US College Entrance Exams (e.g. SAT-I) Perform Useful Functions • Reliably PREDICT Success in College—used in admissions selection • Provide academic STANDARD OF COMPARISON that is consistent across time and location Both goals require extensive time and effort for validity, for a large evolving bank of questions of varying difficulty
Confused reasons for dumping SAT-I (on which its critics disagree) • Regardless of what it is assumed to measure, SAT-I for 40 years has proven predictive of success in college (grades in all years,and also of progress to graduation). • SAT-I predictive power has been under-estimated, for well-understood statistical reasons • Even at face value, SAT-I prediction recently surpassed high school grades, because they have been inflated
Even “small” correlation coefficient is valuable predictor • Ex: if even odds of failure/success randomly, when using r=0.4,0.5,0.6 to select top 10% yields success 78, 84, 90% of the time [Taylor+Russell 1939] • UCLA students, divided into national percentiles, from Top Academic Rank 1 (98-99%), down to lowest Academic Rank 6 (88-89%) show strong systematic differences in success [from 1998 report of Committee on Undergraduate Admissions, next page]
1) Their sample consists only of UC students enrolled under current policy including SAT-I, and II, and High School grades (no correction was attempted for this “restricted range” problem) Thus they missed the real correlation that would be present if they included significant numbers of students with low SAT-I’s Geiser/Studley sample underestimates SAT-I predictive power in several serious ways
2) The Geiser/Studley report was further compromised by admissions decisions which used SAT-I as a compensating factor for lower High School grades • For example: students in the study with weaker high-school grades tended to have higher SAT-I’s, which was the reason they qualified for admission. • So a spurious anti-correlation between SAT-I and high school grades was artificially injected, which does not in fact exist
3) Studies have shown that students entering with higher SAT-I’s (and SAT-II’s) self-select tougher majors, heavier course loads • Geiser/Studley did not account for the intrinsic differences among college grades in different courses and majors, with widely varying grading standards [see: Elliot and Strenta 1988]
College Entrance Exams are a Bad way to “send a message” to secondary schools and students • They are NOT true assessment tests taken by all students (don’t confuse college admissions with proficiency/graduation testing) • Scores received too late for “feedback” to high school students • Students already know they must study hard in their courses
SAT-I is Fair to Underprivileged students • Studies show that SATs predict somewhat higher success at UC for under-represented minorities than they actually obtain. • Therefore these students will not benefit in admissions which transfer weight from SATs to high school grades. On average, both “messengers” tell a similar story. • Manipulating entrance exams cannot accomplish political goals.
Other concerns about SAT are unpersuasive • Expensive exam preparation courses for affluent? Objective studies show they add small gains (20-40 points out of 1600 total), over what is expected from simply re-taking the exams [see Powers, 1983] • Undue burden on students? Then why do most of them take more tests than minimum? • Poor schools wasting too much time drilling? • Damaging to students’ self-esteem? No evidence they suffer from SAT-induced low self-esteem Such “problems” would be worse for Curriculum tests
Curriculum-based tests not better • Nearly impossible to distinguish “achievement” from “ability”, let alone measure one independently of the other • SAT-I material—language use, quantitative reasoning and reading comprehension, is at the heart of all secondary school and college curricula, even without specific courses dedicated to it—skills which evolve slowly over many decades • UC study did not chose curriculum-based tests on their predictive power, coachability, or fairness
Recent Conclusion of “SAT Controversy” Vindicates its Defenders • Not a single other University (out of a thousand) considered dropping the SAT (even the California State University system) • Pres. Atkinson’s original “fast-track” demand for “curriculum-based” exams has been rejected, and completely dropped
New SAT Revisions are Very Modest • SAT-I will simply transfer the writing examination from the SAT-II (“Achievement” test), which was already required at UC • No student in US is obliged to take any particular subject (“Achievement”) exam other than Writing and Mathematic • At most they take one SAT-II Subject exam of their own choice, which is almost worthless because so many select their home language (Spanish or Chinese).
Conclusion of the Debate • University of California system will stay with the SAT-I • So also will every other major University in the United States
SAT-I is by far the most universal of our 3 measures of academic preparation • It is the only direct means of comparing preparation of UC students with the rest of the U.S. (and inside Calif.). Strong academics are same here as elsewhere • Trying to remove UC from the academic competition with leading universities does not seem feasible, even though Pres. Atkinson compares it to “the educational equivalent of a nuclear arms race.” All other universities will continue to measure student qualifications in part with SAT, a process which has served UC well. UC should not seek to seal itself off from comparison with them. • Throwing away the best-established standardized test raises long term danger of eroding academic standards at UC, the foundation of its excellence
Non-academic and/or subjective factors already loom large in UC admissions • It is almost impossible to get admissions officials to state clearly and explicitly how important non-academic factors have become in last few years (“it’s holistic.”) • Even before comprehensive review was adopted this year, among UCLA’s 6400 Academic Rank 2 and 3 applications, 99.2% of those with “high life challenges” were admitted, but only 18.0% of those with “low life challenges were admitted.