Understanding the Relation Between Attitude Involvement and Response Latitude Using Item Response Theory

Abstract

Adapting the original latitude of acceptance concept to Likert-type surveys, response latitudes are defined as the range of graded response options a person is willing to endorse. Response latitudes were expected to relate to attitude involvement such that high involvement was linked to narrow latitudes (the result of selective, careful responding) and low involvement was linked to wide latitudes (the result of disinterested, careless responding). In an innovative application of item response theory, parameters from Samejima’s graded response model were used to examine response latitude width. Other item response theory–based tools (e.g., test characteristic curves, information functions) were used to examine the influence of response latitudes on the psychometric functioning of several attitude surveys. A mix of experimental and nonexperimental methods was employed to create groups of high and low involvement surveys. Comparisons of these surveys showed that high involvement was related to significantly narrower response latitudes than low involvement. Furthermore, wide response latitudes were related to unfavorable psychometric properties such as reduced survey discrimination and reduced internal validity relative to narrow latitudes. Comparisons of information functions in high and low involvement conditions, however, were less consistent. Implications of wide response latitudes are quite unfavorable for researchers and suggest that an element of error is present when respondents feel little involvement with an attitude topic.