A Multitrait-Multimethod Validity Investigation of Scores from a Professional Licensure Examination

Abstract

The construct validity of scores from the Uniform CPA Examination was investigated through the construction of a multitrait-multimethod matrix. The four sections of the exam were treated as distinct traits, and within-section subscores were created according to item format. First, Campbell and Fiske’s four criteria were used to evaluate the matrix. After correcting the correlations for attenuation due to unreliability, evidence was moderate for the presence of four discrete constructs within the exam. The monotrait-heteromethod correlations were large and statistically significant and in a majority of cases were larger than other correlations found within the matrix. Structural equation models were fit to the data to explore whether excessive method variance was present. The results indicated that there was very little method variance present in the data. The implications for understanding the constructs measured by this exam, and for evaluating the current practice of reporting separate exam section scores, are discussed.