Introducing Web in a mixed‐mode establishment survey: Effects on nonresponse

This study reports on a mode design experiment in which a Web starting mode was introduced for the first time in the Institute for Employment Research Establishment Panel. A cross‐sectional sample of establishments was randomized to be interviewed via the traditional face‐to‐face procedure or a Web‐first sequential mixed‐mode design with face‐to‐face follow‐ups. Extensive administrative data were used to estimate and compare nonresponse bias at multiple phases of the sequential mixed‐mode design, and assess the relationship between mode design and establishment characteristics on the likelihood of response. We show that the final response rates and nonresponse bias were similar between both mode designs, but these results contrasted with the results at each phase of the sequential mode design. Larger establishments were significantly more likely to respond in the Web mode compared to the face‐to‐face mode. A moderate cost savings (of about 14% per respondent) was estimated for the Web‐first sequential mode design.