Measuring heterogeneity of the educational system Higher School of Economics, Moscow, 2014 Alina...
-
Upload
marshall-hoover -
Category
Documents
-
view
214 -
download
0
Transcript of Measuring heterogeneity of the educational system Higher School of Economics, Moscow, 2014 Alina...
Measuring heterogeneity of the educational system
Higher School of Economics , Moscow, 2014
www.hse.ru
Alina Ivanova, Fuad Aleskerov, Elena Kardanova, Isak Frumin Higher School of Economics, Moscow, Russia
Higher School of Economics , Moscow, 2014
Heterogeneity in education
photo
photo
photo
Mass higher education
Freshmen with different social, economic and academic backgrounds
To which extent has the system of higher education become non-homogeneous?
Higher School of Economics , Moscow, 2014
Studying heterogeneity in education
photo
photo
photo
Different views
diversification of higher education institutions (Reichert, 2009; Carnevale; Strohl, 2010; Posselt et al., 2012)
selectivity of higher education institutions, (Calmand et al., 2009; Hurwitz, 2011; Pastine & Pastine, 2012)
heterogeneity of the student population (van Ewijk, 2010; De Paola, Scoppa, 2010; Bielinska-Kwapisz and Brown, 2012)
Different methods
statistical and econometric tools• Standard deviation, coefficient of variation (Murdoch, 2002)• Gini coefficient (Bosi, Seegmuller, 2006; Sudhir, Segal, 2008)• Multidigraphs (Fedriani and Moyano, 2011)
Higher School of Economics , Moscow, 2014
About our work
photo
photo
photo
Another approach to estimate heterogeneity in education
Heterogeneity of an educational system
A mathematical model based on the construction of universities’
interval order
The Unified State Examination (USE) scores of Russian students are
used to illustrate how our measure of the system’s heterogeneity
works.
.
Higher School of Economics , Moscow, 2014
Our model:Construction of the interval order
photo
photo
photo
The set of all universities AFor each university i it is known the set of students entrance grades• Assume that each university is defined by some interval [] depending on the entrance
exam grade.• Use the mean value of grades and standard deviation to construct the interval order• If the left boundary for i lies at the right of the right boundary of j, then we include the pair
(i,j) to the interval order
Intervals for 5 universities
Graph of the interval order for the 5 universities
54321
00111
00011
00001
00000
00000
5
4
3
2
1
xxxxx
x
x
x
x
x
P
The incidence matrix
Higher School of Economics , Moscow, 2014
Our model:Evaluation of the heterogeneity
photo
photo
photo
1. The interval order P is constructed
2. The notion of ideal interval order Pid is defined
where ij is equal to 1 if the pair (i ; j) belongs to the interval order B1, otherwise =0; ij is equal to 1 if the pair belongs to the interval order B2 , otherwise it is equal to 0.
Comparing the matrices for the real and the ideal interval orders and using Formula we calculate the Hamming distance between two interval orders
Then as the measure of heterogeneity the Hamming distance is used.
Higher School of Economics , Moscow, 2014
Our model: how does it work?
photo
photo
photo
University Mean USE score
Standard deviation
Respective intervalsMean - SD Mean + SD
A 60 5 55 65B 65 5 60 70C 80 3 77 83D 90 5 85 95
Parameters of “real” universities
A B C D
A 0 0 0 0
B 0 0 0 0
C 1 1 0 0
D 1 1 1 0
Incidence matrix for ‘real’ data
• Any ideal system: as an example, clustering • For each university - its ideal counterpart with the mean value as the center of the cluster
and with the interval [m – SD; m + SD]
Ideal points by clustering
Ideal point left Ideal point right
iA 59 66
iB 59 66
iC 78 92
iD 78 92
Intervals for clustered data
iA iB iC iD
iA 0 0 0 0
iB 0 0 0 0
iC 1 1 0 0
iD 1 1 0 0
The incidence matrix for the interval order for clustered data
The Hamming distance between real interval order and PI is equal to 0.08.
Higher School of Economics , Moscow, 2014
The data
The monitoring database "Quality of students’ enrollment - 2012". The database contains;• Mean scores of Unified State Examination (USE) of full-time students
enrolled in 2012 on the Bachelor Degree Programs, • Information on the forms of admission (on competition, out of competition,
on a tuition-based basis or on a state-financed basis, etc.). • Open data
We examine the heterogeneity in the context of integrated groups of majors – in Economics and Management
Group of Majors Economy and management
Total number of universities 379
Total number of enrolled students 91 336
Score mean (range) 46 – 84
Standard deviation (range) 3.1 – 17.2
Higher School of Economics , Moscow, 2014
The notion of ideal system
• Simulated data• Experts view
Artificially ideal
• Real data• Experts view
Based on real data
Higher School of Economics , Moscow, 2014
Our ideal system
Ideal educational system for Economics and Management: expert view
1) A group of best universities that train managers, strategists, high-class
analysts (about 10% of universities, the average score of the whole
contingent of enrolled students should not fall below 75)
2) A group of strong universities that train strong professionals for regional
labor markets (about 70% of universities, the average score from 65 to 74).
3) Group of universities preparing bachelors on applied programs (about 20%
of universities, an average USE score of the admitted contingent should not
be lower than 55).
Interval orders for real
universities constructed
Real university replaced by
“ideal” counterpart
Interval orders for “ideal” universities constructed
Higher School of Economics , Moscow, 2014
The scores' interval Mean St.Dev. Count (%)
>75 79 2.82 10 (3%)
(65;75] 69 3.29 52 (14%)
(55;65] 59 2.67 220 (58%)
<=55 52 1.26 97 (25%)
Prototype for ideal system.
Comparing the matrices P for real and ideal interval orders and using formula (2) we can calculate the Hamming distance between two interval orders: H(P, Pid)=0.26.
The results
Higher School of Economics , Moscow, 2014
The desirable lower limit of the average USE scores for economics majors lies at the level of 55 points
97 universities to delete
The Hamming distance between real and ideal interval orders becomes H(P,Pid)= 0.16.
Improving system
Higher School of Economics , Moscow, 2014
The Hamming distance between real and ideal interval orders becomes H(P,Pid)= 0.21.
Improving system: stage 2
Input artificial data
Construct ideal orders
Compare matrixes
Groups Mean Count (%)
Top 10% 75 28 (10%)
Middle 70% 68 197 (70%)
Bottom 20% 50 57 (20%)
Higher School of Economics , Moscow, 2014
To conclude
A new method of studying heterogeneity in the higher
education system
Our method is based on the comparison of the
hypothetical educational system to the real system
We showed how our method works on Russian data
The model proposed can be applied for any other
data, educational systems, countries
Higher School of Economics , Moscow, 2014
photo
photo
photo
Questions and comments