1 12 April 2013
Looking at apples and oranges.
Learning from university rankings
and other composite indicators
Andrea Saltelli
Unit Econometrics and Applied Statistics European Commission, Joint Research Centre
JRC – UNIVERSITIES: HOW TO INCREASE THE CONTRIBUTION OF
UNIVERSITIES TO SCIENCE AND INNOVATION
11 APRIL 2013, Berlaymont Building, Rue de la Loi 200, Schuman Room
Saisana, M., d’Hombres, B.,
Saltelli, A.,
Rickety numbers:
Volatility of university rankings
and policy implications
, 2011,
Research Policy, 40 165–177.
Paruolo, P., Saisana, A., Saltelli,
A., 2013,
Ratings and rankings:
Voodoo or Science?
Journal
Royal Statistical Society A, 176
(2), ---.
‘Invasive’ sensitivity analysis: shaking the assumptions
shows that rankings are volatile
Legend:
Frequency lower 15%
Frequency between 15 and 30% Frequency between 30 and 50%
Frequency greater than 50%
Note: Frequencies lower than 4% are not shown
1-5 6-1 0 11 -1 5 16 -2 0 21 -2 5 26 -3 0 31 -3 5 36 -4 0 41 -4 5 46 -5 0 51 -5 5 56 -6 0 61 -6 5 66 -7 0 71 -7 5 76 -8 0 81 -8 5 86 -9 0 91 -9 5 96 -1 00 Original rank
Harvard Univ 100 1USA
Stanford Univ 89 11 2USA
Univ California - Berkeley 97 3USA
Univ Cambridge 90 10 4UK
Massachusetts Inst Tech (MIT) 74 26 5USA
California Inst Tech 27 53 19 6USA
Columbia Univ 23 77 7USA
Princeton Univ 71 9 11 7 8USA
Univ Chicago 51 34 13 9USA
Univ Oxford 99 10UK
Yale Univ 47 53 11USA
Cornell Univ 27 73 12USA
Univ California - Los Angeles 9 84 7 13USA
Univ California - San Diego 41 46 9 14USA
Univ Pennsylvania 6 71 23 15USA
Univ Washington - Seattle 7 71 21 16USA
Univ Wisconsin - Madison 27 70 17USA
Univ California - San Francisco 14 9 14 11 7 10 6 6 18USA
Tokyo Univ 16 16 49 20 19Japan
Johns Hopkins Univ 7 54 21 17 20USA
Simulated rank range - SJTU 2008
Volatility of university ranking, example
ARWU (Shanghai) 2008 data
Internal coherence of ranking systems
Non-invasive sensitivity analysis. Comparing the
internal coherence of ARWU versus THES by testing
the weights declared by developers with ‘effective’
importance measures suggests that ARWU is more
coherent than THES.
5 12 April 2013
A cooperation
between JRC and RTD
Research
Excellence
Indicators
Research Excellence indicators (JRC -RTD)
Code and Name
Definition
Sources
HICIT
(Highly cited
publications)
Field-normalized count of the 10%
most highly cited publications / GDP
Science Metrix
(Scopus)
PCTPAT
(Patent
Cooperation
Treaty
applications)
Patent applications filed under PCT by
inventors’ country of residence
(fractional counting) in all IPC classes /
GDP
OECD
TOPINST
(World class
universities)
Country scores on global top 250
universities (based on the
Leiden
Ranking
) / GDP
CWTS (Web of
Science)
ERC
(ERC grants
received)
Value of ERC grants received by
country of host organization, equally
spread over project duration / GDP
ERC, DG-RTD
CORDIS
Research Excellence – 41 country “global”
comparison
7 8 9 .8 7 4 .8 7 0 .7 6 0 .2 5 8 .2 5 7 .6 4 5 .4 4 3 .2 3 9 .5 3 7 .1 3 6 .4 3 1 .9 2 9 .2 2 7 .5 2 7 .4 2 7 .1 2 6 .1 2 4 .4 2 0 .2 1 8 .6 1 6 .7 1 4 .5 1 2 .8 1 1 .9 1 1 .6 1 0 .5 1 0 .2 9 .2 8 .5 7 .9 7 .4 7 .3 5 .6 5 .4 5 .3 4 .8 4 .6 3 .8 3 .5 3.5 2 .4 0 10 20 30 40 50 60 70 80 90 100 CH SE DK NL IL FI UK BE AT DE NO US IS EU 2 7 IE FR KR JP SI EE IT ES PT GR CY HU CZ CN LU MT HR LT SK TR PL LV BG RO BR IN RU 2008 2005Composite based on 3 indicators:
• Highly cited Publications
• Top universities
• PCT Patent applications
Research Excellence - ERA countries
9 2 .4 7 7 .4 6 8 .6 6 8 .2 6 6 .2 6 3 .3 4 7 .4 4 5 .6 4 3 .6 4 1 .6 3 3 .6 3 2 .9 2 9 .1 2 9 .1 2 8 .7 2 5 .3 1 8 .6 1 7 .8 1 7 .1 1 6 .9 1 6 .4 1 4 .6 1 3 .4 1 0 .4 6 .6 6 .4 5 .9 5 .7 5 .5 5.4 4 .2 4 .2 3 .6 3 .3 0 10 20 30 40 50 60 70 80 90 100 CH SE IL FI DK NL UK IS BE AT DE NO EU 2 7 FR CY IE IT ES SI HU EE GR PT CZ BG LU MT PL HR LT TR SK LV RO 2008 2007Composite based on 4 indicators:
• Highly cited Publications
• Top universities
• PCT Patent applications
• ERC grants received
Research Excellence Index
versus one of its variables:
Top Universities
AT BE BG CZCY DE DK EE GRES FI FR HU IE IT LTLU LV MT NL PL PT RO SE SI SK UK EU27 HR TR CH IS NO IL BR RUIN CN JPKR US 0 20 40 60 80 1 0 0 T O PI N ST p G D P-0 8 0 20 40 60 80 100 Index_WLD_08 Countries with no universities in global top 250 T op Un iv e rs it ie s in d ic a torResearch Excellence Composite
Using ARWU (Shanghai) ranking, EU versus US, 2008-2012 .
Top one hundred in ranking, per capita, US=100
2008
11 12 April 2013
Using ARWU (Shanghai) ranking, EU versus US, 2008-2012 .
Top fifty in ranking, per capita, US=100
2008
Using ARWU (Shanghai) ranking, EU versus US, 2008-2012 .
Top ten in ranking, per capita, US=100
2008
13 12 April 2013
Top 250
universities/GDP and PCT patents/PCT
Looking at invividual US states and
Chinese region a-la-Aghion.
(Leiden ranking)
Using of the EUMIDA dataset
•
25 EU member states
•
Census of all HEIs from 25 EU member states (2,400 HEIs)
•
2008
Vertesy, D., Annoni, P., Nardo, M., University systems: beyond league tables. Engines of growth growth or ivory towers?, Proceedings of the IREG 2012 conference, Taiwan.
University
rankings
at the regional scale
(JRC with DG REGIO)
CODE VARIABLE NAME
HEI density Higher education density (Nr. of HEIs / pop. aged 18-30)
SI5 intensity ISCED5 student intensity (Nr. of ISCED5 Students / pop. aged 18-26)
IS5 mean Regional average of international students share (ISCED5) per HEI
SI6 intensity
Doctoral student (ISCED6) intensity
(Nr. of ISCED6 Students / pop. aged 22-30)
IS6 mean
Regional average of international doctoral student share (ISCED6) per HEI
RAC Ratio of HEIs defined as research active DDA intensity
Intensity of Doctoral Degrees Awarded (DDA per region)/(pop. age 22-30) * 1000
SSR mean Regional average of student to staff ratio per HEI
University
222 regional
scores
25 countries
Berkshire,
Buckinghamshire & Oxfordshire (100) E Scotland (96) NE Scotland (92)
Capital & City
Regions in Top20: Bremen (88) Vienna (85) Gr. London (81) Gr. Manchester (78) Prague (78) Gr. Brussels (75) Stockholm (75)
University
Country_Average Regional_Score
0
10
20
30
40
50
60
70
80
90
100
FI
IE
IT
SI
LT
RSU I
ndex
Country average
Regional scores
University
regional scale
RCI-Labour market performance
index (*) University System Research Index
University
regional scale
(*) This is one pillar out of 11 pillars of RCI, which includes variables on employment, short- and long-term unemployment, employment gender gap.0 10 20 30 40 50 60 70 80 90 100 0 10 20 30 40 50 60 70 80 90 100 RCI L abo r M ar ke t E ff ici ency Ind ex
University System Research Performance index
Not efficient labour market - stronger research performance Spanish/Greek regions (ES41, ES61; GR11, GR13, GR23)
University
regional scale
Summary innovation index 2010
Intergenerational Mobility in
Accessing Tertiary Education; JRC + World Bank
86 countries
Micro data for more than
1 million adults, including information on their
parents’ educational background
d'Hombres, B., and Nguyen-Hoang (2011) International and Time
Comparisons of Inequality in Tertiary Education. Background report to the World Bank's global study on “Equity of access and success in tertiary
education”.
Wide disparities across Europe
23 12 April 2013
How about getting
the environment
right?
-6.00 -4.00 -2.00 0.00 2.00 4.00 6.00 8.00 10.00 2003 2004 2005 2006 2007 2008 2009 2010 2011 GDP GEE GGE/GDP GEE/GGE
Year on year change of General Government expenditure on education and GDP at constant prices, of GG expenditure/GDP and of GG expenditure on education / Total GG expenditure
(EU27) EU Governments rather than “investing” in education,
have constantly reduced education in the total budget during the crisis (Data from Eurostat).
25 12 April 2013 486 487 488 489 490 491 492 493 494 PISA 2006 PISA 2009 U n w e ig h te d A v e ra g e Cycle
Reading domain
EU countries OECD countries 491 492 493 494 495 496 497 498 499 500 PISA 2006 PISA 2009 U n w e ig h te d A v e ra g e CycleMathematics domain
EU countries OECD countriesPISA data: EU versus the rest of OECD; boys and girls inn their 15’s
495 496 497 498 499 500 501 502 503 504 505 PISA 2006 PISA 2009 U n w e ig h te d A v e ra g e Cycle
Sciences domain
EU countries OECD countriesEducation
6th dimension of the Rule of Law Index
(World Justice Project)
• 83 survey questions • 97 countries (20 EU)
7th dimension of the Rule of Law
Index
(World Justice Project)
• 56 survey questions
29 12 April 2013 Corruption Perceptions Index (Transparency International) • 13 sources • 176 countries (27 EU)
0 20 40 60 80 100 120 140 160 180 200 S in g ap o re U n it e d S ta te s D en m a rk N o rw a y U n it e d K in g d o m A u st ra li a F in la n d M a la y si a S w ed en Ic e la n d Ir e la n d C a n a d a G er m a n y E st o n ia S a u d i A ra b ia Ja p a n L atv ia L it h u a n ia S w it ze rl an d A u str ia P o rt u g a l N et h er la n d s B e lg iu m F ra n ce S lo v en ia C y p ru s C h il e P e ru S p a in C o lo m b ia S lo v ak R e p u b li c M e x ic o S t. L u c ia H u n g ar y P o la n d L u x e m b o u rg S a m o a G h a n a C z e ch R e p u b li c B u lg ar ia A ze rb a ij an T u rk ey R o m an ia It al y S e y c h e ll es B a h a m a s, T h e G re e ce C ro a ti a Ja m a ic a C h in a M a lt a P a ra g u ay C e n tr al A fr ic an R e p u b li c
Ease of Doing Business Rank Paying Taxes
Ease of Doing Business Rank
Starting a Business
Dealing with Construction Permits Getting Electricity
Registering Property Getting Credit Protecting Inv estors
Paying Taxes
Trading Across Borders Enf orcing Contracts Resolv ing Insolv ency
World Bank Index ‘Ease of Doing Business’ + one of its sub-indices: ‘Paying taxes’; blue bars = EU countries
31 12 April 2013
Acknowledgements and thanks
Beatrice d’Hombres,
Patricia Dinis Mota Da Costa,
Luca Pappalardo,
Fiammetta Rossetti,
Michaela Saisana,
Daniel Vertesy
Unit Econometrics and Applied
Statistics, European Commission
Joint Research Centre