Predictive Validity of the Graduate Record Examination with and without Range Restraints

Abstract

Most studies show that scores on the Graduate Record Examination (GRE) have low predictive validity for graduate grade point average. One suggested explanation is that the input and output variables have restricted ranges. Restriction of the input variables results when students with low GRE scores are omitted from the analysis. The output variable is constricted since most students receive an A or B in courses. However, the present study shows that the GRE has low predictive validity even when both the input and output variables are more widely distributed for a sample of graduate students. The need for better screening mechanisms to select from among applicants to graduate school is discussed.