College Performance Predictions and the SAT - Working Paper 45 - Research findings and estimator proposal

Sep 14, 2026

Uploaded by Riley

Studies assessing how well the SAT predicts college performance often omit many other available predictors and depend on implausible sample selection assumptions. This working paper proposes a new estimator that remains consistent under more reasonable conditions, incorporating high school demographic variables. Using University of California data, the SAT’s measured contribution is about 75% lower than standard estimates. The result reflects the SAT acting as a proxy for omitted background characteristics, which inflates its apparent predictive power, with an application to the UC “Four Percent Plan.”

LanguageEnglish

CategoryResearch & Report

FormatPDF

Pages45

Reading Time1 h 53 min

AccessFree to download

Abstract

Introduction

SAT criticism and de-emphasis in admissions

Defense of the SAT and how it predicts performance

Econometric issues in SAT validity literature

Key omissions and bias in measured contribution

Evidence from across-high-school variation

Baseline specification with school-level predictors

CENTER FOR LABOR ECONOMICS UNIVERSITY OF CALIFORNIA, BERKELEY WORKING PAPER NO. 45 College Performance Predictions and the SAT* Jesse M. Rothstein Department of Economics University of California, Berkeley [jrothst@econ.berkeley.edu](jrothst@econ.berkeley.edu) March 2002 Abstract Previous studies of the SAT's contribution to predictions of academic performance ignore many of the other variables available for prediction and are consistent only under unrealistic sample selection assumptions. I propose a new estimator of the SAT's contribution that is consistent under plausible assumptions and I include high school demographic variables as predictors. In University of California (UC) data, the SAT's contribution is about 75% lower than the usual estimates, a result equally attributable to the two innovations. One important implication is that the SAT functions as a proxy for omitted background characteristics in sparse prediction models, and that this serves to inflate the SATís apparent contribution. An application evaluates the UC's new ìFour Percent Plan,î which enhances academic quality but about which the usual models are overly optimistic. JEL Codes: C24, C53, I20, J78 Keywords: SAT validity, college admissions, sample selection, omitted variables * I thank Jared Bernstein, David Lee, Darren Lubotsky, Justin McCrary, Jack Porter, Eugene Smolensky, the participants of the Berkeley and Princeton Labor Lunches, and especially David Card for helpful comments. Errors are my own. This material is based upon work supported under a National Science Foundation Graduate Research Fellowship and a grant from the University of California's Center for Studies of Higher Education (CSHE) . Opinions, findings, conclusions and recommendations expressed in this publication are those of the author and do not necessarily reflect the views ofthe NSF or the CSHE. 1 Introduction The SAT college entrance exam has been widely criticized: as an impediment to minority students' college admission, as too expensive, and as the cause of distortions to high school curricula (Lemann, 1999; Nairn & Associates, 1980; Leonard & Jiang, 1999; Slack & Porter, 1980; Jencks & Crouse, 1982) . In recent years, several prominent colleges have de-emphasized the SAT score in admissions. The University of California, for example, recently enacted an admissions rule whereby students ranked near the top of their high school class are admitted without regard to their SAT scores; a proposal to abandon the SAT entirely|replacing it with subject exams tied to California'so􀀎cial high school curriculum|is under consideration (Atkinson, 2001) . 1 The defense of the SAT has been simple: The SAT helps predict academic performance (Camara & Echternacht, 2000; Stricker, 1991; Willingham et al., 1990; Camara, 2001; Izumi, 2001) . In this view, the acknowledged correlation between SAT scores and student socioeconomic status is an unfortunate side e􀀋ect of educational inequality in the United States: Students from disadvantaged backgrounds are simply not as well prepared to succeed in college, and the SAT should be credited, not blamed, for measuring this shortfall (Caperton, 2001) . The College Board, publisher of the SAT, has the following description on its web site: The SAT I measures developed verbal and math reasoning abilities related to successful performance in college. It provides a standard by which the skills of students applying to colleges and universities can be compared. Studies show that using both SAT I scores together with high school records provides a more accurate prediction of future academic success than using either alone. ([http://www.collegeboard.org/sat/html/admissions/about001.html](http://www.collegeboard.org/sat/html/admissions/about001.html)) This defense relies heavily on the demonstration that the SAT contributes to predictions. A typical study uses freshman grade point averages of students at a single college as a met

FAQs

Why do many SAT validity studies give misleading results?

Many studies ignore other predictors beyond high school grades and racial indicators and rely on unrealistic sample selection assumptions, which can bias the measured SAT contribution upward.

What does the paper propose to improve estimation?

It proposes a new estimator of the SAT’s contribution that is consistent under plausible assumptions, and it adds high school demographic variables as predictors.

What is the main finding from UC data about the SAT’s contribution?

In UC data, the SAT’s contribution to predicting academic performance is about 75% lower than usual estimates, reflecting that the SAT can proxy omitted background characteristics.

Sep 14, 2026

Uploaded by Riley