Why don't they answer? Unit non-response in the IAB Establishment Panel

"With the IAB Establishment Panel the Institute for Employment Research (Institut für Arbeitsmarkt- und Berufsforschung - IAB) has conducted a large-scale annual establishment survey in western Germany since 1993 and also in eastern Germany since 1996, covering some 15,500 establishments in the meantime. In this panel survey the establishments are asked in face-to-face interviews1 to provide information on key determinants of employment. The IAB Establishment Panel is a survey in which the same establishments are contacted each year. New establishments are added to the sample each year in order to depict structural change. Furthermore additional establishments have to be included in the sample to replace those which have dropped out of the sample in the course of time. For despite all efforts, one problem arises in every survey: individual establishments' unwillingness to participate, which leads to so-called non-response. Two forms of non-response can be distinguished. First, an establishment may refuse to participate in the survey at all (unit non-response). Second, participating establishments may fail to answer individual questions in the questionnaire (item non-response). Both types of non-response can lead to biased results if the cases of non-response are not random. Unit non-response leads to greater problems, however, as no interview is available for these establishments and it is not just the case that individual questions are not answered. Experience made with the IAB Establishment Panel shows that the willingness of establishments surveyed for the first time to participate in the survey is clearly lower, at 36 percent, than that of establishments which have already been included in the survey at least once. Furthermore the unit non-response of establishments surveyed for the first time has increased in the past few years. The willingness of the panel establishments, in other words the repeat respondents from previous years, to complete the questionnaire is considerably higher at about 80 percent. There is no indication that the willingness to participate is declining over the years. The advantage of panel surveys is also that a wealth of establishment information is available from previous years for the establishments which have been surveyed repeatedly but which no longer respond and this can be used to model the non-response process. On the other hand little information is available about the establishments which are included in the survey for the first time. It is important to analyse the non-response processes in order to gain the most precise insight possible into the survey process. The findings obtained in this way make it possible to optimise the fieldwork management, thus contributing to quality improvements and possibly to cost reductions. In addition to this, the analyses can reveal any selectivities that may lead to biased estimates. The representativeness of the projection of all variables can also be jeopardised by possible selectivities. Furthermore, in panel surveys selectivities may intensify over time. The aim of this paper is to examine the unit non-response of establishments which have already taken part in the survey at least once and are approached again. On the basis of an extended conceptual framework for establishment surveys, determinants that influence the non-response process are to be brought out. For the first time for establishment surveys the interviewer's influence on the success of the interview is taken into account both in the conceptual framework and in the analyses." (Author's abstract, IAB-Doku) ((en))