Split-Sample Tests of "No Opinion" Responses in an Attribute-Based Choice Model

Researchers using questionnaires to elicit preferences must decide whether to include response options that allow respondents to express “no opinion.” Using a split-sample design, we explore the implications of alternative answer formats including and not including no-opinion responses in an attribute-based choice experiment. The results indicate that using multiple no-opinion responses may enable researchers to differentiate between respondents who choose no-opinion options due to satisficing and those expressing utility indifference. Existing literature suggests no-opinion responses may be treated as “no,” but our results show treating no-opinion responses as “no” can yield substantially disparate preference estimates. (JEL C25, Q24, Q25, Q51)