View in EDS HTML Full Text PDF Full Text

The University of California Was Wrong to Abolish the SAT: Admissions When Affirmative Action Was Banned

Saved in:
Bibliographic Details
Title: The University of California Was Wrong to Abolish the SAT: Admissions When Affirmative Action Was Banned
Language: English
Authors: Donald Wittman (ORCID 0000-0002-1073-4345)
Source: Educational Measurement: Issues and Practice. 2024 43(2):55-63.
Availability: Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us
Peer Reviewed: Y
Page Count: 9
Publication Date: 2024
Document Type: Journal Articles
Information Analyses
Education Level: Higher Education
Postsecondary Education
High Schools
Secondary Education
Descriptors: College Entrance Examinations, Admission Criteria, Grade Point Average, Disproportionate Representation, Socioeconomic Status, Disadvantaged, Affirmative Action, Minority Group Students, Equal Education, Educational Policy, Trustees, Scores, Data Analysis, High School Graduates, Grades (Scholastic), Grade Prediction
Geographic Terms: California
Assessment and Survey Identifiers: SAT (College Admission Test)
DOI: 10.1111/emip.12598
ISSN: 0731-1745
1745-3992
Abstract: I study student characteristics and academic performance at the University of California, where consideration of an applicant's ethnicity has been banned since 1996 and SAT scores were used in admitting students to the university until fall 2021. I show the following: (1) SAT scores were more important than high school grades in predicting first-year university GPA; (2) the use of SAT scores alone or with high school grades in determining admission is biased in favor of admitting underrepresented minorities and students who are socioeconomically disadvantaged; (3) SAT scores are more important and high school grades are less important in predicting GPA for underrepresented minorities and/or those students from low-income families than they are for those students who are white and/or from high-income families; and (4) the University of California found ways to admit a significant number of underrepresented minorities despite many of them having low SAT scores.
Abstractor: As Provided
Entry Date: 2024
Accession Number: EJ1425105
Database: ERIC
Full text is not displayed to guests.
FullText Links:
  – Type: pdflink
    Url: https://content.ebscohost.com/cds/retrieve?content=AQICAHj0k_4E0hTGH8RJwT4gCJyBsGNe_WN95AvKlDbXJGqwxwFZ5Ejz2nxGJtdJcBVS-fd0AAAA4zCB4AYJKoZIhvcNAQcGoIHSMIHPAgEAMIHJBgkqhkiG9w0BBwEwHgYJYIZIAWUDBAEuMBEEDLzJJm0cJnSH5ASMdQIBEICBm19zUgs19mOd3IFfTro4wzq9xtThvIToWq-X-k656QSm_b0kZUMlHY5Zf-JP5Ey9dYAnmU8SmtlylSoHiAYa8XVUEk0HLN0FARg6ezscvbTfAw5crG85SKXC-gD4eoGoXMYKG5EeyPhn6JY_vthQdY9A2D30naZEfe_RyAPufRNLoSiVpjBFKvs-D92A7moWKY1Emz3NWgoQjIlB
Text:
  Availability: 1
  Value: <anid>AN0177337602;ems01jun.24;2024May22.04:42;v2.2.500</anid> <title id="AN0177337602-1">The University of California Was Wrong to Abolish the SAT: Admissions When Affirmative Action Was Banned </title> <p>I study student characteristics and academic performance at the University of California, where consideration of an applicant's ethnicity has been banned since 1996 and SAT scores were used in admitting students to the university until fall 2021. I show the following: (<reflink idref="bib1" id="ref1">1</reflink>) SAT scores were more important than high school grades in predicting first‐year university GPA; (<reflink idref="bib2" id="ref2">2</reflink>) the use of SAT scores alone or with high school grades in determining admission is biased in favor of admitting underrepresented minorities and students who are socioeconomically disadvantaged; (<reflink idref="bib3" id="ref3">3</reflink>) SAT scores are more important and high school grades are less important in predicting GPA for underrepresented minorities and/or those students from low‐income families than they are for those students who are white and/or from high‐income families; and (<reflink idref="bib4" id="ref4">4</reflink>) the University of California found ways to admit a significant number of underrepresented minorities despite many of them having low SAT scores.</p> <p>Keywords: affirmative action; HSGPA; Proposition 209; SAT; underrepresented minorities</p> <p>In November 2021, the University of California Regents voted unanimously to end the use of SAT and ACT scores for admission. Although the official statement regarding the policy change was vague, the Regents seem to agree with arguments presented at an earlier meeting that SAT and ACT scores were biased against low‐income groups and that there was only a modest increase in explained variation of first‐year GPA when adding SAT or ACT scores to high school GPA (HSGPA) as an explanatory variable.</p> <p>I show that the arguments against the SAT are wrong and that there is a great misunderstanding regarding how students were actually admitted to the University of California, and presumably elsewhere. In particular, I show the following: (<reflink idref="bib1" id="ref5">1</reflink>) SAT scores were more important than high school grades in predicting performance at the University of California. (<reflink idref="bib2" id="ref6">2</reflink>) The use of SAT scores in admission shows a bias in <emph>favor</emph> of students who come from socioeconomically disadvantaged (SED) families. (<reflink idref="bib3" id="ref7">3</reflink>) SAT scores for SED students are even more important in predicting relative performance than they are for students coming from socioeconomically advantaged families, while the reverse is true for high school grades. (<reflink idref="bib4" id="ref8">4</reflink>) The University of California admitted a great many socioeconomically disadvantaged students when they had low SAT scores and low high school grade‐point averages by either implicitly or explicitly giving points for being SED. That is, the University of California gave preference to those who faced greater challenges in gaining the tools needed for college because of bias or other circumstances.</p> <p>In 1996, voters in California passed Proposition 209, which prohibited state governmental institutions from considering race, sex, or ethnicity in public education and other governmental areas. Nevertheless, as will be shown in Section 4, a significant number of underrepresented minorities were admitted despite having low SAT scores. The recent Supreme Court decision outlaws affirmative action in universities that receive public funds. The admission policy at the University of California before 2021 serves as a case study for those universities that want their student body to more closely mirror the population.</p> <p>The article is organized along the following lines: The four arguments above comprise four different sections. The relevant literature and method used are presented in each section. The first section on the relative importance of SAT scores is a survey of previous work. The results are well‐known to researchers in the field, but not so much elsewhere. Later sections shift the focus to new results.</p> <hd id="AN0177337602-2">Data</hd> <p>These results rely on the following data: I have grades for every student in every class taken from fall 2009 through summer 2013 for a total of 440,000 observations at the University of California Santa Cruz (UCSC). I also have SAT scores and high school GPA (HSGPA) for everyone who entered as a freshman from fall term 2009 through fall term 2012. Most of the results are based on first‐year performance from 2009 to 2013. This includes 13,325 students taking approximately 8 courses on average, resulting in 103,700 observations.</p> <hd id="AN0177337602-3">Method and Results</hd> <p></p> <hd id="AN0177337602-4">SAT Scores Are Better than HSGPA for Predicting University of California Grades</hd> <p>Proponents of abolishing the SAT commonly argue that the explained variation (R<sups>2</sups>) of first‐year college GPA (fyGPA) does not sufficiently increase when first‐year GPA is regressed against both SAT and HSGPA instead of HSGPA alone. However, at the University of California, this is not at all the case as can be seen by looking at Table 1.</p> <p>1 Table Explained Variation of First‐Year GPA (in Percentages)</p> <p> <ephtml> <table><thead><tr><th>Year</th><th>2001</th><th>2005</th><th>2007</th><th>2012</th><th>2015</th><th>2019</th><th>2019</th></tr></thead><tbody><tr><td>HSGPA alone</td><td>17</td><td>18</td><td>20</td><td>15</td><td>13</td><td>13</td><td>23</td></tr><tr><td>SAT alone</td><td>17</td><td>19</td><td>21</td><td>22</td><td>22</td><td>21</td><td>28</td></tr><tr><td>HSGPA + SAT</td><td>22</td><td>27</td><td>28</td><td>27</td><td>26</td><td>26</td><td>36</td></tr><tr><td>(Row 3)/(Row 1)</td><td>1.29</td><td>1.50</td><td>1.40</td><td>1.80</td><td>2.00</td><td>2.00</td><td>1.57</td></tr></tbody></table> </ephtml> </p> <p>1 <emph>Note</emph>: The data for 2001–2015 are from the University of California Office of Institutional Research ([<reflink idref="bib25" id="ref9">25</reflink>]), table 3. The data for 2019 are from Kurlaender ([<reflink idref="bib13" id="ref10">13</reflink>]) who reported correlations. The first 2019 column is produced in the same way as the preceding columns. The last column controls for the restriction of range. The SAT value represents the total score. The SAT changed format in 2005 and 2016. During that time, a writing component was included.</p> <p>Starting with the second to the last column, HSGPA alone explained 13% of the variance in first‐year GPA; SAT scores alone explained 21% of the variance in first‐year GPA; and when added together, HSGPA and SAT explained 26% of the variability in first‐year GPA. So, adding SAT scores to HSGPA doubled the explained variation of first‐year college GPA (last row). Even in 2001, the increase in explained variation from adding SAT scores to HSGPA was very large—29%.</p> <p>Suppose that the University of California only accepted students who had perfect SAT scores (800 on each part). Then SAT scores would not explain any of the variation in GPA. While not so dramatic, the admission process reduces the range of both SAT scores and HSGPA because students with low HSGPA and/or low SAT scores are <emph>less</emph> likely to be admitted or even apply in the first place. The numbers in the last column try to correct for this restriction in range (presumably on both the low end and the high end, as students may choose to go elsewhere even if admitted). While the absolute increase in explanatory power when adding SAT to HSGPA remains the same as in the previous column, the percentage change is smaller but still very large (36/23 −1 = .57). There is evidence that the increase in explanatory power when adding SAT is less for less selective universities and colleges. For example, Kurlaender and Cohen ([<reflink idref="bib14" id="ref11">14</reflink>]), when accounting for the restriction of range, found that the increase in explanatory power when adding SAT to HSGPA was 14% for the California State University system.</p> <hd id="AN0177337602-5">Admissions Based on SAT Scores Are Biased in Favor of Students Who Are SED</hd> <p>The impetus for the University of California dropping the SAT was the Compton Unified School District lawsuit that accused the University of being biased against students from low‐income households, racial and ethnic minorities, and/or students with disabilities by requiring the SAT and ACT in the admissions process.</p> <p>The first question to ask is how to determine bias. The answer is that using SAT scores alone or in conjunction with HSGPA for admission is biased against social‐economically disadvantaged students if they on average perform better at the university than predicted. This methodology is often called the Cleary ([<reflink idref="bib5" id="ref12">5</reflink>]) model of predictive bias. The most commonly used measure of college performance in the literature is first‐year GPA (fyGPA) and that is what I will be using here. More formally, one determines whether using the SAT in admissions is biased against those students who are SED by running the following regressions: 1 <ephtml> <math display="block" altimg="urn:x-wiley:07311745:media:emip12598:emip12598-math-0001" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><mrow><mi>fyGPA</mi><mspace width="0.33em" /></mrow><mo linebreak="badbreak">=</mo><msub><mi>B</mi><mn>10</mn></msub><mspace width="0.33em" /><mo linebreak="goodbreak">+</mo><msub><mi>B</mi><mn>11</mn></msub><mi>SAT</mi><mo linebreak="goodbreak">+</mo><msub><mi>B</mi><mn>12</mn></msub><mi>SED</mi><mo>,</mo></mrow><annotation encoding="application/x-tex">$$\begin{equation}{\mathrm{fyGPA\ }} = {B}_{10}\ + {B}_{11}{\mathrm{SAT}} + {B}_{12}{\mathrm{SED}},\end{equation}$$</annotation></semantics></math> </ephtml> 2 <ephtml> <math display="block" altimg="urn:x-wiley:07311745:media:emip12598:emip12598-math-0002" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mrow><mrow><mi>fyGPA</mi><mspace width="0.33em" /></mrow><mo linebreak="badbreak">=</mo><msub><mi>B</mi><mn>20</mn></msub><mspace width="0.33em" /><mo linebreak="goodbreak">+</mo><msub><mi>B</mi><mn>21</mn></msub><mi>SAT</mi><mo linebreak="goodbreak">+</mo><msub><mi>B</mi><mn>22</mn></msub><mi>SED</mi><mo linebreak="goodbreak">+</mo><msub><mi>B</mi><mn>23</mn></msub><mi>HSGPA</mi><mo>.</mo></mrow><annotation encoding="application/x-tex">$$\begin{equation}{\mathrm{fyGPA\ }} = {B}_{20}\ + {B}_{21}{\mathrm{SAT}} + {B}_{22}{\mathrm{SED}} + {B}_{23}{\mathrm{HSGPA}}.\end{equation}$$</annotation></semantics></math> </ephtml></p> <p>If the coefficient of SED is positive, then those students who are SED will tend to have a higher fyGPA than other students who have the same SAT score (and HSGPA in the second equation) but are not SED. If the coefficient of SED is negative, then there is bias against students who are not SED when applying to the university. Let us stick with the latter case for the remainder of this paragraph. The true expected fyGPA requires that SED be included in the regression and then using all of the regression coefficients for prediction. If the university does this and sets a fixed expected fyGPA value that students must obtain for admission, then there is no bias, because SED and non‐SED students with the <emph>same</emph> expected fyGPA will be treated the <emph>same</emph> way by the admission committee. But if the university does not use the variable SED in the regression equation for admission, which is actually the case, then admission is biased against students who are not SED because they will do better than predicted, while students who are SED will do worse. The same logic holds for determining whether there is bias regarding the admission of underrepresented minorities or cisgender females.</p> <p>This is the standard used to prove bias in other realms, as well. Consider the U.S. Equal Employment Opportunity Commission Section 10, where a variety of scenarios are presented. For example, if men and women are in the <emph>same</emph> job classification and there are no other <emph>relevant</emph> differences and men make more than women, then there is bias against women. Section 10 also considers previous steps in the employment ladder, such as being assigned to a higher classification. If men and women have the <emph>same</emph> qualifications for the higher classification and men are more likely to be advanced to the higher classification, then there is bias (discrimination) against women. That is, the key to ascertaining employment bias is to compare men and women who are otherwise the <emph>same</emph> but are treated differently. The words "similar," "equal," and "same," respectively, appear 20, 107, and 93 times in the 5‐page document. Note that if women do not have the qualifications needed for the higher classification and as a consequence there are relatively few women employed in the higher classification, that is not evidence that the firm is engaged in biased behavior, although bias in society may account for women having fewer qualifications.</p> <p>There is a long and varied empirical literature that regresses college performance against SAT scores, HSGPA, and socioeconomic characteristics. Depending on the data available to the researcher, some articles focus on underrepresented minorities, others on the characteristics of the student's family, and still others on the nature of the high school that the student attended. Ramist et al. ([<reflink idref="bib18" id="ref13">18</reflink>]), in a study of 28 colleges, found that relative to white students, the prediction of fyGPA was underpredicted for Asian students and overpredicted for Black and Latinx students. Geiser and Studley ([<reflink idref="bib9" id="ref14">9</reflink>]), making use of University of California data from 1996 through 1999, found that relative to white students, fyGPA was overpredicted for Black, Latinx, and Asian students. Rothstein ([<reflink idref="bib19" id="ref15">19</reflink>]) made use of a long list of variables, including the ethnicity and gender of the student attending the University of California, as well as characteristics of the student's high school such as percent of underrepresented minorities, percent of students getting free lunch and average parental income. With the exception of the free‐lunch variable (possibly due to multicollinearity), all the results showed that fyGPA was overpredicted for those students who were underrepresented minorities and/or who attended high schools where many of the other students were underrepresented minorities, and/or average parental income was low. Kurlaender ([<reflink idref="bib13" id="ref16">13</reflink>]) made use of two variables SED and HS‐CCI. Kurlaender's definition of SED was that the student either qualified for a reduced‐price school lunch program or did not have a parent who graduated from high school. HS‐CCI is a California Department of Education index based on the percentage of high school graduates who are prepared for <bold>c</bold>ollege or a <bold>c</bold>areer (hence the double CC). The larger HS‐CCI is, the stronger the high school. While not explicitly stated in Kurlaender's Table 1, the coefficient of SED is negative and the coefficient of HS‐CCI is positive when regressing fyGPA against SAT, HSGPA, SED, and HS‐CCI. Again, this shows that admission is biased in favor of social‐economically disadvantaged students when SAT scores are used. I note that I am not interested in showing that there is bias against those who are who are white or wealthy. It is just a powerful way to show that using the SAT does not result in admission bias against those who are SED.</p> <p>Here, I will be using another measure of SED: being eligible for the Educational Opportunity Program (EOP). "EOP provides assistance through mentorship, academic programs, financial assistance, counseling/advising, and other campus support services to those who are first‐generation college students, and/or from low‐income and educationally disadvantaged backgrounds" ([<reflink idref="bib24" id="ref17">24</reflink>].).</p> <p>I will also use a different measure of fyGPA. It is well known that grading standards vary across courses. For example, STEM courses are, in general, less generous than grading in social sciences courses (Achen & Courant, 2009; Tomkin & West, [<reflink idref="bib22" id="ref18">22</reflink>]). The problem in using plain fyGPA as a measure of academic success can be seen by considering the following example. Suppose that every course is graded on a curve with B being set as the median grade. If everyone takes course 1 and the top half of the students who received an A take course 2 and the bottom half of the students who received a C take course 3, then there is a downward bias in GPA for those in the upper half of the students in class 1 when they take classes 1 and 2 and an upward bias for students at the bottom half when they take courses 1 and 3. All of this is relevant here because there is a tendency for academically weaker students to take courses where the grading is more generous (Wittman, [<reflink idref="bib27" id="ref19">27</reflink>]), thereby reducing the measured influence of SAT and HSGPA on fyGPA.</p> <p>These issues are accounted for by jointly estimating <emph>course fixed effects</emph> (cfe) and <emph>student fixed effects</emph> (sfe). This is equivalent to a two‐way ANOVA for students and courses. The use of fixed effects in predicting grades is not unique to this article. Kurlander used campus fixed effects; in addition, Rothstein used department fixed effects. Arcidiacono et al. ([<reflink idref="bib2" id="ref20">2</reflink>]) also used department fixed effects in their study of performance at Duke. For those who are unacquainted with these procedures, think of these as adjusting a student's grade upward when the courses the student takes are filled with academically competitive students who tend to get higher grades than the average grade in their other courses and the faculty members teaching these academically strong students do not adjust their grading upward. To keep these ideas in mind, I have created the acronym, fyGPAsfe. At times, I also use plain fyGPA to highlight the interesting differences that arise when these different measures are employed.</p> <p>Table 2 presents the results for six regressions. For now, I focus on regressions 1, 2, and 5. In each of these regressions, the coefficient of EOP is negative and congruent with the literature. The evidence is clear: regardless of how SED is measured, whether the regression includes HSGPA, and whether fyGPA or fyGPAsfe is used, the coefficients of SED are negative. The logic is also clear—a negative SED coefficient means that predictions of fyGPA or fyGPAsfe that do not include SED will overpredict fyGPA and fyGPAsfe for students who are SED and will underpredict fyGPA for students who are not.</p> <p>2 Table FyGPAsfe Is Overpredicted for Those Who Are EOP</p> <p> <ephtml> <table><tbody><tr><td>(1) fyGPAsfe = 1.31 +1.092 SAT − .146 EOP</td><td>Adj. R<sup>2</sup> = .177</td></tr><tr><td>(2) fyGPAsfe = −1.019 + .988 SAT + .688 HSGPA − .104 EOP</td><td>Adj. R<sup>2</sup> = .246</td></tr><tr><td>(3) fyGPAsfe = −1.290 + 1.099 SAT + .702 HSGPA</td><td>Adj. R<sup>2</sup> = .243</td></tr><tr><td>(4) fyGPAsfe = −.090 + .871 HSGPA</td><td>Adj. R<sup>2</sup> = .116</td></tr><tr><td>(5) fyGPAsfe = −.429 + .765 HSGPA − .362 EOP</td><td>Adj. R<sup>2</sup> = .173</td></tr><tr><td>(6) fyGPAsfe = .994 + 1.252 SAT</td><td>Adj. R<sup>2</sup> = .170</td></tr></tbody></table> </ephtml> </p> <p>2 <emph>Note</emph>: SAT is the SAT score divided by 1,000. Stata reports all coefficients having <emph>P</emph> >| <emph>t</emph> | = 0.000 with the exception of the intercept in regression 3. Only those students with a reported HSGPA (not being homeschooled) are included. The number of observations in Table 2 is 13,235.</p> <p>It is useful to consider evidence and arguments that run counter to the above. Geiser ([<reflink idref="bib7" id="ref21">7</reflink>]) spends a number of pages arguing that the SAT is unintentionally biased because (<reflink idref="bib1" id="ref22">1</reflink>) the SAT may be written in such a way that is harder for minority students to understand since terms and issues used in the SAT may not be common in their culture, and (<reflink idref="bib2" id="ref23">2</reflink>) African Americans face "stereotype threat" in that they get nervous and distracted when they take the SAT because the stereotype is that African Americans do poorly on such exams. A third argument is that the SAT is biased against students from low‐income families because they are less likely to take SAT preparation courses or retake the exam. With regard to the latter, Goodman et al. ([<reflink idref="bib10" id="ref24">10</reflink>]) found that, on average, students' scores improved by nearly 90 points of out 2,400 when they retook the SAT. A fourth argument is that very wealthy parents may hire substitute SAT takers (the Varsity Blues scandal). It is possible that all four have an impact in the suggested direction.</p> <p>But how important are these biases quantitatively? Ultimately, the question here is whether these biases against minorities and the socioeconomically disadvantaged in admissions are larger than the biases in their favor. If the answer is yes, then the <emph>net</emph> bias against minorities should show up in the regressions discussed in the above paragraphs. That is, underrepresented minorities should on average have a higher fyGPA or fyGPAsfe than predicted by the regressions. However, as already seen, regression results show that high school students from economically disadvantaged families tend to have a lower, not higher fyGPA and fyGPAsfe than predicted.</p> <p>A different line of attack has been undertaken by Geiser ([<reflink idref="bib8" id="ref25">8</reflink>]). Since much of it is inspired by Rothstein's ([<reflink idref="bib19" id="ref26">19</reflink>]) article mentioned earlier, it is enlightening to review his article in greater detail as it is both carefully executed and important. Rothstein was concerned that sparse prediction models (employing only SAT scores and HSGPA) would miss important factors determining fyGPA. So, he made the regression models richer by including student and high school demographic variables. To simplify, Rothstein added variables for high school quality (HS‐CCI) and variables for socioeconomic status (SES). SES is basically the opposite of SED, but SES is not restricted to 2 levels. His results show that these demographic characteristics add predictability regarding first‐year performance at the University of California, and, at the same time, reduce the importance of SAT scores (both the SAT coefficient and its contribution to explained variation are reduced). Similar results can be seen by comparing the coefficients of SAT in regressions 2 and 3 in Table 2. Regression 2 adds EOP to regression 3 and the coefficients of both SAT and HSGPA are reduced, but the latter not by as much as the former. This is an important result because it suggests that there are other important factors such as high school quality or socioeconomic status that influence student performance at the university. In turn, this suggests that society might, for example, want to spend more resources on low HS‐CCI high schools.</p> <p>But what does this say about admission to the University of California? First, recall that being SED is basically the opposite of being high SES. I consider four options. The first option is to use the second regression instead of the third regression in Table 2. This would reduce the role of the SAT by about 10% but would penalize students from SED backgrounds. Here is Rothstein's view: "[My research] indicates that admissions offices could admit better‐prepared entering classes by giving explicit admissions preferences to high‐SES students and to students from high‐SES high schools. SAT scores would receive some weight in 'best predictor' admissions rules, but considerably less than is indicated by sparse models. Few would advocate this sort of admissions rule, which might be called 'affirmative action for high SES children,' and even fewer would consider it meritocratic." Thus, we have eliminated regression 2 in Table 2, from guiding admission to the University.</p> <p>Let us go to the opposite extreme, which more closely resembles, but not exactly, the present‐day admission policy at the University of California. Geiser ([<reflink idref="bib8" id="ref27">8</reflink>]) claims that using only SAT and HSGPA (regression 3) has an omitted variable problem as it omits SES (negative SED) in regression 2. But the last thing that he would want to do is use the regression results when SES is a factor for admission (regression 2). Instead, he wants to omit SAT scores, as well (resulting in regression 4), which increases omitted variable bias. Furthermore, employing the same logic as Geiser used against the SAT, the University of California should not use HSGPA alone for admission (regression 4) because when you add EOP and get regression 5, the coefficient of HSGPA decreases by a larger percentage than SAT did when adding EOP to regression 3 (for further arguments about the relationship between SES and HSGPA, see Zwick & Green, [<reflink idref="bib30" id="ref28">30</reflink>]; Zwick, [<reflink idref="bib29" id="ref29">29</reflink>])</p> <p>Next, let us briefly consider some so far unmentioned alternatives. The California Smarter Balanced Assessment test, promoted by Kurlaender et al. ([<reflink idref="bib15" id="ref30">15</reflink>]), was also shot down by the Regents as well as by STTF ([<reflink idref="bib21" id="ref31">21</reflink>]). Many of the other alternatives were already employed along with SAT scores. These include students' personal statements and taking Advance Placement Exams. The question is how good they are in predicting academic success, whether they are uncorrelated with SES, and are they better as substitutes for SAT scores than they are as complements. Substituting these alternatives for SAT scores (that is, adding them to Regression 4) is likely to be a lot less useful in predicting fyGPA than using them as complements to SAT scores (that is, adding them to Regression 3). With regard to the issue of bias against minorities, it is less clear whether the bias in admission in favor of SED observed in Regression 2 will increase or decrease when adding these alternatives to Regression 2. However, in Section 4, we will see that regardless of the regression chosen, there is an easy way to increase the bias in favor of those who are SED.</p> <p>It is true that SAT scores are positively correlated with socioeconomic status (SES), but high school graduation rates (Wodtke et al., [<reflink idref="bib28" id="ref32">28</reflink>]), high school grade‐point average (HSGPA), math, reading, and science achievement scores (Hanushek et al., [<reflink idref="bib11" id="ref33">11</reflink>]), and even cognitive skills of children entering kindergarten are positively correlated with SES (Lee & Burkam, [<reflink idref="bib16" id="ref34">16</reflink>]). And more relevant, as already shown, grades in college are negatively correlated with SED even after controlling for SAT scores and HSGPA. That is, even with identical HSGPA and SAT scores, the student with a higher socioeconomic status background will have higher college grades than a student with a lower socioeconomic status background. And because admissions criteria justifiably do not treat lower SES negatively in their decisions regarding whom to admit, this means that <emph>the admission process is biased in favor of students with lower socioeconomic status</emph>. SAT scores just mirror the disadvantages that disadvantaged groups have in doing well at the University of California and presumably elsewhere. To be clear, the disadvantages that underrepresented minorities face in doing well at the University of California are very likely due to the challenging environment they faced before arriving at the university and possibly never fully remediated. In turn, these challenges may be due to underlying biases toward minorities.</p> <p>It is important to understand what this article says and what it does not say.</p> <p></p> <ulist> <item> The article does not have anything to say about the existence or extent of bias that exists precollege although the observation that underrepresented minorities often go to low‐performing high schools is consistent with their underperforming at the university.</item> <p></p> <item> The article does not provide any data that would explain why underrepresented minorities have lower fyGPA and fyGPAsfe than predicted by their HSGPA and SAT scores.</item> <p></p> <item> Across time and across universities, empirical studies have shown that students from low SES families and underrepresented minorities on average have lower grades than predicted by SAT scores alone or in conjunction with HSGPA. There are many possible reasons for this, including lower rates of preschool attendance and weaker high schools attended. On the other side of the equation, the design of the SAT may, purposely or not, be biased in favor of these groups. In turn, this means that the underperformance may no longer be the case if the underlying causes change.</item> <p></p> <item> Expected first‐year GPA (fyGPA) is the most common criterion for judging whether a student should be admitted to the university. The criterion itself maybe biased one way or another. This is not easy to assess and the Cleary method is of no use in determining whether the criterion is biased.</item> </ulist> <p>It is quite possible that some readers are not convinced by the evidence I have presented in this section and continue to believe that SAT scores are biased against socioeconomically disadvantaged and underrepresented minorities. It turns out that even if such a belief is correct or that one believes that there are other values to achieve than maximizing expected academic performance, there is a much better solution than getting rid of the SAT, but the reader will have to wait until the fourth section to find out the answer.</p> <hd id="AN0177337602-6">SAT Scores Are More Predictive for SED than They Are for Those Not SED</hd> <p>In this section, I show that SAT scores are more predictive and HSGPA is less predictive of first‐year college performance for social‐economically disadvantaged students and underrepresented minorities than they are for social‐economically advantaged students.</p> <p>I introduce the following variables: EOP × SAT, EOP × HSGPA, LATIN × SAT, LATIN × HSGPA, ASIAN × SAT, ASIAN × HSGPA, BLACK× SAT, and BLACK × HSGPA. Each of these variables represents products. Regressing fyGPAsfe on EOP × SAT determines the <emph>marginal</emph> or interactive effect of SAT on fyGPAsfe from being EOP eligible.</p> <p>I now turn to the expanded multiple regressions in Table 3. I first focus on the fyGPAsfe plus columns. Looking at those variables ending in SAT, the marginal effect of SAT scores on fyGPAsfe is positive for LATINX, ASIAN, and BLACK students relative to WHITE students (but the coefficient of BLACK × SAT is not significant). That is, for each of these groups, SAT is relatively more important in predicting fyGPAsfe than it is for being WHITE. For example, the predicted increase in fyGPAsfe from a 1,000‐point <emph>increase</emph> in a Latinx student's SAT score (.625 +.466) will be greater than the predicted <emph>increase</emph> in fyGPAsfe from a 1,000‐point increase in a white student's SAT score (.625). In contrast, the marginal effect of HSGPA on fyGPAsfe is negative for LATINX, ASIAN, and BLACK relative to WHITE, meaning HSGPA is less important for these groups in predicting fyGPAsfe, but still positive overall. For example, an increase in HSGPA by 1 results in an expected increase in fyGPAsfe equal to (.786 −.06) for a LATINX student, which is less than it is for a white student (.786). The same pattern holds for EOP versus non‐EOP students. The marginal effect of SAT scores on fyGPAsfe is positive for EOP eligible students relative to non‐EOP eligible students. Similarly, the marginal effect of HSGPA is less when the student is EOP eligible than when the student is not EOP eligible. For EOP, the intercept is shifted downward; however, the coefficient is not at all significant, possibly due to multicollinearity of EOP with BLACK and LATINX.</p> <p>3 Table Regressions When There Are Additive and Multiplicative Effects</p> <p> <ephtml> <table><thead><tr><th>Variables</th><th>fyGPA</th><th>fyGPAsfe</th><th>fyGPAsfe plus</th></tr><tr><th /><th>Coefficient</th><th><italic>T</italic></th><th>Coefficient</th><th><italic>T</italic></th><th>Coefficient</th><th><italic>T</italic></th></tr></thead><tbody><tr><td>SAT</td><td>1.029</td><td>34.63</td><td>.954</td><td>32.32</td><td>.625</td><td>12.23</td></tr><tr><td>HSGPA</td><td>.591</td><td>28.74</td><td>.696</td><td>34.06</td><td>.786</td><td>23.57</td></tr><tr><td>EOP</td><td>−.074</td><td>−5.15</td><td>−.080</td><td>−5.57</td><td>−.048</td><td>−.25</td></tr><tr><td>EOP × SAT</td><td /><td /><td /><td /><td>.275</td><td>4.25</td></tr><tr><td>HSGPA × EOP</td><td /><td /><td /><td /><td>−.129</td><td>−2.78</td></tr><tr><td>LATINX</td><td>−.074</td><td>−4.5</td><td>−.097</td><td>−5.98</td><td>−.635</td><td>−2.8</td></tr><tr><td>LATIN × SAT</td><td /><td /><td /><td /><td>.466</td><td>5.95</td></tr><tr><td>LATIN × HSGPA</td><td /><td /><td /><td /><td>−.060</td><td>−1.08</td></tr><tr><td>ASIAN</td><td>−.045</td><td>−3.3</td><td>.067</td><td>4.65</td><td>.037</td><td>.16</td></tr><tr><td>ASIAN × SAT</td><td /><td /><td /><td /><td>. 153</td><td>2.15</td></tr><tr><td>ASIAN × HSGPA</td><td /><td /><td /><td /><td>−.069</td><td>−1.31</td></tr><tr><td>BLACK</td><td>−.054</td><td>−1.66</td><td>−.090</td><td>−2.76</td><td>.627</td><td>1.44</td></tr><tr><td>BLACK × SAT</td><td /><td /><td /><td /><td>.152</td><td>1.06</td></tr><tr><td>BLACK × HSGPA</td><td /><td /><td /><td /><td>−.281</td><td>−2.45</td></tr><tr><td>MALE</td><td>−.107</td><td>−9.28</td><td>−.016</td><td>−1.40</td><td>−.012</td><td>−1.07</td></tr><tr><td>CONSTANT</td><td>−.967</td><td>−10.82</td><td>−.978</td><td>−11.00</td><td>−.733</td><td>−5.02</td></tr><tr><td>Adj. R<sup>2</sup></td><td>.2446</td><td /><td>.2596</td><td /><td>.2662</td><td /></tr></tbody></table> </ephtml> </p> <p>3 <emph>Note</emph>: Compound names are multiplied. Cisgender is included to more closely match the regressions in Rothstein ([<reflink idref="bib19" id="ref35">19</reflink>]). The number of observations in each case is 12,549.</p> <p>In a nutshell, SAT scores are relatively more predictive of university performance and HSGPA is relatively less predictive of university performance for these ethnic minorities and EOP students than for their white and non‐EOP counterparts.</p> <p>With coefficients being both positive and negative, it is hard to determine the net effect of being a minority or being classified as being EOP eligible. Therefore, I have provided the fyGPAsfe columns, which provide the average net effect of being an EOP eligible student and/or being a particular minority. Given the data, the predicted first‐year GPAsfe when the student is EOP and/or Black or Latinx will <emph>on average</emph> be less than their non‐EOP and/or white counterparts.</p> <p>It is insightful to also look at the fyGPA column. For the most part, the coefficients are roughly similar to their fyGPAsfe counterparts. There are two striking differences. First, the coefficient of ASIAN turns from negative to positive (both coefficients are significantly different from zero) when considering fyGPAsfe instead of fyGPA. The reason for this change is that Asian students are more likely than white students to take STEM courses that grade lower. fyGPAsfe takes this into account. The second notable difference is that the coefficient of MALE is about 1/7 as large and less significant when fyGPAsfe is the dependent variable. The reason is the same as above: MALES are more likely to take STEM courses. With regard to the coefficients of EOP, LATINX, and BLACK, they become more negative and more significant when fyGPAsfe is employed, suggesting that, relative to WHITE and/or non‐EOP eligible students, these groups tend to take courses where the grading is more generous. It is well recognized that individual differences in college course choice reduce the measured influence of HSGPA and SAT scores on college GPA. See for example, Berry and Sackett ([<reflink idref="bib3" id="ref36">3</reflink>]) who look at cross‐sectional data, and Wittman ([<reflink idref="bib27" id="ref37">27</reflink>]) who looks at UCSC data.</p> <p>Another way to understand the importance of SAT scores for EOP eligible students is to go back to the first section and see how much the explained variation (adjusted <emph>R</emph><sups>2</sups>) in fyGPA increases when SAT scores are added to HSGPA, but now dividing the entering class into those students who are EOP eligible and those students who are not. More technically, we are comparing the following ratio when EOP = 1 and the number of observations is 4,920 to when EOP = 0 and the number of observations is 8,315: <ephtml> <math display="block" altimg="urn:x-wiley:07311745:media:emip12598:emip12598-math-0003" xmlns="http://www.w3.org/1998/Math/MathML"><semantics><mfrac><mrow><mrow><mi>e</mi><mi>x</mi><mi>p</mi><mi>l</mi><mi>a</mi><mi>i</mi><mi>n</mi><mi>e</mi><mi>d</mi><mspace width="0.33em" /><mi>v</mi><mi>a</mi><mi>r</mi><mi>i</mi><mi>a</mi><mi>t</mi><mi>i</mi><mi>o</mi><mi>n</mi><mspace width="0.33em" /><mi>o</mi><mi>f</mi><mspace width="0.33em" /><mi>f</mi><mi>y</mi><mi>G</mi><mi>P</mi><mi>A</mi><mspace width="0.33em" /><mi>w</mi><mi>h</mi><mi>e</mi><mi>n</mi><mspace width="0.33em" /><mi>S</mi><mi>A</mi><mi>T</mi><mspace width="0.33em" /><mi>a</mi><mi>n</mi><mi>d</mi><mspace width="0.33em" /><mi>H</mi><mi>S</mi><mi>G</mi><mi>P</mi><mi>A</mi></mrow><mspace width="0.33em" /><mrow><mi>a</mi><mi>r</mi><mi>e</mi><mspace width="0.33em" /><mi>t</mi><mi>h</mi><mi>e</mi><mspace width="0.33em" /><mi>i</mi><mi>n</mi><mi>d</mi><mi>e</mi><mi>p</mi><mi>e</mi><mi>n</mi><mi>d</mi><mi>e</mi><mi>n</mi><mi>t</mi><mspace width="0.33em" /><mi>v</mi><mi>a</mi><mi>r</mi><mi>i</mi><mi>a</mi><mi>b</mi><mi>l</mi><mi>e</mi><mi>s</mi></mrow></mrow><mrow><mrow><mi>e</mi><mi>x</mi><mi>p</mi><mi>l</mi><mi>a</mi><mi>i</mi><mi>n</mi><mi>e</mi><mi>d</mi><mspace width="0.33em" /><mi>v</mi><mi>a</mi><mi>r</mi><mi>i</mi><mi>a</mi><mi>t</mi><mi>i</mi><mi>o</mi><mi>n</mi><mspace width="0.33em" /><mi>o</mi><mi>f</mi><mspace width="0.33em" /><mi>f</mi><mi>y</mi><mi>G</mi><mi>P</mi><mi>A</mi><mspace width="0.33em" /><mi>w</mi><mi>h</mi><mi>e</mi><mi>n</mi><mspace width="0.33em" /><mi>H</mi><mi>S</mi><mi>G</mi><mi>P</mi><mi>A</mi><mspace width="0.33em" /><mi>i</mi><mi>s</mi></mrow><mspace width="0.33em" /><mrow><mi>t</mi><mi>h</mi><mi>e</mi><mspace width="0.33em" /><mi>i</mi><mi>n</mi><mi>d</mi><mi>e</mi><mi>p</mi><mi>e</mi><mi>n</mi><mi>d</mi><mi>e</mi><mi>n</mi><mi>t</mi><mspace width="0.33em" /><mi>v</mi><mi>a</mi><mi>r</mi><mi>i</mi><mi>a</mi><mi>b</mi><mi>l</mi><mi>e</mi></mrow></mrow></mfrac><annotation encoding="application/x-tex">$$\begin{equation*} \frac{{explained\ variation\ of\ fyGPA\ when\ SAT\ and\ HSGPA}\ {are\ the\ independent\ variables}}{{explained\ variation\ of\ fyGPA\ when\ HSGPA\ is}\ {the\ independent\ variable}}\end{equation*}$$</annotation></semantics></math> </ephtml></p> <p>The respective ratios are 3.18 and 1.56. That is, relative to HSGPA alone, SAT scores plus HSGPA are twice as important for EOP students in predicting fyGPA as they are for non‐EOP students.</p> <hd id="AN0177337602-7">How Students Were Admitted to the University of California before 2020</hd> <p>The University of California did make use of SAT scores and HSGPA, but even before 2020, other factors influenced the decision to admit a student, including the application essay, which was judged not only for the quality of the exposition, but also for what it said about the circumstances faced by the student. Every year, catalog copy said that the University wanted to promote diversity in admissions, and the admissions director for UCSC made positive statements about increased diversity in the incoming class and applauded becoming a Hispanic Serving Institution. The most detailed description of what the University was trying to achieve in admission is found in Table 4. Table 4 contains an online message to students about the criteria for admission and makes explicit what before had been implicit. Not surprisingly, academic performance and ability, including SAT scores and HSGPA, compromise the first four points. However, later points with italicized portions weight achievement with respect to the student's situation. For example, they make it possible for a student with lower HSGPA and lower SAT scores and who goes to a lower HS‐CCI school to be accepted instead of another student with higher HSGPA and SAT scores and who goes to a higher HS‐CCI school. "Eligibility in the local context" (ELC) means that a student is guaranteed admission to the University, but not to any particular campus if the student's HSGPA is in the top 9% in their high school even though the high school has a low HS‐CCI score.</p> <p>4 Table How Applications Are Reviewed</p> <p> <ephtml> <table><tbody><tr><td>... we look beyond grades and test scores. We spend time evaluating your academic achievements in light of the opportunities available to you ... The 14 factors we consider are: <list list-type="Bullet"><list-item><p>Academic grade point average in all completed A‐G courses, including ...</p></list-item><list-item><p>Scores on the following tests: ACT with Writing or the SAT with Essay.</p></list-item><list-item><p>Number of, content of and performance in academic courses beyond the minimum ...</p></list-item><list-item><p>Number of and performance in UC‐approved honors, Advanced Placement ...</p></list-item><list-item><p>Identification by UC as being ranked in the top 9% of your high school class at the end ofyour junior year (Eligible in the Local Context, or ELC).</p></list-item><list-item><p>Quality of your senior‐year program ...</p></list-item><list-item><p>Quality of your academic performance relative to the educational opportunities availablein your high school.</p></list-item><list-item><p>Outstanding performance in one or more specific subject areas.</p></list-item><list-item><p>Outstanding work in one or more special projects in any academic field of study.</p></list-item><list-item><p>Recent, marked improvement in academic performance</p></list-item><list-item><p>Special talents, achievements and awards in a particular field, ... experiences that demonstrate unusual promise for leadership, such as significant community service ...</p></list-item><list-item><p>Completion of special projects undertaken in the context of your high school curriculum ...</p></list-item><list-item><p>Academic accomplishments in light of your life experiences and special circumstances, including but not limited to: disabilities, low family income, first generation to attendcollege, need to work, disadvantaged social or educational environment, difficult personal and family situations or circumstances, refugee status or veteran status.</p></list-item><list-item><p>Location of your secondary school and residence.</p></list-item></list></td></tr></tbody></table> </ephtml> </p> <p>4 <emph>Note</emph>: The ellipses are employed to eliminate less important details. Italics represent student characteristics measured relative to the student's high school or group. Table 4 was downloaded from the Internet Archives Wayback Machine.</p> <p>There are many ways to determine whether a student is from a low SES background. This year's admission questions, like previous years' questions, asked about challenges faced and overcome, and, more specifically, there is a question about overcoming an educational barrier. The UC application asks students to report their family income and whether they will be the first in their immediate family to graduate from a 4‐year college. Low SES can also be determined by looking at the HS‐CCI index for the high school attended. There are other indicators available, such as percent of students who were eligible for a free lunch and whether the student lives in a low‐income census tract. The university provides information to those in charge of admissions about each student's performance relative to other students in the high school. In determining which applicants to admit, UCSC explicitly or implicitly gave (and presumably will continue to give) extra points for students who have attended academically weak high schools and/or were from SED backgrounds, thereby compensating to some degree for the fact that these students have faced a more challenging learning environment and/or bias. When a holistic approach (comprehensive review) is used in admissions, the process is less transparent, but the reviewers can put more weight on the challenges that a student has faced. In a nutshell, the University of California undertook a variety of ways to increase the number of SED students and/or underrepresented minorities admitted to the University even under the previous system where SAT scores entered into the decision and admission on the basis of race or ethnicity was prohibited by Proposition 209.</p> <p>The data used in this section are based on UCSC, but there is little reason to believe that the other University of California campuses did not engage in similar practices, albeit at different levels. At the top end of the applicant pool, there were a significant number of students from disadvantaged backgrounds who were admitted because their HSGPA and SAT scores were superior. But this was not enough to make the percentage of UCSC students who were from socioeconomically disadvantaged families and/or who had gone to academically weak high schools and/or who were members of minorities that were historically underrepresented at UCSC to closely mirror the percentage of Californians with these characteristics. So, UCSC targeted students who had one or more of these three overlapping characteristics. Admission committees implicitly or explicitly weighted low socioeconomic status as a positive factor for admission. The more points that a student received for being low SES, the more the student's SAT deficit was overcome relative to those who were high SES. In principle, any disadvantages underrepresented minorities and/or low SES students faced before college could be counter‐balanced by these admission weights in their favor. Pre‐2020, the university was <emph>capable</emph> of admitting those students from low SES backgrounds who were most likely to be academically successful by considering both their HSGPA and SAT scores. Similarly, among high SES students, the most likely to be academically successful could also be admitted.</p> <p>For those who think that groups were not targeted, that the number of students involved was trivial, or that SAT scores must have been high in order to be admitted to UCSC, Table 5 may alter your preconceptions. Table 5 shows that EOP students comprised 38% of those freshmen who identified as being white, Latinx, Asian, or Black. Thirty‐nine percent of the incoming class were white (and at times being EOP eligible) and 29% were Latinx (and at times not being EOP eligible). The best forensic evidence that admissions were able to overcome low SAT scores is to compare the low SAT column to the high SAT column and the different student compositions. For example, only 6.4% of students with SAT scores below 1,450 were white, while 63.4% of students with SAT scores above 1,845 were white.</p> <p>5 Table Characteristics of the First‐Year Classes between 2009 and 2012</p> <p> <ephtml> <table><thead><tr><th align="center">Percent</th></tr><tr><th /><th>First year</th><th>SAT < 1,450</th><th>SAT > 1,845</th><th>HSGPA < 3.4</th><th>HSGPA > 3.84</th></tr></thead><tbody><tr><td>WHITE</td><td>39.3</td><td>6.4</td><td>63.4</td><td>26.8</td><td>52.2</td></tr><tr><td>LATINX</td><td>29.1</td><td>61.7</td><td>9.4</td><td>38.4</td><td>23.7</td></tr><tr><td>ASIAN</td><td>28.2.</td><td>26.6</td><td>25.6</td><td>29.9</td><td>22.0</td></tr><tr><td>BLACK</td><td>3.4</td><td>5.3</td><td>1.6</td><td>4.8</td><td>2.1</td></tr><tr><td>EOP</td><td>38.0</td><td>83.5</td><td>10.3</td><td>52.3</td><td>30.0</td></tr></tbody></table> </ephtml> </p> <p>5 <emph>Note</emph>: Each of the last four columns encompasses about 21% of the first‐year students. The SAT was worth 2400 during this time</p> <p>For students whose SAT scores were less than 1,450, the mean SAT score at UCSC was 1,298. An SAT score of 1,298 places the individual in the 27th percentile of all students who took the test. The mean SAT score for those who had an SAT score greater than 1,845 is 1,960. An SAT score of 1,960 places the individual in the 91st percentile. The mean HSGPA for those who had an HSGPA less than 3.4 is 3.23. The mean HSGPA for those who had an HSGPA greater than 3.84 is 4.01.</p> <p>While there is considerable variation across the University of California campuses in their willingness to pursue a student body more reflective of California's diverse population, it is clear that neither Proposition 209, which does not allow admission to be based on ethnic grounds, nor the SAT prevented a significant proportion of underrepresented minorities from gaining admission to the university, albeit still not mirroring the general population. If the University of California Regents wanted more students from disadvantaged backgrounds, they should not have eliminated the SAT but instead instructed some of the campuses to admit more SED even if this required the campuses to admit students with lower SAT scores. To illustrate this last sentence, let us consider eligibility in the local context (ELC). In practice, the more selective campuses only admitted a few ELC students and many of those would have been admitted anyway. The least selective campus was Merced. But for those students who were only accepted by Merced, a significant proportion decided to go elsewhere outside the University of California system. The Regents could have instructed each campus to accept either a certain percentage of ELC students who applied or a certain percentage of students who were eligible for a free lunch (even if not ELC), but in each case allowing SAT scores to influence the campus decision. And if that did not result in sufficient numbers, the Regents could have established a goal for matriculation by students who were SED.</p> <hd id="AN0177337602-8">Discussion</hd> <p>Some people argue that a university's student body should reflect the proportion of minorities and/or low‐income individuals in society, while other people argue that only those who will be academically most successful should be admitted. This article is completely neutral with regard to such arguments. What this article does is show the best way to achieve either of these goals or anywhere in between. Contrary to what many believe, this article has shown that SAT scores are an important predictor of academic success for all groups, but especially so for underrepresented minorities.</p> <p>The Supreme Court of the United States has determined that university affirmative action programs are not constitutional. California is a case study on what this will mean. California Proposition 209 essentially outlawed admissions to state colleges and universities based on race, ethnicity, or gender. As I have shown in Section 4, the University of California, to a significant extent, mitigated this proposition by recruiting students that were socioeconomically disadvantaged, which is highly correlated, but by no means perfectly so, with being an underrepresented minority. Those states that want to increase the proportion of historically underrepresented minorities attending their universities can follow the University of California's lead in admitting students. But one thing that they should not do is follow the University of California by abolishing the use of the SAT or ACT.</p> <p>As a final remark, the usual caveats hold true for this article. I cannot guarantee the results presented here hold for all universities. Contrary results might arise in barely selective public universities or in law schools where college grades and LSAT scores are used for admissions. The only way that we will know for sure is to undertake similar empirical strategies for these cases.</p> <p>[Correction added on February 29, 2024, after first online publication: A sentence in the Discussion section has been corrected]</p> <hd id="AN0177337602-9">Conflict of Interest Statement</hd> <p>My work does not involve any conflict of interest.</p> <ref id="AN0177337602-10"> <title> Footnotes </title> <blist> <bibl id="bib1" idref="ref1" type="bt">1</bibl> <bibtext> Long ([17]) has argued that the University of California and other universities will have to stop saying that they want to achieve ethnic diversity and instead say that they are trying to achieve economic diversity.</bibtext> </blist> <blist> <bibl id="bib2" idref="ref2" type="bt">2</bibl> <bibtext> The assumptions made to account for the restriction of range are questionable. For example, the calculations assume that if a university starts to admit academically weak students, many will fail. However, the likely outcome is that faculty would lower their standards for passing the course. Few faculties would want to flunk a much higher proportion of their students. See Rothstein ([19]) for another critique.</bibtext> </blist> <blist> <bibl id="bib3" idref="ref3" type="bt">3</bibl> <bibtext> Westrick et al.'s ([26]) study of 171 colleges included 28% that accepted above 75% of applicants and 43% that accepted between 51% and 75% of applicants. This mix may explain why, accounting for the restriction of range, their reported increase was on average 15%.</bibtext> </blist> <blist> <bibl id="bib4" idref="ref4" type="bt">4</bibl> <bibtext> However, SAT scores continue to explain more of the variation in fyGPAsfe than does HSGPA, as do a large number of fyGPA studies reported in Sackett and Kuncel ([20]).</bibtext> </blist> <blist> <bibl id="bib5" idref="ref12" type="bt">5</bibl> <bibtext> Although K, R & R were advocating the Smarter Balanced Assessment test as a substitute for the SAT, it might have been even better as a complement to the SAT.</bibtext> </blist> <blist> <bibl id="bib6" type="bt">6</bibl> <bibtext> There is a large literature devoted to explaining the SAT's bias against cisgender females. See Keiser et al. ([12]) and a meta‐analysis by Fischer et al. ([6]).</bibtext> </blist> <blist> <bibl id="bib7" idref="ref21" type="bt">7</bibl> <bibtext> Table S3 in Kurlaender ([13]) considers similar ratios for SED and not SED, but Kurlaender's table S3 is not comparable to the ratios considered here because she controls for HS‐CCI. Nevertheless, the ratio of the explained variation of SAT plus HSGPA over the explained variation of HSGPA alone is also larger for SED than non‐SED, but the difference is not as dramatic.</bibtext> </blist> <blist> <bibl id="bib8" idref="ref25" type="bt">8</bibl> <bibtext> Individual campuses each had their own way of determining whom to accept. Each campus was required to review all applications from high school students with a 3.00 or better HSGPA. The main difference post‐2020 is eliminating the SAT.</bibtext> </blist> <blist> <bibl id="bib9" idref="ref14" type="bt">9</bibl> <bibtext> One possible reason is that minority students at the top end of the distribution chose to go elsewhere.</bibtext> </blist> <blist> <bibtext> In determining HSGPA, the University of California gives additional points for high school advanced placement courses. The University of California System is premised on accepting the top 12.5% of graduating high school seniors, but exceptions are allowed.</bibtext> </blist> <blist> <bibtext> Bleemer ([4]) provides evidence that expanding ELC from 4% to 9% had little effect on the number of underrepresented minorities at the University of California.</bibtext> </blist> </ref> <ref id="AN0177337602-11"> <title> References </title> <blist> <bibtext> Achen, A., & Courant, P. (2009). What are grades made of. Journal of Economic Perspectives, 23 (3), 77 – 92.</bibtext> </blist> <blist> <bibtext> Arcidiacono, P., Aucejo, E. M., & Spenner, K. (2012). What happens after enrollment? An analysis of the time path of racial differences in GPA and major choice. IZA Journal of Labor Economics, https://izajole.springeropen.com/articles/10.1186/2193‐8997‐1‐5</bibtext> </blist> <blist> <bibtext> Berry, C. M., & Sackett, P. R. (2009). Individual differences in course choice result in underestimation of the validity of college admissions systems. Psychological Science, 20 (7), 822 – 830. https://doi.org/10.1111/j.1467‐9280.2009.02368.x</bibtext> </blist> <blist> <bibtext> Bleemer, Z. (2023). Affirmative action and its race‐neutral alternatives. Journal of Public Economics, 220, 104839. https://doi.org/10.1016/j.jpubeco.2023.10483</bibtext> </blist> <blist> <bibtext> Cleary, T. A. (1968). Test bias: Prediction of grades of Negro and white students in integrated colleges. Journal of Educational Measurement, 5 (2), 115 – 124.</bibtext> </blist> <blist> <bibtext> Fischer, F. T., Schult, J., & Hell, B. (2013). Sex‐specific differential prediction of college admission tests: A meta‐analysis. Journal of Educational Psychology, 105 (2), 478 – 488.</bibtext> </blist> <blist> <bibtext> Geiser, S. (2017). Norm‐referenced test and race blind admissions. In Joseph Soares (Ed.), The Scandal of standardized tests: Why we need to drop the SAT & ACT. Teachers College Press.</bibtext> </blist> <blist> <bibtext> Geiser, S. (2020). SAT/ACT scores, high‐school GPA and the problem of omitted variable bias: Why the UC taskforce's findings are spurious. UC Berkeley Center for Studies in Higher Education. https://cshe.berkeley.edu/publications/satact‐scores‐high‐school‐gpa‐and‐problem‐omitted‐variable‐bias‐why‐uc‐taskforce's</bibtext> </blist> <blist> <bibtext> Geiser, S., & Studley, R. (2002). UC and the SAT. Predictive validity and differential impact of the SAT I and SAT II at the University of California. Educational Assessment, 8 (1), 1 – 25</bibtext> </blist> <blist> <bibtext> Goodman, J., Gurantz, O., & Smith, J. (2020). Take two! SAT retaking and college enrollment gaps. American Economic Journal: Economic Policy, 12 (2), 115 – 158.</bibtext> </blist> <blist> <bibtext> Hanushek, E., Light, J., Peterson, P. E., Talpey, L. M., & Woessmann, L. (2022). Long‐run trends in the U.S. SES achievement gap. Education Finance and Policy, 17 (4), 608 – 640.</bibtext> </blist> <blist> <bibtext> Keiser, H. N., Sackett, P. R., Kuncel, N. R., & Brothen, T. (2016). Why women perform better in college than admission scores would predict: Exploring the roles of conscientiousness and course‐taking patterns. Journal of Applied Psychology, 101 (4), 569 – 581. https://doi.org/10.1037/apl0000069</bibtext> </blist> <blist> <bibtext> Kurlaender, M. (2019) Predicting college success: How do different high school assessments measure up? Supplemental Tables. Policy Analysis for California Education (PACE). https://edpolicyinca.org/sites/default/files/R_Supplemental_Kurlaender_Mar19.pdf</bibtext> </blist> <blist> <bibtext> Kurlaender, M., & Cohen, K. (2019). Predicting college success: How do different high school assessments measure up. Policy Analysis for California Education (PACE). https://edpolicyinca.org/publications/predicting‐college‐success‐how‐do‐different‐high‐school‐assessments‐measure‐2019</bibtext> </blist> <blist> <bibtext> Kurlaender, M., Reber, S., & Rothstein, J. (2020) UC Regents should consider all evidence and options in decision on admissions policy. Policy Analysis for California Education (PACE). https://edpolicyinca.org/newsroom/uc‐regents‐should‐consider‐all‐evidence‐and‐options‐decision‐admissionspolicy</bibtext> </blist> <blist> <bibtext> Lee, V., & Burkam, D. (2002). Inequality at the starting gate: Social background differences in achievement as children begin school. Economic Policy Institute. Washington D.C.</bibtext> </blist> <blist> <bibtext> Long, M. C. (2015). Is there a "workable" race‐neutral alternative to affirmative action in college admissions? Journal of Policy Analysis and Management, 34 (1), 162 – 183.</bibtext> </blist> <blist> <bibtext> Ramist, L., C. Lewis, & L. McCamley‐Jenkins (1994). Student group differences in predicting college grades: Sex, language, and ethnic groups. College Board Report No. 93‐1.</bibtext> </blist> <blist> <bibtext> Rothstein, J. M. (2004). College performance predictions and the SAT. Journal of Econometrics, 121 (1–2), 297 – 317. https://doi.org/10.1016/j.jeconom.2003.10.003</bibtext> </blist> <blist> <bibtext> Sackett, P. R., & Kuncel, N. R. (2018). Eight myths about standardized testing. In J. Buckley, L. Letukas, & B. Wildavsky (Eds.), Measuring success: Tests, grades, and the future of admissions testing (pp. 13 – 39). Baltimore, MD : Johns Hopkins University Press.</bibtext> </blist> <blist> <bibtext> Standardized Testing Task Force, STTF. (2020). University of California systemwide academic senate report on standardized testing. https://senate.universityofcalifornia.edu/_files/committees/sttf/sttf‐report.pdf</bibtext> </blist> <blist> <bibtext> Tomkin, J. H. D., & West, M. (2022). STEM courses are harder: Evaluating inter‐course grading disparities with a calibrated GPA model. International Journal STEM Ed, 9 (1), Article 27. https://doi.org/10.1186/s40594‐022‐00343‐1</bibtext> </blist> <blist> <bibtext> U.S. Equal Employment Opportunity Commission. (n.d.). Section 10 Compensation Discrimination. https://<ulink href="http://www.eeoc.gov/laws/guidance/section‐10‐compensation‐discrimination#:~:text=The%20EPA%20requires%20employers%20to,need%20be%20only%20substantially%20equal%2C">www.eeoc.gov/laws/guidance/section‐10‐compensation‐discrimination#:~:text=The%20EPA%20requires%20employers%20to,need%20be%20only%20substantially%20equal%2C</ulink></bibtext> </blist> <blist> <bibtext> University of California Admissions. (n.d). https://admission.universityofcalifornia.edu/campuses‐majors/campus‐programs‐and‐support‐services/educational‐opportunity‐program‐eop.html#:~:text=The%20Educational%20Opportunity%20Program%20(EOP,income%20and%20educationally%20disadvantaged%20backgrounds)</bibtext> </blist> <blist> <bibtext> University of California Office of Institutional Research and Academic Planning. (2020). Relationship of the SAT/ACT to college performance at the University of California. https://<ulink href="http://www.ucop.edu/institutional‐research‐academic‐planning/%5ffiles/sat‐act‐study‐report.pdf">www.ucop.edu/institutional‐research‐academic‐planning/%5ffiles/sat‐act‐study‐report.pdf</ulink></bibtext> </blist> <blist> <bibtext> Westrick, P. J., Marini, L., Young, H., Ng, H., Shmueli, D., & Shaw, E. (2019). Validity of SAT for predicting first‐year grades and retention to the second year. College Board. https://satsuite.collegeboard.org/media/pdf/national‐sat‐validity‐study.pdf</bibtext> </blist> <blist> <bibtext> Wittman, D. (2022) Average rank and adjusted rank are better measures of college student success than GPA. Educational Measurement: Issues and Practice, 41 (4), 23 – 34. https://doi.org/10.1111/emip.1252</bibtext> </blist> <blist> <bibtext> Wodtke, G. T., Harding, D. J., & Elwert, F. (2011). Neighborhood effects in temporal perspective: The impact of long‐term exposure to concentrated disadvantage on high school graduation. American Sociological Review, 76 (5), 713 – 736.</bibtext> </blist> <blist> <bibtext> Zwick, R. (2013). Disentangling the role of high school grades, SAT® scores, and SES in Predicting College Achievement. ETS Research Report, 13, 1 – 10. https://<ulink href="http://www.ets.org/Media/Research/pdf/RR‐13‐09.pdf">www.ets.org/Media/Research/pdf/RR‐13‐09.pdf</ulink></bibtext> </blist> <blist> <bibtext> Zwick, R., & Green, J. G. (2007). New perspectives on the correlation of SAT scores, high school grades, and socioeconomic factors. Journal of Educational Measurement, 44 (1), 23 – 45.</bibtext> </blist> </ref> <aug> <p>By Donald Wittman</p> <p>Reported by Author</p> </aug> <nolink nlid="nl1" bibid="bib25" firstref="ref9"></nolink> <nolink nlid="nl2" bibid="bib13" firstref="ref10"></nolink> <nolink nlid="nl3" bibid="bib14" firstref="ref11"></nolink> <nolink nlid="nl4" bibid="bib18" firstref="ref13"></nolink> <nolink nlid="nl5" bibid="bib19" firstref="ref15"></nolink> <nolink nlid="nl6" bibid="bib24" firstref="ref17"></nolink> <nolink nlid="nl7" bibid="bib22" firstref="ref18"></nolink> <nolink nlid="nl8" bibid="bib27" firstref="ref19"></nolink> <nolink nlid="nl9" bibid="bib10" firstref="ref24"></nolink> <nolink nlid="nl10" bibid="bib30" firstref="ref28"></nolink> <nolink nlid="nl11" bibid="bib29" firstref="ref29"></nolink> <nolink nlid="nl12" bibid="bib15" firstref="ref30"></nolink> <nolink nlid="nl13" bibid="bib21" firstref="ref31"></nolink> <nolink nlid="nl14" bibid="bib28" firstref="ref32"></nolink> <nolink nlid="nl15" bibid="bib11" firstref="ref33"></nolink> <nolink nlid="nl16" bibid="bib16" firstref="ref34"></nolink>
Header DbId: eric
DbLabel: ERIC
An: EJ1425105
AccessLevel: 3
PubType: Academic Journal
PubTypeId: academicJournal
PreciseRelevancyScore: 0
IllustrationInfo
Items – Name: Title
  Label: Title
  Group: Ti
  Data: The University of California Was Wrong to Abolish the SAT: Admissions When Affirmative Action Was Banned
– Name: Language
  Label: Language
  Group: Lang
  Data: English
– Name: Author
  Label: Authors
  Group: Au
  Data: <searchLink fieldCode="AR" term="%22Donald+Wittman%22">Donald Wittman</searchLink> (ORCID <externalLink term="https://orcid.org/0000-0002-1073-4345">0000-0002-1073-4345</externalLink>)
– Name: TitleSource
  Label: Source
  Group: Src
  Data: <searchLink fieldCode="SO" term="%22Educational+Measurement%3A+Issues+and+Practice%22"><i>Educational Measurement: Issues and Practice</i></searchLink>. 2024 43(2):55-63.
– Name: Avail
  Label: Availability
  Group: Avail
  Data: Wiley. Available from: John Wiley & Sons, Inc. 111 River Street, Hoboken, NJ 07030. Tel: 800-835-6770; e-mail: cs-journals@wiley.com; Web site: https://www.wiley.com/en-us
– Name: PeerReviewed
  Label: Peer Reviewed
  Group: SrcInfo
  Data: Y
– Name: Pages
  Label: Page Count
  Group: Src
  Data: 9
– Name: DatePubCY
  Label: Publication Date
  Group: Date
  Data: 2024
– Name: TypeDocument
  Label: Document Type
  Group: TypDoc
  Data: Journal Articles<br />Information Analyses
– Name: Audience
  Label: Education Level
  Group: Audnce
  Data: <searchLink fieldCode="EL" term="%22Higher+Education%22">Higher Education</searchLink><br /><searchLink fieldCode="EL" term="%22Postsecondary+Education%22">Postsecondary Education</searchLink><br /><searchLink fieldCode="EL" term="%22High+Schools%22">High Schools</searchLink><br /><searchLink fieldCode="EL" term="%22Secondary+Education%22">Secondary Education</searchLink>
– Name: Subject
  Label: Descriptors
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22College+Entrance+Examinations%22">College Entrance Examinations</searchLink><br /><searchLink fieldCode="DE" term="%22Admission+Criteria%22">Admission Criteria</searchLink><br /><searchLink fieldCode="DE" term="%22Grade+Point+Average%22">Grade Point Average</searchLink><br /><searchLink fieldCode="DE" term="%22Disproportionate+Representation%22">Disproportionate Representation</searchLink><br /><searchLink fieldCode="DE" term="%22Socioeconomic+Status%22">Socioeconomic Status</searchLink><br /><searchLink fieldCode="DE" term="%22Disadvantaged%22">Disadvantaged</searchLink><br /><searchLink fieldCode="DE" term="%22Affirmative+Action%22">Affirmative Action</searchLink><br /><searchLink fieldCode="DE" term="%22Minority+Group+Students%22">Minority Group Students</searchLink><br /><searchLink fieldCode="DE" term="%22Equal+Education%22">Equal Education</searchLink><br /><searchLink fieldCode="DE" term="%22Educational+Policy%22">Educational Policy</searchLink><br /><searchLink fieldCode="DE" term="%22Trustees%22">Trustees</searchLink><br /><searchLink fieldCode="DE" term="%22Scores%22">Scores</searchLink><br /><searchLink fieldCode="DE" term="%22Data+Analysis%22">Data Analysis</searchLink><br /><searchLink fieldCode="DE" term="%22High+School+Graduates%22">High School Graduates</searchLink><br /><searchLink fieldCode="DE" term="%22Grades+%28Scholastic%29%22">Grades (Scholastic)</searchLink><br /><searchLink fieldCode="DE" term="%22Grade+Prediction%22">Grade Prediction</searchLink>
– Name: Subject
  Label: Geographic Terms
  Group: Su
  Data: <searchLink fieldCode="DE" term="%22California%22">California</searchLink>
– Name: SubjectThesaurus
  Label: Assessment and Survey Identifiers
  Group: Su
  Data: <searchLink fieldCode="SU" term="%22SAT+%28College+Admission+Test%29%22">SAT (College Admission Test)</searchLink>
– Name: DOI
  Label: DOI
  Group: ID
  Data: 10.1111/emip.12598
– Name: ISSN
  Label: ISSN
  Group: ISSN
  Data: 0731-1745<br />1745-3992
– Name: Abstract
  Label: Abstract
  Group: Ab
  Data: I study student characteristics and academic performance at the University of California, where consideration of an applicant's ethnicity has been banned since 1996 and SAT scores were used in admitting students to the university until fall 2021. I show the following: (1) SAT scores were more important than high school grades in predicting first-year university GPA; (2) the use of SAT scores alone or with high school grades in determining admission is biased in favor of admitting underrepresented minorities and students who are socioeconomically disadvantaged; (3) SAT scores are more important and high school grades are less important in predicting GPA for underrepresented minorities and/or those students from low-income families than they are for those students who are white and/or from high-income families; and (4) the University of California found ways to admit a significant number of underrepresented minorities despite many of them having low SAT scores.
– Name: AbstractInfo
  Label: Abstractor
  Group: Ab
  Data: As Provided
– Name: DateEntry
  Label: Entry Date
  Group: Date
  Data: 2024
– Name: AN
  Label: Accession Number
  Group: ID
  Data: EJ1425105
PLink https://search.ebscohost.com/login.aspx?direct=true&site=eds-live&db=eric&AN=EJ1425105
RecordInfo BibRecord:
  BibEntity:
    Identifiers:
      – Type: doi
        Value: 10.1111/emip.12598
    Languages:
      – Text: English
    PhysicalDescription:
      Pagination:
        PageCount: 9
        StartPage: 55
    Subjects:
      – SubjectFull: College Entrance Examinations
        Type: general
      – SubjectFull: Admission Criteria
        Type: general
      – SubjectFull: Grade Point Average
        Type: general
      – SubjectFull: Disproportionate Representation
        Type: general
      – SubjectFull: Socioeconomic Status
        Type: general
      – SubjectFull: Disadvantaged
        Type: general
      – SubjectFull: Affirmative Action
        Type: general
      – SubjectFull: Minority Group Students
        Type: general
      – SubjectFull: Equal Education
        Type: general
      – SubjectFull: Educational Policy
        Type: general
      – SubjectFull: Trustees
        Type: general
      – SubjectFull: Scores
        Type: general
      – SubjectFull: Data Analysis
        Type: general
      – SubjectFull: High School Graduates
        Type: general
      – SubjectFull: Grades (Scholastic)
        Type: general
      – SubjectFull: Grade Prediction
        Type: general
      – SubjectFull: California
        Type: general
      – SubjectFull: SAT (College Admission Test)
        Type: general
    Titles:
      – TitleFull: The University of California Was Wrong to Abolish the SAT: Admissions When Affirmative Action Was Banned
        Type: main
  BibRelationships:
    HasContributorRelationships:
      – PersonEntity:
          Name:
            NameFull: Donald Wittman
    IsPartOfRelationships:
      – BibEntity:
          Dates:
            – D: 01
              M: 06
              Type: published
              Y: 2024
          Identifiers:
            – Type: issn-print
              Value: 0731-1745
            – Type: issn-electronic
              Value: 1745-3992
          Numbering:
            – Type: volume
              Value: 43
            – Type: issue
              Value: 2
          Titles:
            – TitleFull: Educational Measurement: Issues and Practice
              Type: main
ResultId 1