Non-sampling error
Non-sampling error is error in a study's results caused by anything other than sampling itself — bad wording, wrong frames, non-response, or data mistakes — and it does not shrink with a bigger sample.
See it move
Non-sampling error comes from anything other than sampling itself — bad wording, wrong frames, non-response, data mistakes. A firm mails a survey to its full 800-customer list; only 240 respond, a 30% response rate, and 12 of those 240 forms are logged incorrectly, 5% of usable responses. Unlike sampling error, which shrinks predictably as the sample grows, non-response and data-entry slips do not shrink just because more people are asked.
Check yourself
A university draws a correct random sample of 400 students from its full enrolment list and surveys them about study habits. Only 100 students respond, and follow-up records show the 300 non-respondents were disproportionately students working long part-time hours. Which type of error does this best illustrate?