Non-sampling error refers to any deviation between the results of a survey and the truth which are not caused by the random selecting of observations. That is, non-sampling error is the total of all forms of error other than sampling error. Common types of non-sampling error include non-response error, measurement error, interviewer error, adjustment error, and processing error.
Non-response error refers to errors that are caused by differences between people that participate in surveys versus people who do not participate in surveys. For example, surveys that ask people about how they spend their time likely have large amounts of non-response error, as people who spend their time doing surveys are likely quite different from those who do not.
Non-response error can be narrowly defined as relating to whether people selected to participate actually do participate (e.g., the difference between people who opened their email invitation and completed a survey versus those who did not); or it can be more broadly defined to include all non-random aspects of sampling (e.g., selection of the list to be used in the research). Errors relating to whether the lists used in research are representative are also known as list selection error and coverage error.
Measurement error refers to all the errors relating to the specific measurement of each sampling unit (as opposed to errors relating to how they were selected to be measured). For example, these could include confusing question wordings, low-quality data due to respondent fatigue, and low quality multi-item scales being used to measure abstract concepts.
Interviewer error occurs when an interviewer makes an error in how they administer the survey or record responses. For example, in qualitative research, an interviewer may “lead” a respondent to a certain answer, and in quantitative research a bored interviewer may choose to ask a question in words that they regard as superior to those in the questionnaire.
Adjustment error occurs where the analysis of the data inadvertently adjusts the data in such a way that it becomes less accurate. The main forms of adjustment error are errors with weighting, data cleaning, and imputation.
Processing error occurs when the processing of the data has caused an error of some kind, such as when it is incorrectly entered or corrupted.
Check out our other handy What Is... guides for more!