There is considerable interest in using mixed mode strategies for conducting social surveys, in particular to enhance response rates and fieldwork efficiency while at the same time reducing costs. There is also interest in mode differences in survey measures. In Wave 2, Computer Assisted Personal Interviewing (CAPI) and Computer Assisted Telephone Interviewing (CATI) were examined. In Waves 5 and 6, we contrasted CAPI and Web interviewing.
In the Wave 2 mixed mode experiment, the IP2 sample was divided into three equal sized experimental groups, and each group received a different treatment in terms of questionnaire mode and sequence of modes. Within PSUs, households were randomly assigned to experimental treatment – and all individuals within households were treated the same way. There were three experimental groups:
Group 1, CAPI: Households in this group were only eligible for face-to-face interviews.
Group 2, CATI “Move one, move all”: All households in this group were issued for telephone interviewing; if one person could not be interviewed by telephone all remaining members were transferred to CAPI. CATI was the mode of first contact, if that mode failed; the case was transferred to face-to-face. If enumeration was completed in CATI, individual interviews would be attempted within CATI until any household member indicated that they were unable to complete the interview by telephone. This may be because they refused, were classified as a noncontact or were unwilling/unable to complete the interview by phone. As soon as one individual interview could not be obtained, all outstanding household members were allocated to field and attempts were made to interview the remaining sample members face-to-face.
Group 3, CATI “Try all”: All households in this group were issued to telephone interviewing and more attempts were made to complete interviews by telephone. CATI was the mode used to contact the household. If that mode failed; the case was transferred to face-to-face for follow-up. If enumeration was completed in CATI, individual interviews were attempted within CATI. Households in this group were only eligible to be transferred once attempts had been made to contact and interview all household members by telephone. Attempts to contact each household member by telephone continued even if one household member was unable to be interviewed by telephone. The case was transferred to field for face-to-face interviews only after all eligible adults had either refused, were classified as non-contacts or were unwilling/unable to complete the interview by phone.
Experimental allocation is given by the variable b_ff_modew2 in the data file b_hhsamp_ip. It is important to realise that the household enumeration and household questionnaire could be completed in one mode, with the individual questionnaires being completed in different modes from these or from other individuals interviewed in the household. The variable b_trtocapi on the record b_hhsamp_ip indicates that the case was transferred at some point from CATI to CAPI. The variable b_hhmodetype on the record b_hhsamp_ip indicates the mode the household enumeration was completed in, the variable b_hhmodetype in the data file b_hhresp_ip indicates the mode of administration for the household questionnaire. Finally, the variable b_indmode in the data files b_indall_ip and b_indresp_ip indicates the mode of administration for the individual questionnaire.
The same questionnaires were used for CAPI and CATI, with only some necessary adaptations for telephone, such as dropping references to showcards.
Twenty households and 37 individuals were issued face-to-face and interviewed face-to-face but used the telephone instrument. The variable b_modeallerr on the datafile b_indresp_ip flags the cases with this error in mode allocation.
