The most common method for eliciting attitudes and beliefs in surveys is to employ rating scales where respondents are asked to choose the alternative that best describes their belief in, attitude towards or agreement with a statement. Such response scales typically contain between three and seven alternatives and can be bipolar (strongly agree, agree, neither agree nor disagree, disagree, strongly disagree) or unipolar (always, often, sometimes, occasionally, never).
Attitude items that use a single bipolar rating scale are ubiquitous in surveys. Evidence suggests that a two-step branched method may be more advantageous in terms of response processes and statistical properties. In the branched or unfolding bipolar format, respondents are first asked about the overall direction of their belief or attitude (e.g. satisfied or not satisfied). Then they are asked about degree (very satisfied, somewhat satisfied, slightly satisfied). A composite score can then be computed for the pair of items.
This experiment was first conducted in Wave 3. Households within PSUs were assigned to one of two experimental treatments, with special attention to being independent of the any showcard experiments.
The controlling variable is for Wave 3 is c_ff_branchingw3 on record c_hhsamp_ip. The two groups are:
Group 1 – Branched
Group 2 – Unbranched
The affected questionnaire items are c_nbrcoh1_a to c_nbrcoh4_d2, and c_poleff1_a to c_poleff_d2 on record c_indresp_ip. The questions are about neighbourhood cohesion and political efficacy.
At Waves 4 and 5, the experiment was repeated, with the same allocation to treatments as at Wave 3. The controlling variable is w_ff_branchingw3 on record w_hhsamp_ip, with relevant questionnaire items w_nbrcoh1_a to w_nbrcoh4_d2, and w_poleff1_a to w_poleff_d2 on record w_indresp_ip where “w” indicates “D” and “E” for Waves 4 and 5.
