How Do Admission Preferences Affect Minority Student Outcomes? (March, 1999)

Summary

This study traces the impact of preferential admissions for targeted minorities on their admission, retention, and graduation rates, examining the impact of eliminating such preferences, and explores the “perceived” effect of these preferences on non-targeted applicants.¹If preferential admission for targeted minorities was eliminated, here is what would likely happen:

the enrollment of targeted minority freshmen would immediately decline by 30 percent;
the gap in second-year retention rates for targeted minorities would decrease by half after the first year, and the gap in six-year graduation rates for targeted minorities would decrease by almost half in six years;
if existing diversity resources were redirected to support the now smaller number of targeted minority students, minority retention and graduation rates would be further increased, and these gaps might even be eliminated before the year 2008;
if this happy event were to occur, the number of targeted minorities graduating from UW-Madison would equal the current number who are graduating despite the one-third decrease in the number targeted minorities admitted and enrolling here.

These findings emerge from an analysis of admission, retention, and graduation data for targeted minority and non-targeted applicants, and also for targeted minority and non-targeted students. The analysis is unique in that it focuses on how these rates vary by high school class rank, ACT score, and both high school class rank and ACT scores. The analysis, based on data from the UW-Madison Office of Budget, Planning, and Analysis, is restricted to those applicants for whom information is available on both their high school class rank and ACT scores.

The logic of these results is compelling. When the admission standard applicable to non-targeted applicants is applied to targeted minorities, the number of targeted minority applicants most at risk of not graduating is substantially reduced. As a result, those who are admitted would have greater chances of succeeding academically and, consequently, their retention and graduation rates would rise. Because many targeted minorities with good high school academic records often perform less well than do non-targeted students with similar academic records, current levels of diversity funding could now focus on helping this smaller group of minority students succeed in college. Indeed, the resulting one-third increase in per-student support offers the prospect of improving both retention and graduation rates for targeted minorities. The only question is how best to utilize the available resources to help these students.

Preferential admission for targeted minorities is not without some impact on non-targeted applicants and students. The extent to which non-targeted applicants would be displaced by targeted minority students is generally viewed as minimal. For example, if 130 targeted minority applicants are admitted when there is an overall limit on total enrollment, this means that 130 fewer non-targeted applicants would be admitted. These 130 displaced non-targeted applicants constitute less than 1.5 percent of all non-targeted applicants.

The policy of preferential admission also means that approximately 1,500 equally qualified non-targeted applicants would qualify for admission by the standard applied to targeted minority applicants. But, because the enrollment limit prevents them from being admitted, they may “perceive” they are being discriminated against in admission decisions. The number of non-targeted applicants in this category is larger, representing approximately 15 percent of all non-targeted applicants. Cumulated over several years, the potential negative effects of UW-Madison’s preferential admissions policy may outweigh its positive effects, and, in turn, undermine public support for the campus and perhaps jeopardize its budgetary support.

Introduction

The impact of preferential admissions on targeted minority students is extended in this study beyond that contained in the author’s “Do Minority Groups Get Admission Preference at UW-Madison.”² Rather than examining differences in admission rates between targeted minority and non-targeted applicants by high school class rank alone as that study does, this analysis looks at differences in admission rates by both high school class rank and ACT scores. The first part of the paper describes these differences. The second part examines the impact on enrollment, retention, and graduation of eliminating admission preferences for targeted minorities, and then explores several implications of eliminating preferential admissions.Before turning to the data, it is helpful to think about the likely patterns of admission by high school class rank and ACT scores for both non-targeted applicants and targeted minority applicants. Based on faculty-established admission standards, high school class rank takes clear priority in determining who is admitted. Hence, we would expect admission rates, except at the top high school rank intervals (e.g., those in the top 20 percent), to drop fairly sharply at each successive high school class rank interval. Admission rates by ACT scores, which are also used to help estimate for each applicant an expected first-year college GPA, would display a similar pattern. Since admission rates vary inversely with both high school class rank and ACT scores, we would expect a sharp drop off in admission rates at successively lower combinations of high school class rank and ACT scores. The likely pattern is shown in the top panel of Figure 1.

Because targeted minority applicants are admitted by a different standard, i.e., all minimally qualified targeted minority applicants are “normally admitted,” we would expect them to have high admission rates without regard to their high school class rank. This means that the admission rates for targeted minority applicants in the top half of their high school class would not vary much by high school class rank interval. Nor would much variation be expected in admission rates for targeted minority applicants by ACT score intervals, particularly for those applicants who graduated in the upper half of their high school class. Thus, rather than the sharp drop off expected for non-targeted applicants, admission rates for targeted minority applicants would remain high over the range of those minimally qualified (those in the top half of their high school class) and then drop off sharply for those targeted minority applicants not in the top half of their high school class. This pattern in shown in the bottom panel of Figure 1.

In conclusion, based on faculty-determined admission policy, we would expect to find quite different patterns in the admission rates of targeted minorities and non-targeted applicants. Specifically, admission rates for targeted minorities would generally exceed admission rates for non-targeted applicants except at higher class rank and ACT intervals. To test these expectations, we turn to the data.

Figure 1: Hypothesized Patterns of Admission Rates by High School Percentile Rank and ACT Composite Score

But, first, a description of the data. The analysis is restricted to those applicants whose records contain both class rank and ACT scores; excluded are applicants who submitted SAT scores (out of state applicants may opt to do this) and those for whom class rank or ACT scores were not available.³ The high school class rank intervals are by decile; the ACT score intervals of three points are standard in the literature. The data are analyzed for targeted minority applicants as a group, which includes Blacks, Native Americans, Hispanics, and SE Asians; and for non-targeted applicants as a group, which includes Whites and Asian Americans.With data on Fall 1997 applicants, it becomes possible to construct comparable admission rates for targeted minorities and the non-targeted population. With data for entering freshmen from Fall 1988-91, it becomes possible to construct comparable retention and graduation rates for both populations. To help readers understand the information used here, the initial analysis of admission rates is presented in considerable detail through a succession of charts and tables.

Looking at the Evidence

Admission Rates. Evidence of preferences in admission decisions comes from UW-Madison data for Fall 1997. These data permit calculating admission rates for applicants and admitted applicants by their high school percentile rank (abbreviated as high school class rank or class rank), by their ACT scores, and by class rank and ACT scores together. As background, the 82 percent overall admission rate for targeted minority applicants exceeds the 79 percent admission rate for non-targeted applicants.The percentages of targeted minority and non-targeted applicants admitted by high school class rank in Fall 1997 are shown in Figure 2; the results are quite similar to those based on somewhat different data that were presented in the author’s “Do Minority Groups Receive Admission Preferences at UW-Madison?” What is most apparent are the quite different patterns of admission rates for targeted minorities and for non-targeted applicants.

To elaborate, whereas the admission rates for non-targeted applicants drop substantially at each class rank interval below the 80th percentile, the admission rates for targeted minorities remain relatively high and drop off sharply but only below the 50th percentile. As might be expected, the admission rates for applicants in the top two class rank intervals (80+) are very similar. Oddly, this pattern continues but only for targeted minority applicants. By the 60-69 class rank interval, less than 40 percent of non-targeted applicants are admitted as contrasted to 93 percent for targeted applicants. The disparities are even greater at the 50-59 and the under 50 class rank intervals; in both intervals, the admission rates for targeted minorities exceed those for non-targeted applicants by a margin of roughly 5 to 1. The likelihood that these results occurred by chance is remote.

Figure 2: Percentage of Applicants Admitted by Class Rank, Fall 1997

Figure 3: Percentage of Applicants Admitted by ACT Scores, Fall 1997

These differences are not unexpected, as noted earlier. They arise because for non-targeted applicants, high school class rank is a key determinant of admission. By contrast, targeted applicants are admitted on a different basis. Considerable numbers are “normally admitted” even if they do not meet the “competitive” academic standard applied to non-targeted applicants, and other targeted minority applicants are admitted after being given “special consideration” on the basis of their race/ethnicity.⁴A different slant on how the admission process operates is provided by contrasting the percentages the of targeted minority and non-targeted applicants admitted by ACT score, as shown in Figure 3. Recall that ACT scores are used to help estimate a predicted first-year college GPA for each applicant, and this predicted value is used to rank applicants. The differences in admission rates shown by ACT scores are not quite as dramatic as those by high school class rank because the range of differences in scores is smaller for both groups. In another sense, these differences are even more dramatic, because at every ACT score interval targeted minorities are more likely to be admitted than are non-targeted applicants. Though these data are for only a single year, the likelihood is remote that the differences between these two population groups are the result of chance alone. Again, these patterns reflect the preferential admission status accorded targeted minorities.

These two perspectives can be combined by comparing admission rates by both high school class rank and ACT scores, as shown in Figure 4. This particular table shows admission rates for applicants with comparably high class rank and ACT scores, e.g., 90-100 for high school class rank and 30+ on the ACT, 80-89 for high school class rank and 27-29 on the ACT, etc. As might be expected, 100 percent of the top class rank-top ACT score groups (80+/27+) are admitted for both targeted minority and non-targeted applicants. At the next level (70-79 high school rank and 24-26 ACT), the admission rate for non-targeted applicants exceeds that for targeted applicants, i.e., 93 percent versus 72 percent. Below that level the gap in admission rates for non-targeted applicants becomes progressively wider except for applicants with the lowest class rank and ACT scores. The reason for the widening gap should be apparent: because all minimally qualified targeted minorities are “normally admitted” without regard to either their class rank or ACT scores, their admission rates exceed 85 percent in every interval except the lowest interval (<50/<18). Now that readers have a general understanding of these patterns, the data on admissions rates are presented in another more informative format. Table 1 shows admission rates for each cell in the 6 x 6 grid of high school class rank and ACT score intervals used in Figures 2-4. Admissions rates for targeted minority applicants appear in the top panel and those for non-targeted applicants appear in the bottom panel. The cells are shaded to highlight differences in the patterns of admission rates between the two groups.

Figure 4: Percentage of Applicants Admitted by High School Class Rank and ACT Scores, Fall 1997

As might be expected, admission rates for non-targeted applicants in the top panel are highest for applicants with excellent high school records and top ACT scores; these cells are identified by the darkest shading. Admission rates drop off sharply thereafter, with the rates exceeding 90 percent in only 1 of the 22 cells with entries below the 80th percentile class rank. The three cells with no entries, meaning there were no applicants with these combinations of class rank and ACT scores, are distinguished by their low ACT scores (under 18).The contrast with targeted minority applicants is dramatic. Admission rates for targeted minority applicants exceed 90 percent in 13 of the 23 cells with entries below the 80th percentile class rank. Expressed another way, admission rates exceed 80 percent in 26 of the 35 cells for targeted minority applicants as compared to only 14 of the 33 cells for non-targeted applicants.

The inescapable conclusion is this: these results confirm the speculations that prompted this analysis. One, targeted minority applicants are admitted on a different standard because of their race and ethnicity, as specified by faculty legislation. Two, the rules and procedures used by the Admissions Office to process freshmen applications for admission effectively implement the preferential admission policy established by faculty legislation.

Retention Rates. The analysis now shifts to describe second-year retention of freshmen from targeted minority and non-targeted groups. As noted, this analysis is based on data for freshmen who enrolled in the Fall semesters of 1988, 1989, 1990, and 1991. By averaging over several years, the analysis that follows is less affected by particular year-to-year variations. The format of this analysis follows that of the previous section but skips more quickly through Figures 5-7 which mirror Figures 2-4, and Tables 2 which mirrors Table 1.

Second-year retention rates by high school class rank for non-targeted students exceed those for targeted students across the entire spectrum of class rank intervals (Figure 5); the same is true for ACT intervals except at the top interval of 30+ (Figure 6). The pattern is somewhat mixed for students with comparable class rank and ACT scores (Figure 7). Here, retention rates for targeted minorities exceed those for non-targeted students in three of the six class rank/ACT intervals: 90-100/30+, 60-69/21-23, and <50/<18. What accounts for this pattern is not clear. Perhaps targeted minority students in these cells took fuller advantage of the opportunities open to them.

Table 1: Admission Rates for Fall 1997 Entering Freshmen Applicants, By High School Percentile Rank and Composite Score

Figure 5: Second-Year Retention Rate for Freshman Who Entered In Fall Semester 1988-1991 By High School Percentile Rank

Figure 6: Second-Year Retention Rate for Freshman Who Entered In Fall Semester 1988-1991 By ACT Score

Figure 7: Second-Year Retention Rate for Freshman Who Entered In Fall Semester 1988-1991 By High School Percentile Rank and Composite ACT Score

More illuminating is Table 2 which shows second-year retention rates by class rank-ACT cells. The retention pattern for non-targeted applicants is about what might be expected, with a high overall retention and with somewhat lower retention rates for students with lower class rank and lower ACT scores. By contrast, the pattern for targeted minorities is more variable. Some cells with high (90+) retention rates (shown by the dark shading) are next to cells with considerably lower retention rates (indicated by the lighter shading or in some cases by the absence of any shading). This pattern suggests that class rank and ACT scores are less accurate predictors of second-year retention for targeted minorities. Whether this pattern of retention reflects differences in academic performance, which these variables are designed to predict, or the effects of other personal and financial factors cannot be determined from these data. Graduation Rates. Similar data on six-year graduation of freshmen from targeted minority and non-targeted groups are available for these same freshmen who entered in the Fall semesters of 1988, 1989, 1990, and 1991. Again, the same pattern of presentation is followed.

Whether by high school class rank (Figure 8), ACT score (Figure 9), or a combination of the two measures (Figure 10), six-year graduation rates for non-targeted students exceed those for targeted students across the entire spectrum of intervals except for targeted minority students with the lowest combination of class rank and ACT scores; the graduation rate for this group exceeds by a wide margin that for the next higher class rank and ACT group. Close inspection of Table 3 shows that the difference arises in part because the graduation rate for this group of non-targeted students is so low.

Otherwise, the pattern of six-year graduation rates for non-targeted students is about what might be expected, with lower graduation rates at successively lower combinations of class rank and ACT scores. In 13 of the 36 cells, graduation rates exceed 70 percent, and in another 16 cells they exceed 51 percent. In only one cell, that for the lowest combination of class rank and test scores, is the graduation rate less than 35 percent.

The results for targeted minority students are more varied. The number of cells with graduation rates exceeding 70 percent is smaller (6 as compared to 13 for non-targeted students); the number of cells with rates exceeding 51 percent is also smaller (16 as contrasted to 29 for non-targeted students). Even more disturbing are the zero graduation rates in the high ACT cells where targeted minorities might be expected to do well, especially those with strong high school records. These results are also puzzling because some students in these same cells had high second-year retention rates. Again, these data by themselves cannot throw light on why graduation rates differ as they do, both between the two groups and within the targeted minority group.

Table 2: Second-Year Retention Rate for New Freshman Who Entered In The Fall Semesters of 1988-1991 By High School Percentile Rank and Composite ACT Score

Figure 8: Six-Year Graduation Rates for Freshmen Who Entered In Fall Semester 1988-1991 By High School Percentile Rank

Figure 9: Six-Year Graduation Rates for Freshmen Who Entered In Fall Semester 1988-1991 By Composite ACT Score

Figure 10: Six-Year Graduation Rates for Freshmen Who Entered In Fall Semester 1988-1991 By High School Percentile Rank and Composite ACT Score

Table 3: Six-Year Graduation Rates for Freshmen Who Entered In the Fall Semesters of 1988-1991, By High School Percentile Rank and Composite ACT Score

Elaborating on the Implications

Implications for Enrollment Rates. On average, considerable numbers of targeted minorities admitted to UW-Madison, though minimally qualified, are not academically competitive with the student body as a whole. On this basis alone, they face greater academic challenges than do their fellow targeted minority students who are admitted on the basis of their academic records.

If the same admission standard were applied to all applicants, without regard to their race and ethnic status, the number of targeted minority applicants admitted to UW-Madison, would decline by an estimated 30 percent,⁵ from about 430 to about 300 based the Fall 1997 data used here. The remaining 70 percent of targeted minorities would be admitted by the same standard of academic merit used to admit non-targeted applicants. While some of these targeted minority students might still face problems associated with their race and ethnicity, they as a group would be more likely to succeed academically and eventually graduate.
Applying the admission rate for non-targeted applicants to targeted minority applicants would reduce the admission rate for targeted minorities and reverse the gap in admission rates between the two groups. The admission rate for targeted minorities would fall from its current level of 82 percent to about 57 percent. And whereas the targeted minority admission rate currently exceeds the 79 percent rate for non-targeted applicants, it would now be substantially lower.
If, by contrast, the admission rates for targeted minority applicants were applied to non-targeted applicants, approximately 20 percent more non-targeted applicants would qualify for admission.⁶ Based on applicants who supplied information on both class rank and ACT scores, almost 1,600 additional non-targeted applicants would have been admissible. Of course, with the limit on enrollment imposed by the UW System’s Enrollment Management Policy, these additional non-targeted applicants could not be admitted. Thus, though the actual displacement effect of preferential admissions for targeted minorities is the 130 additional targeted minorities admitted (430-300), the “perceived” displacement effect as seen by non-targeted applicants is 1600. This figure, which represents more than 15 percent of all non-targeted applicants, can be viewed as the number of non-targeted applicants who might legitimately claim to be discriminated against in admission solely on the basis of their race/ethnicity.⁷ When cumulated over the years during which targeted minorities have received preference in admission, the number of people who may be disillusioned about UW-Madison’s admission policy could be substantial; what the effects of this ever growing group might be university budget requests and contributions to UW-Madison is not known.

To summarize, the data indicate that targeted minority applicants to UW-Madison benefit from race and ethnic-based preferences in admissions decisions. As a consequence, a substantial proportion of admitted applicants from targeted minority groups, though minimally qualified, are not academically competitive with admitted applicants from non-targeted groups. As a further consequence, substantial numbers of non-targeted applicants are denied admission even though they rank higher on two key determinants of admission, namely their high school class rank and their ACT Score. The evidence shows clearly that the University’s quest for a more diverse student body does entail some costs even though these costs cannot be defined or quantified with any precision.

This analysis also demonstrates that targeted minority applicants are evaluated on the basis of their group characteristics, i.e., race and ethnicity. The extent to which they may have been admitted because they as individuals “have been disadvantaged as a result of substandard education, family income level, or ethnic background” is not apparent. Even there were a desire to take these factors into account, how this would be done is not obvious. For example, family income cannot be an important consideration in admissions decisions because no information on family income is sought in the application form or available to the Admissions Office. Substandard education may be deduced by admissions personnel because of information available to them about the quality of an applicant’s high school. But, how substandard school quality may have affected any individual applicant’s academic performance at such a school is difficult to ascertain; invariably, in poor quality schools, some targeted minority students perform well and others perform less well.

Implications for Retention Rates. Differences in retention rates by the quantitative measures of academic merit used in making admissions decision are smaller than differences in admission rates based on these same measures. What seems surprising is how high the retention rates remain for students at the lower range of class rank and ACT scores. Moreover, second-year retention rates for targeted minority students are lower by class rank, ACT scores, and by various combinations of these measures, and thus they are at odds with the more favorable admission rates for targeted minority applicants.

If the same admission standard used to admit non-targeted applicants were applied to targeted minorities, and if targeted minority retention by class rank and ACT scores continued at current levels, the gap in retention rates would drop by 55 percent. Put another way, the retention rate for targeted minorities would rise from 81 to 87 percent, in contrast to the 92 percent figure for non-targeted students.
If targeted minorities admitted had to be academically competitive with non-targeted admitted applicants, and if the resources currently devoted to academic support programs for targeted minorities were focused on the academically competitive targeted minority students, it seems plausible to believe that the minority retention rate could be further improved and thereby help to close the remaining five percentage point gap between the two groups. Were this to happen, much might be learned about the effectiveness of different approaches to improving the academic performance of targeted minority students. And, such information might be helpful in sorting out the relative importance of prior academic achievement, campus climate, and other personal factors affecting minority retention.

Implications for Graduation Rates. The gap in six-year graduation rates is substantial. The overall graduation rate for targeted minorities is 49 percent, which is 24 percentage points below the 73 percent for non-targeted students.

If no preference in admissions were given to targeted minorities, the gap in six-year graduation rates would drop by more than two-fifths (actually, 42 percent). This means that the current six-year graduation rate of 49 percent would increase to 59 percent, still well below the 73 percent rate for non-targeted students.⁸
If this big an improvement can result from eliminating preferential admissions, should UW-Madison continue this policy? How can preferences be defended when the graduation rate data indicate that on average the chance that a targeted minority student will graduate within six years is less than the toss of a coin? Is this an acceptable basis for a university to admit what would clearly seem to be at-risk students? In fact, in 14 of the 29 class rank-ACT score cells for targeted minorities, as compared to only 7 out of the 36 cells for non-targeted students, the graduation rate is below 50 percent. Whether admitting applicants with such rates reflects for them a “reasonable probability of academic success” is open to questions. So also are the role of preferences in “implementing the University’s goals of maximizing the success [likelihood of graduation] of students who are admitted to UW-Madison and of achieving a heterogeneous and ethnically diverse student body.”
If concentrating academic support resources on academically competitive targeted minority students could increase their retention and graduation rates, the remaining gap in graduation rates might be narrowed or even eliminated. Any improvement would be beneficial, of course. In the process, something could be learned about the effectiveness of different approaches to this challenge.
If the gap in graduation rates could be eliminated, by bringing the graduation rate for targeted minorities up to that for non-targeted students, the number of targeted minority students graduating within six years would still equal the number now graduating under preferential admissions, despite a one-third reduction in the number of entering minority freshman.

This last finding is quite startling. While for many reasons it might be impossible to bring the targeted minority graduation rate up to that for non-targeted students, bringing these two rates much closer together would represent an impressive accomplishment. By focusing academic support resources on academically able targeted minority students and by removing the stigma minority students suffer as a result of preferential admissions, the possibility exists of bringing about a substantial narrowing in the graduation rate gap.What we don’t know is how effective this greater concentration of resources might be in making it possible for more targeted minority students to persist and graduate. Presumably, there are documented reports of successful support programs. These programs need to be examined to determine their relevance to UW-Madison and their potential success. It is not enough to assert that new and continuing support programs discussed in Madison Plan 2008 will work. We don’t really know whether they will work. What we do know is that such programs have not succeeded in the past. For these reasons, careful thought must be given to devising a strategy that will help minority students, who by virtue of their color and environment in which they grew up, have the ability to succeed, indeed, even to excel, in their undergraduate study at UW-Madison.

Conclusions

Wide gaps remain in new freshmen enrollment rates, second-year retention rates, and six-year graduation rates. These gaps persist despite a substantial and aggressive series of diversity programs implemented over the past several decades and a long history of preferences in admissions for targeted minority applicants.⁹Available data on applicants and those admitted by high school class rank and ACT scores reveals that targeted minority applicants are systematically treated more favorably than non-targeted applicants in admissions decisions. Except for the most academically able applicants, targeted minorities are admitted at appreciably higher rates than similarly qualified non-targeted applications. Such a result is not surprising because the UW-Madison faculty authorizes these preferences, which are then faithfully implemented by the Admissions Office.¹⁰ The impact of these preferences can be assessed in still another way; if targeted minority applicants were admitted at the same rate as non-targeted applicants, the number admitted would fall by 30 percent. This policy means that as many as 15 percent of all non-targeted applicants may have reason to “perceive” they are being displaced by because of race and ethnic based preferences.

The gaps in retention and graduation rates could be reduced by approximately 50 percent by admitting only those targeted minority applicants who are academically competitive with non-targeted applicants. If the resources currently devoted to diversity programs were then concentrated on the somewhat smaller number of more academically capable targeted minority students, the prospect of further narrowing, if not eliminating, the gaps in retention and graduation rates looks promising.

Finally, the UW-Madison community must consider the consequences its preferential admissions policy. How appropriate is it, from an educational and ethical standpoint, to admit to this campus substantial numbers of targeted minority applicants whose academic records indicate they have less than a 50-50 chance of graduating?