We analyze whether there are differences in kind and quality of ego-centered network data that were collected on the web compared to data collected with the help of an interviewer. Many mode effects on the quality of web survey data have been analyzed (Couper 2000) for more or less standard questions. The collection of ego-centered network data, however, is notoriously more difficult than the collection of other survey data because it faces the respondent with a more complex answering task. Until now, we do not know whether web surveys lead to a substantial reduction of data quality and quantity (see Lozar Manfreda, Vehovar & Hlebec 2004).
In our study university researchers of different disciplines at a Dutch university (n=110), either filled out ego-centered data through a web questionnaire or were probed about their network in a personalized interview. Ego-centered network data collection makes use of so-called “name generators” that are both hard to explain properly and moreover give the respondent a special opportunity to shorten the answer procedure. For this reason, interviewers are often deemed critical in the proper collection of ego-centered network data. Hence our analysis provides a strong test of the feasibility of data collection through the Internet. The emphasis of our analysis is on whether there are differences in the ego-centered network data between the groups of respondents. One could argue that respondents who are not interviewed in person are tempted to answer in a much more time saving manner than other respondents, and are moreover more likely to make mistakes or simply quit. We focus on drop out rates, the number of missing values, the size and density of the ego-centered networks, as well as other properties of the networks. The paper presents the results of the hypotheses testing.