-
1.
REVISITING A SELECTIVE
STEM HIGH SCHOOL
ADMISSIONS POLICY:
7 YEARS LATER
Christopher Kolar
Carissa Rosine
NCSSS/NAGC Baltimore 2014
-
2.
The Revision
-
3.
Problems with TCS
• Verbal score weighed as
much as mathematics
• Skewed GPA (mean
admitted 3.77)
• 4-person review teams,
significant interrater-severity
effect
• Eligibility requirement of
“lab science” not
meaningfully verifiable
and discouraged
qualified applicants from
under-resourced districts
-
4.
Summary of changes
TCS: Ranked By 2006 Revision: Ordered By
Half of Class:
by formula
Second Half:
in/out decision
followed by
balancing
SAT-M
GPA
Review Committee
Be attentive to tilt!
-
5.
The Applicants
-
6.
Applicant Profile: Who is applying to IMSA?
2003-2008
• Total number of applicants steadily increases each year
• Slightly more male applicants than female
• Majority from Chicagoland area
• 3875 Applicants over 6 years
RACE PERCENT
White 42.9
Asian 36.0
Black 9.1
Hispanic 5.1
Multi-racial (2 or more) 4.4
Not reported 2.5
American Indian/Native Alaskan 0.1
-
7.
Applicant Profile: Who is applying to IMSA?
-
8.
The Admitted
-
9.
Applicants vs. Admitted Students
Pre Policy Change
AVERAGE APPLIED ADMITTED
GPA 3.69 3.86
SAT Score 1152 1242
SAT Math 598 646
SAT Critical Reading 554 596
RCE Score 62 66
Post Policy Change
AVERAGE APPLIED ADMITTED
GPA 3.70 3.88
SAT Score 1132 1237
SAT Math 590 646
SAT Critical Reading 543 591
RCE Score 60 66
-
10.
Race: Before and After
-
11.
Admitting students
-
12.
Matriculation
70
37
60
62
34
19
-
13.
Matriculation + SATM
70/
735
37/623
60/629
62/696
34/564
19/478
-
14.
Matriculation + Tilt
70/
47
37/130
60/-36
62/130
34/-10
19/13
-
15.
Matriculation + STEM Degree
70/
73
37/71
60/62
62/67
34/51
19/54
-
16.
The Outcomes I
-
17.
Some applicants are missing
Simulated
TCS run
showed that
Accept rate
had
decreased to
78% of “first
half”
-
18.
SIR Participation
Participation
in Student
Inquiry and
Research
increased,
with a
significant
bump in
SBES.
-
19.
Graduation Rate
-
20.
The Outcomes II
-
21.
Degree Profile
DEGREE FIELD PRE POLICY CHANGE
STEM 63.4
SBES 18.1
Professional 9.7
Humanities 6.6
Fine Arts 1.4
Unidentifiable 0.8
Percentage of students earning degrees in each field as reported by
National Student Clearinghouse.
-
22.
Degree Profile
DEGREE FIELD PRE POLICY
CHANGE
POST POLICY
CHANGE
STEM 63.4 69.0
SBES 18.1 15.3
Professional 9.7 7.8
Humanities 6.6 5.4
Fine Arts 1.4 1.4
Unidentifiable 0.8 1.0
Percentage of students earning degrees in each field as reported for 781
graduates by National Student Clearinghouse.
-
23.
The Conclusions
-
24.
Summary
Implementation matters
Increased student research participation, but
no increase in the STEM pipeline
This is a singular value decomposition of applicants from 5 dimensions to 2. From a policy perspective, we need to see how we can parse cluster 3 from cluster 2 “in the real world.”
The dark blue cluster on the far right represents students who are the generally “gifted” population of SAT scores over 1500. They are difficult to turn down and present themselves well because of their high verbal skills. Approximately 70% of the students in this group enter IMSA.
The light blue group in the lower right represent students with SATM scores, mean 696, as well as a tilt of 130 points over verbal. According to our redesigned admission policy, this is our target group: high achievement/potential in mathematics as well as a significant tilt away from verbal skills. The research indicates that students with these traits are the ones mostly likely to attain STEM outcomes in the long term. (Lubinski and Benbow). The rate at which applicants from this group matriculate is 62%.
The cluster in the upper right are students with above average SATM scores (per our applicant pool), who have a tilt on average of -36 in the direction of verbal/critical reading. These students present themselves in their applications very well at excel at communicating their vision to the review committees. Nevertheless, these students are supposed to be ones that we look at more closely – the research indicates that they are more like to pursue the social sciences or careers in fields such as law that allow them to apply their reasoning skills. While our revised application process is supposed to put greater scrutiny on this group, they are admitted at a rate of 60% -- almost the same as the high-tilt group. This means that we are not necessarily discounting a high overall SAT in order to preference mathematics aptitude. Indeed, it is very difficult to take a look at a student with a score of 1400 and make the case that they are not likely to be an ideal candidate for our program.
Finally, the gold cluster. This group consists of above average SATM scores, mean 623. While effectively the same as the light green group, their tilt, however, is 130 points. As viewed by the selection committee, however, this makes an interesting condition. The same SATM score, but an overall SAT difference that averages 180 points because of the SAT-CR differential. The rate of admission for this group is only 37 percent, though according to the literature this cluster should be preferred over the verbally talented group.
This is the slide the sums it all up. Our gold group, admitted at a 37% rate, show that 71% eventually attain an initial degree in a STEM field – a rate that is functionally identical to the highest group in the testing. Repeat that: these kids that we admit that we are so-so about, they are just as likely to attain a STEM degree. In the same program, these students are not tracked in any way. They go through the same curriculum as the highest gifted applicants, this is the big takeaway that I need to stress another three or four times in front of the audience, and this is what I think distinguishes us from the talent development models.
The gold group consists of students who would not show up on the radar or in the programs of the TD programs, but we can see that they are capable of accomplishing outcomes of significant strategic importance..
Odds of 7.62 became 5.21. The odds were 42% higher before. No differences in GPA overall, and disaggregated by STEM and humanities. Around 3.3
Chi square no significant difference.
