Collapsing Scale Categories: Comparing the Psychometric Properties of Resulting Scales

This study compares a scale, the Rosenberg Self-Esteem Scale, that was administered with four response categories to versions of the same scale that were administered with six and eight response categories. Respondents were randomly assigned to take one of the three versions of RSES. A rating scale utility analysis was conducted on all three versions creating two new four-category versions after collapsing (or combining) adjacent categories. The three different four-category versions were compared on such properties as average scores, correlations with external variables, and factor structure. While most of the psychometric properties were similar across all versions, there were moderate differences related to criterion validity: the scale that did not need to be collapsed had the strongest relationship with external variables, even though there were slightly stronger correlations for the collapsed versions compared to their original scales. A recommendation is made that if it is found that too many categories are being used for a scale then new administrations of the scale should also use the new format, however, an argument could be made to continue administering the survey in its original format, but then collapse responses before analysis