Upload
joe-brennan-phd
View
129
Download
5
Tags:
Embed Size (px)
DESCRIPTION
DE-MYSTIFYING THE U.S. NEWS RANKINGS: HOW TO UNDERSTAND WHAT MATTERS, WHAT DOESN’T AND WHAT YOU CAN ACTUALLY DO ABOUT
Citation preview
1
De-mystifying The U.S. News Rankings:
How To Understand What Matters,What Doesn’t, and
What You Can Actually Do About It
Joe Brennan, Ohio UniversityRobert Brodnick, Univ. of the Pacific
Diana Pinckley, Zehno Cross-Media Communications
2
Overview
1. How does U.S. News rank schools?2. Can you affect the peer assessment
score?3. Do the rankings actually differentiate
schools?4. Do prospective students care about
rankings?5. What does all this mean for
marketers?6. What do you think?
3
1. U.S. News ranking system
• Schools are grouped by Carnegie classification
• Editors get data on 15 variables– 14 are objective– 1 (peer assessment) is subjective
• Each variable is weighted by the editors
• A composite score is calculated for each school
• Rank-order is based on composite score
4
Table 1. U.S. News Ranking Model for National Universities Category
weight Sub-factor
weight Relative
weight
Peer assessment 25% 25%
Survey of presidents, provosts, admissions directors Retention and graduation 20%
Freshman retention, fall to fall (4-year average of freshman cohorts) 20% 4%
6-year graduation rates (4r-year average of freshman cohorts) 80% 16%
Faculty resources 20%
Percent of undergraduate classes with fewer than 20 enrolled 30% 6%
Percent of undergraduate classes with 50 or more enrolled 10% 2%
Faculty pay and benefits, 2-year average, adjusted for regional cost-of-living differences 35% 7%
Percent of full-time faculty with terminal degree 15% 3%
Student/faculty ratio, FTE basis 5% 1%
Proportion of FTE faculty members that are full-time 5% 1%
Student selectivity 15%
SAT/ACT (middle 50 percent) 50% 8%
Percent 2005 freshmen in top 10 percent of HS class 40% 6%
Freshman acceptance rate 10% 2%
Financial resources Average spending per FTE student (2-year average) 10% 10% Alumni giving rate 5% 5% Percent of undergraduate alumni of record who made a gift (2-yr. ave.)
Graduation rate performance 5% 5%
Difference between predicted and actual 6-year graduation rate; predictors are SAT/ACT scores and expenditures per student
Total 100% 100%
5
Peer Assessment Score
• Survey of presidents, provosts, chief enrollment officers
• Rate academic programs• Scale of 1 (“marginal”) to 5
(“distinguished”)• Single most important factor in
ranking• Comprises 25% of overall score
6
Peer Assessment Score
Before spending $$$ to move this measure, you need to know:
1. Is it valid (does it actually measure academic quality)?
2. Can it be influenced?
7
What does the Peer Assessment measure?
• We compared peer assessment scores to five other possible indicators of quality:– Median SAT/ACT score– Percent of classes under 20 students– Graduation rate– Alumni giving rate– Public or private control of institution
8
Relationship of other indicators to Peer Assessment Score
Indicator Relationship to Peer Score
Coefficient
Graduation Rate
Direct, large 1.2953
Alumni Giving Direct, moderately large
0.8420
Public Direct, moderately small
0.4332
Freshmen Test Scores
Direct, small 0.0035
Small Class Sizes
No relationship N/A
9
Regression Model
Predicted Score = -2.2149 + 1.2953(Graduation Rate) + 0.8420(Giving Rate) + 0.4332(Control) + 0.0035(Median SAT Score) + e
• How to use this:– Messaging strategies– Compare your predicted to your actual
score
10
Behavior of the Peer Score
• Compared peer scores in 2000 and 2005 for national universities - how much did they change?
• Distribution of peer scores
11
Figure A-1. Distribution of changes in
peer assessment scores, 2000-05
4 7
74
113
46
4
≤ -0.3 -0.2 -0.1 0 +0.1 +0.2
Change in Score
95% of all schools had same score or changed +/- 0.1
12
Figure A-2. Distribution of fi ve-year
mean scores, 2000-05
9
58
83
4227
15 14
1.5-1.9
2.0-2.4
2.5-2.9
3.0-3.4
3.5-3.9
4.0-4.4
4.5-5.0
5-year mean score
(μ = 3.00 σ = 0.73)
Scores cluster tightly around the midpoint and skew towards the higher end of the scale.
13
Do the Rankings Tell Us How Schools are Different?
• Institutional researchers collect data on 100s of variables (magazine uses 15)
• Three questions:1. What factors underlie this vast array of
data?2. Can we create a model to differentiate
schools?3. Does well does the U.S. News model fit
the observed data?
14
Dimensions of higher education
• Examined 170 variables on 247 national universities.
• Pared the data set to 33 variables• Identified seven factors underlying
those variables
15
Factor correlations
IO Control DiversityResearc
h AidAffluenc
e Size
IO 1 -0.349 -0.113 0.333 -0.162 0.501 -0.194
Control -0.349 1 -0.061 0.1 -0.077 -0.341 -0.264
Diversity -0.113 -0.061 1 -0.026 0.058 0.101 -0.027
Research 0.333 0.1 -0.026 1 -0.189 0.315 -0.239
Aid -0.162 -0.077 0.058 -0.189 1 -0.252 0.202
Affluence 0.501 -0.341 0.101 0.315 -0.252 1 -0.059
Size -0.194 -0.264 -0.027 -0.239 0.202 -0.059 1
16
Variable-Factor Relationships
IO Control Research Affluence Aid Size DiversityPERST RATE CLASS < 20 EXP RSCH PCT END / FTES FR AID PCT SUM UNIT % NON WHITE
GRAD RATE PCT UG ENR EXP RSCH / FTES ENDOWMENT INST GRANT % FTES TOTAL FED GRANT %
USN RANK CLASS > 50 EXP INST PCT EXP SVC / FTES LOAN % EXP TOTAL
SAT MEDIAN SF RATIO PCNT FT FAC STATE GRANT %
TOP 10 HS TUITION
AAUP SALARY
REP SCORE
SELECTIVITY
GIVING RATE
AID FR AMT
EXP / INST FTES
EXP / FTES
17
Simple Orthogonal Model
.85FR_PERST_RATE
.85FR_GRAD_RATE_AVG
.88USN_RANK
.90SAT_MEDIAN
.81TOP_10_HS
.76AAUP_SALARY
.83USN_REP_SCORE
.48SELECTIVITY
.58GIVING_RATE
.49AID_FR_AMT
.53EXP_INST_FTES_RATIO
.54EXP_FTES_RATIO
.58CLASS_UNDER_20
.58ENR_PCT_UG_TOTAL
.25CLASS_OVER_50
.72STUD_FAC_RATIO
.70TUITION
.96EXP_RSCH_PCT
.72EXP_RSCH_FTES_GR_PR
.49EXP_INST_PCT
.28USN_PCNT_FT_FAC
1.04END_FTES_RATIO
.72ENDOWMENT
.35EXP_STSVC_FTES_RATIO
2.52AID_FR_AID_PCT
.09AID_FR_INST_GRANT_PCT
.06AID_FR_LOAN_PCT
.12AID_FR_STATE_GRANT_PCT
.50SUM_UNIT
.66FTES_TOTAL
IO
.51EXP_TOTAL
29.79ENR_PCT_UG_NON_WHITE
.01AID_FR_FED_GRANT_PCT
CONTROL
RESEARCH
AFFLUENCE
AID
SIZE
DIVERSITY
e1
e2
e3
e4
e6
e5
e7
e8
e9
e10
e11
e12
e13
e14
e15
e16
e17
e18
e19
e20
e21
e22
e23
e24
e25
e26
e28
e27
e29
e30
e31
e32
e33
.92.92-.94.95.90.87.91-.69.76.70
.84
-.50-.76.76
.53-.70.85.98
1.02.85.59
1.59.30.24.34
.71
.81
.72
5.46.09
MODEL = OrthogonalGFI = .438
AGFI = .365CFI = .556
RMSEA = .185
.73
.74
-.85
18
Model Used byU.S. News FR_PERST_RATE
FR_GRAD_RATE_AVG
USN_RANK
SAT_MEDIAN
TOP_10_HS
AAUP_SALARY
USN_REP_SCORE
SELECTIVITY
GIVING_RATE
CLASS_UNDER_20
CLASS_OVER_50
STUD_FAC_RATIO
USN_PCNT_FT_FAC
SELECTIVITY
FAC RESOURCES
RETENTION
e11
e21
e31
e41
e61
e51
e71
e81
e91
e131
e151
e161
e211
MODEL = USN onlyGFI = .000
AGFI = .000CFI = .000
RMSEA = .402USN_GRAD
EXP_UG_G
e34
e351
1
1
1
1
19
Model showing what U.S. News
should use.87
FR_PERST_RATE.86
FR_GRAD_RATE_AVG.89
USN_RANK.89
SAT_MEDIAN.80
TOP_10_HS.75
AAUP_SALARY.84
USN_REP_SCORE.46
SELECTIVITY.57
GIVING_RATE
.34CLASS_UNDER_20
.88CLASS_OVER_50
.00STUD_FAC_RATIO
.21USN_PCNT_FT_FAC
IO
SIZE
e1
e2
e3
e4
e6
e5
e7
e8
e9
e13
e15
e16
e21
MODEL = USN CorrectedGFI = .606
AGFI = .486CFI = .718
RMSEA = .228
.00USN_GRAD
.51EXP_UG_G
e34
e35
.58
-.94
-.46
.93
.93-.94.95.90.87.92
-.68
.76
.72
20
Model which best explains observed data
.93FR_PERST_RATE
.92FR_GRAD_RATE_AVG
.88SAT_MEDIAN
.83USN_REP_SCORE
.52SELECTIVITY
.60GIVING_RATE
.91AID_FR_AMT
.92EXP_INST_FTES_RATIO
.95EXP_FTES_RATIO
.49CLASS_UNDER_20
.54ENR_PCT_UG_TOTAL
.47CLASS_OVER_50
.65STUD_FAC_RATIO
.94TUITION
.88EXP_RSCH_PCT
.53EXP_INST_PCT
.48END_FTES_RATIO
.46EXP_STSVC_FTES_RATIO
.38AID_FR_AID_PCT
.84AID_FR_INST_GRANT_PCT
.86FTES_TOTAL
IO
.94EXP_TOTAL
CONTROL
RESEARCH
AFFLUENCE
AID
SIZE
e1
e2
e4
e7
e8
e9
e10
e11
e12
e13
e14
e15
e16
e17
e18
e20
e22
e24
e25
e26
e30
e31
.96.96
.67
.57-.25.45.48
1.08
-.50-.74.70
-.73
.94
.77
.34
.62
.78
.81
.77
-.16-.21
-.61
.46
.35
.72
-.02
.57
-.08
.76
.03
.09
.33
-.13
.38
-.23
MODEL = Best FitGFI = .774
AGFI = .661CFI = .876
RMSEA = .131
-.20
-.11
.47-.20.93.03.09
.38
.46
-.54
.07
.23
.21
.18
.30
-.52.38-.321.05.97
-.10-.49
.40
.51-.26
.45
.28
21
Model Summary
Model GFI AGFI CFI RMSEA
Simple Orthogonal 0.438 0.365 0.556 0.185
USN Only (what they do use) 0.000 0.000 0.000 0.402
USN Corrected (what they should use) 0.606 0.486 0.718 0.228
Best Fit 0.774 0.661 0.876 0.131
22
What does this mean?
• U.S. News model does not explain observed data (only 2 of 7 known factors)
• The kinds of data collected about schools doesn’t help individual students find the school which “fits” them best.
• New measures are needed:– Learning– Service to society– Satisfaction of stakeholder needs and wants
23
Do Rankings Matter to Students?
• Yes and no (but mostly no).• Studies show little impact on
undergraduate decision making.• May be more influential for graduate
students.
24
How Undergrads See Rankings
• Art & Science Group (2002)– Only 20% can recall reports or articles about
rankings– Just 8% used rankings info in college
decision
• Lipman Hearne (2006)– Two-thirds don’t use rankings as info source– Mid-Atlantic students and “academic
superstars” are more likely to use rankings
25
How Grad Students See Rankings
• GMAC studies show:– Rankings are strongest influence of MBA
prospects– Students older than 27, and men, are
more likely to be influenced by rankings– For younger students, Web site more
important– For women, personal contact more
important
26
Focusing on What Matters
• Rankings aren’t going away – but don’t overvalue them
• Use the facts about rankings with your administrators
• Use your data to find competitive advantages that really matter
• Deliver messages which differentiate your school and communicate meaningful benefits
27
Questions?
• Diana Pinckley, Zehno Cross-Media Communications
• Rob Brodnick, University of the Pacific
• Joe Brennan, Ohio University