General-Specific Questions in Survey Research: A Confirmatory Factor Analysis Approach

• Activity Number: 577
• Type: Contributed
• Date/Time: Wednesday, August 3, 2011 : 2:00 PM to 3:50 PM
• Sponsor: Section on Survey Research Methods
• Author(s): Fred Lorenz and Virginia Lesser and Laura Hildreth*+ and Ulrike Genschel
• Companies: Iowa State University and Oregon State University and Iowa State University and Iowa State University
• Keywords: survey research ; part-whole questions ; general-specific questions ; serial correlation ; structural equation modeling

Abstract

Questionnaires often contain "general - specific" questions about the quality of community life in which a general question (G) either precedes (GS) or follows (SG) a series of specific items (S) related to the general question. One hypothesis is that the R2 is larger in a regression of G on the specific items in the SG conditions as respondents interpret G as a call to summarize. The magnitude of the regression coefficients can provide insight into the importance of specific items, but interpretation is difficult due to collinearity. An alternative approach is confirmatory factor models where each item is a manifestation of one or more latent variables. In this model, the same insight can be gained by comparing factor loadings in the GS and SG conditions. In addition, insight can be gained into the processes by which people answer questions. We hypothesize that respondents answer questions in an orderly sequence, represented as first order serial correlations between adjacent items. Our results support this hypothesis as evidenced by an improvement in the fit of the model over what one would expect if pairs of items were correlated at random.