sample size rises, becoming zero for a full census.

Non-sampling error covers every other source of error (questionnaire design, non-response, interviewer bias, recording mistakes, an incomplete sampling frame) - it can occur in a census exactly as easily as in a sample survey, and does not necessarily shrink as sample size grows.

1. A respondent misunderstanding a poorly worded question is a flaw in the questionnaire design/administration - this exact same mistake could happen even if every single member of the population were surveyed (a census), so it is a non-sampling error.
2. Two different random samples of the same size from the same population giving slightly different sample means is exactly the phenomenon that only arises because sampling (rather than a full census) was used - this is the defining case of sampling error.
   ✓Final answer

   (a) Non-sampling error (a questionnaire/administration flaw that would occur even in a census). (b) Sampling error (natural sample-to-sample variation that disappears in a census).

Students often assume any error found in survey data must be 'sampling error' - but any error that would also occur in a full census (bad questions, biased interviewers, wrong recording) is non-sampling error, regardless of how it was discovered.