skip to main content skip to footer

Trends in SAT Content and Statistical Characteristics and Their Relationship to SAT Predictive Validity SAT

Marco, Gary L.; Crone, Carolyn R.; Braswell, James S.; Curley, W. Edward; Wright, Nancy K.
Publication Year:
Report Number:
ETS Research Report
Document Type:
Page Count:
Subject/Key Words:
History of Testing, Scholastic Aptitude Test (SAT), Test Construction, Test Validity, Trend Analysis


Recent analyses by Morgan (1989) and Ramist and Weiss (1990) estimated that the average multiple correlation of the SAT with college freshman grades increased by about .05 in the early 1970s and decreased by about the same amount from the mid-1970s to the mid-1980s. Intensive analyses were made of changes associated with the test during this time. This report provides the results of these analyses and evaluates the possible effect on predictive validity of changes in the SAT's content and format, statistical characteristics, and equating procedures. The editions of the SAT that provided data for the study were those taken by the classes entering college from 1971 to 1985, the classes that provided data for the validity studies conducted through the College Board's Validity Study Service (VSS). Since more than half of the SAT scores for the high school seniors who graduated the previous spring came from the tests administered in November and December of the senior year, the test-specific analyses for this study used item and test data from the November and December SAT editions. The analyses of item and test statistics from November and December SAT test administrations from 1970 and 1984 provided little evidence linking changes in the test to changes in predictive validity. None of the patterns of change in the various statistical or equating indices that were studied was consistent with the trends in predictive validity from 1971 to 1985. The few noticeable trends were too slight to have more than a marginal effect on predictive validity. (260pp.)

Read More