When drawing conclusions about the target population based on a sample, the assumptions are subject to survey errors that can have different origins:
- Noncoverage. The possibility of being included in the sample does not exist for all members of the target population (e.g. no internet access, an incomplete sampling frame, etc.).
- Nonresponse. The responses of the individuals that were not included in the survey differ from the responses of people that were included.
- Sampling. Measurements are not performed on all units of the target population.
- Measurement. Respondents’ answers deviate from their actual values. This can occur due to unsuitable operationalization, different contextual factors, etc.
The first three types of errors happen because certain units are not observed, whereas measurement errors occur during unit observation, either due to the respondent, measurement instrument, data collection methods, or the interviewer (if present). The latter does not apply to web questionnaires because they are self-administered.
Let us take a closer look at different measurement errors:
- Errors due to the respondent occur because different respondents submit data with a different error rate, for they differ in the terms of cognitive capabilities and motivation when entering data. If we consider web surveys, this occurrence does not differ greatly when compared with other survey modes.
- Errors due to the measurement instrument occur when the questionnaire wording or the sequence of questions influence the quality of answers. It is therefore important to make sure that the questionnaire is easily understood and to employ methods that prevent effects of the sequence of questions. With the help of computer-assisted surveys and appropriate software we can reduce item nonresponse, inconsistent answers and wrong answers, thus improving data quality.
Thanks to the branching functionality, 1KA offers the creation of dynamic questionnaires. This enables the collecting of highly consistent data and reduces the probability of unsuitable answers. Automatic skips of unsuitable questions mean that there are no interruptions in the survey process.