Non-response bias

The concept of silence distortion (English non-response bias) referred to in the empirical distortion of the result which arises because respondents give different answers than those who would give that have not responded. Cause is thus a systematic non-response. If the bias in the survey on (unit nonresponse ) is also called selectivity of the sample or sample selectivity.

Example

In an office are ready for a day questionnaires on which the office staff to document their workload. Depending on the office staff it is (slightly ) different, is to be measured, the average workload.

• An office employee A, who finds no opportunity to fill due to its heavy workload, leaves his high reading is not included in the final result ( he conceals it).
• An office employee B, who can complete the questionnaire due to its lower workload, however, can incorporate its low measured value.

The end result of the average workload is distorted by silence of A, which would have given a different answer than B.

Check for distortion of silence

There are several ways to reduce a distortion of silence. For a non-responding respondents can be interviewed specifically for the causes of non-response, such as by telephone demand. Also, they can be brought to an answer by repeatedly presenting the questionnaire yet.

However, there are statistical ways to recognize a moment of silence distortion. As can be compared with incoming replies early arriving late, such as by the first compared with the last third of the incoming questionnaires under a t-test. If there are no significant differences in response behavior, this can indicate a lack of silence distortion, since it is assumed that the last third of the respondents is the non-responding respondents closest. However, it was pointed out that by this method is not the behavior of non- respondents, but only the behavior of late respondents is considered, whereby it is indeed suitable for an examination on Zögerverzerrung (late -response bias), unsuitable for testing on silence distortion, however. For a late reply, an entirely different reason than a lack of response.

It can be compared instead of a priori known variables (eg, firm size, industry ) of respondents and non- respondents. This can often also be used for homogeneity t- test or chi-square test.