Découvrez les nouvelles fonctionnalités de la version 4.0.

An Experimental Study of Two Data Collection Strategies: Face-to-face Interviews (CAPI) versus Telephone Interviews (CATI)

Ref. 6970

Description générale

Période concernée

-

Région géographique

Informations géographiques additionnelles

Agglomeration of Bern

Résumé

In a methodological experiment, carried out within the framework of the Swiss Household Panel, two data collection strategies are evaluated: Computer Assisted Personal Interviews (CAPI) and Computer Assisted Telephone Interviews (CATI). The study was designed as a split-ballot in combination with a Multitrait-Multimethod design. In the first stage of the test, an initial sample is split up into two halves. One halve of the respondents is interviewed by telephone, the other halve is visited by an interviewer. In the second stage, the groups are partially crossed over: Part of those interviewed face-to-face the first time are interviewed by telephone the second time, and vice-versa. In all interviews in both waves, a small selection of questions is repeated within the same interview. A broad range of criteria for comparison are provided by this combined design: costs, speed, response rates, answer distributions and summary statistics, as in the traditional comparison studies. But also estimates of data quality defined as reliability and validity or random and systematic measurement error, associated with the different modes of data collection. In addition, we plan to look at the interaction effects of the mode of data collection and some characteristics of questions that are known to induce response effects, such as sensitivity to social desirable answering and complexity of questions. These aspects are hypothesized to have differential effects within different mode of data collections. Both modes of data collection included in this study are computer assisted (CATI and CAPI), which augments its significance for the current survey practice and diminishes the lack of knowledge in the literature.

Résultats

The results of the first part of the study show that participation rates in the two modes were similar on both the household and the individual level. In addition, the effects of question characteristics on the answer distributions were the same in both modes. These findings imply that, for panel studies, there is no reason to prefer either CAPI or CATI with regard to initial response rates and answer distributions. Moreover, it is shown that CATI saves costs and time in comparison with CAPI. In the second part of the study, estimates of reliability and validity were obtained for all data by using a Multitrait-Multimethod design. Comparing these estimates, the conclusion must be that the choice of CATI versus CAPI has no implications for the quality of the data. Once the choice has been made, it does not make a very large difference either which type of response scale is used, although 10 point scales and real number scales are slightly better than categorical scales in both interviewing modes. What can really make a difference is the formulation of the question: The form in which respondents have to indicate their agreement- disagreement with statements can better be rephrased into a direct question. In addition, sensitive topics can negatively influence the validity of the data, especially in CATI.