SAMPLE ATTRITION AND CRONBACH-ALPHA - A 5-YEAR LONGITUDINAL-STUDY

Conventional wisdom indicates that as attrition in a field sample occu rs, the remaining sample becomes more similar, e.g., homogeneous. This narrowing of the differences in the sample will cause a decrease in r eliability as measured by Cronbach alpha. The current research is a 5- yr. longitudinal study that involved a group of auditors from five Big Six firms. The initial sample (Bernardi, 1994a) of 494 auditors was f ollowed for a period of five years to assess the effect of attrition o n Cronbach alpha for the Defining Issues Test (Rest, 1979a). At the be ginning of the research, alpha was .348; however, by the end of Year f ive, alpha had decreased to .309. When the sample was stratified by se x and geographic location, the data indicate that, while alpha increas ed for the majority of the sample, it dramatically decreased for women in the New York City Metropolitan area. This study should assist othe r researchers mho are examining data on a longitudinal basis when the sample is presumed to become more similar over time.