Short CATI-questionnaire as foot-in-the-door technique to refresh a CAPI establishment panel study
Jul 25, 11:30
Due to panel attrition and changes in the panel population, there is a need in long-term panel studies to refresh the sample. Especially in establishment surveys and longitudinal studies, it has become more and more difficult to recruit (new) sample members. While telephone pre-contact is a common technique to narrow the relevant sample, there is less knowledge about the effect of using short telephone interviews in a panel study as a foot-in-the-door.
We tested a possible way of recruiting new members - asking a short questionnaire in the first wave as foot-in-the-door technique. The underlying idea is that this two-step approach will lead to a higher response rate compared to if the respondents had been asked the full questionnaire in the first run. Indeed computer assisted telephone interview (CATI) studies are known to lead to lower response rates than computer assisted personal interview (CAPI) studies. Nevertheless, we expect, using a short CATI questionnaire as foot-in-the-door, to increase the likelihood that establishments will participate in a panel study in the long run. Therefore, the question is if it is more efficient to recruit panel refreshers with a short-CATI-questionnaire first than to use a full CAPI mode.
To answer this question we conduct an experimental design on two recruitment strategies. Database for our investigation are the 2016 and 2017 waves of the BIBB-qualification-panel, a yearly establishment panel study with 3500 establishments in Germany. We refreshed the 2016 wave of the sample in two ways: a) the full questionnaire via computer assisted personal interview (CAPI) and b) a short version of the questionnaire via CATI technique. Based on this data and the participation in the next wave, we can evaluate the likelihood of participating in the panel study in the next wave depending on being a) a “normal” refresher or b) a CATI-short-version-refresher. Furthermore, as we have additional information on participating and non-participating establishments (size, industry, region) within both survey modes, we can test for environmental effects using logistic regression models.