Developing a Set of Decision Rules for a Responsive Split Questionnaire Design

• Activity Number: 226
• Type: Topic Contributed
• Date/Time: Monday, August 1, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Jeffrey Gonzalez*+
• Companies: Bureau of Labor Statistics
• Keywords: Composite estimation ; Cost-error tradeoffs ; Matrix sampling ; Multiphase sampling ; Paradata ; Propensity model

Abstract

Responsive designs use data collected about the sample to inform mid-survey decisions, or actions, affecting the error properties of the resulting statistics. A key element of these designs is the decision rule, a function of the collected data to an action. Before survey practitioners can choose actions that will positively affect the statistics' error properties, they must understand the relationships in the data so that they are more informed about tradeoffs when choosing among actions. Responsive designs can be utilized in many settings, e.g., split questionnaires. A responsive split questionnaire would entail administering subsets of questions of a full questionnaire to sample members based on the data collected about the sample during an initial phase. We use data from the Consumer Expenditure Survey, a panel survey collecting information on the spending habits of consumers, to explore the extent to which information from the first interview, or initial phase, is related to events, e.g., purchases, in later interviews. With this information, we develop decision rules regarding which subsets of questions to administer to sample members at subsequent phases of data collection.