Admission%20to%20Selective%20Schools - PowerPoint PPT Presentation

About This Presentation
Title:

Admission%20to%20Selective%20Schools

Description:

Formalization with noisy admission tests suggests the ability gap between Zs' ... ST - admission test score threshold for Directly Admitted ... – PowerPoint PPT presentation

Number of Views:42
Avg rating:3.0/5.0
Slides: 29
Provided by: jf15
Category:

less

Transcript and Presenter's Notes

Title: Admission%20to%20Selective%20Schools


1
Admission to Selective Schools
  • ALPHABETICALLY

Štepán Jurajda and Daniel Münich CERGE-EI
CentER, 2008
2
Introduction
  • Alphabetical order is omnipresent. Makes people
    ask about advantage to A individuals. (The
    Economist 2001)
  • Sorting of applicants is important for allocating
    a prize even when allocation is based on
    performance (ability). (van Ours and Ginsburgh,
    2003)
  • We ask whether Z students face lower chances of
    being admitted to oversubscribed Czech schools
    than A ones.
  • Anecdotal evidence
  • Sorted lists with multiple characteristics
  • Oral exams
  • Breaking ties
  • Repeated use of alphabetical sorting (the same
    lottery ticket) at entry to both secondary and
    tertiary education can lead to efficiency losses.
  • We find evidence consistent with substantial
    effects, even though no one seems to be aware of
    this.

3
Names Research
  • Alphabetical sorting in academia and citation
    bias
  • Einav and Yariv (2006), Praag and Praag (2008)
    Katuscak et al. (2005).
  • Racial attributes of first names Bertrand and
    Mullainathan (2003) and Fryer and Levitt (2003)
  • Last names and marriage Golding and Shim (2004)

4
Hypotheses and tests
  • Zs face lower chances of admission to selective
    schools.
  • Hence, ability-alphabet sorting arises within
    school types.
  • Assume ability and alphabetical position are
    independent.
  • Ability
  • A High Marginal Low
  • Z High Marginal Low
  • Admission / Rejection
  • 3. Formalization with noisy admission tests
    suggests the ability gap between Zs and As is
    larger in more selective schools.
  • 4. A natural check do first-name initials
    matter?

5
Some formalization
  • Admissions based on lexicographic order on S
    (score) and N (alphabet position)
  • discrete admission test S reflects ability a
    N(0,1)
  • ST - admission test score threshold for Directly
    Admitted
  • selection among Marginal applicants based
    alphabet their share is m.
  • NT - last-name initial of the last marginal
    student admitted.
  • ability gap between Z and A students grows
    with admission selectivity

6
The Czech Education System in our Data
  • A selective education system
  • academic secondary programs cover only 15 of
    cohort
  • tertiary attainment rate very low at 12
  • Admission rate 63 for academic secondary
    schools
  • 30 of applications (50 of applicants) admitted
    to colleges
  • Population data
  • 1999 national high-school leaving examination
    Maturita administrative register of
    all applications to universities, but we dont
    see the college admission test.
  • Cant differentiate schools that actually use the
    alphabet
  • department-specific procedures, but faculty-level
    code,
  • there are no records of 1999 practices,
  • it is difficult to ask directly.

7
Data
.
8
Data
.
9
Test Score Analysis
  • Assume that Maturita test scores reflect
    ability and look across schools displaying
    different degree of selection (admission rates).
  • Except for 40 in short apprenticeship programs
    with no hope of further education, everyone takes
    the test gt some alphabet-ability sorting across
    entire districts.
  • Selective schools have higher admission standards
    (even within academic ones) as reflected in share
    of applicants admitted to college, in Maturita
    test scores, and in grades at primary school of
    last 3 admitted (SET96).
  • Density at margin is higher in academic programs
    (PISA) (m does not shrink).

10
.
11
So, what do we do?
  • Test for predictions of the alphabet-based
    admissions
  • ability-alphabet sorting across secondary schools
  • direct evidence on college admissions

10
12
Data
.
13
.
14
.
15
.
16
Secondary School Test Score Analysis
  • Students with surnames sorted low in the alphabet
    do achieve higher test scores on average.
  • This is fully robust to including school fixed
    effects.
  • This sorting effect is stronger in more
    over-subscribed schools, as predicted.
  • gt Ability-alphabet sorting suggestive of
    alphabet-based admissions at the secondary school
    level.
  • Esp. given some evidence of no alphabet-ability
    sorting among students graduating from
    primary-level programs 10k 2005 commercial
    practice test scores of 9th graders.

17
College Admission Analysis
  • Want to test whether marginal applications are
    affected. But do not observe school-specific
    admission test scores.
  • identify school-specific marginal applications
    from data.
  • Assign applications percentile rankings based on
    school-specific admission regressions using
    Maturita test scores and secondary school avg.
    success rate.
  • Look at those close to the median predicted
    admission probability. Control for
    faculty/college excess demand.

18
.
19
LAST-NAME INITIAL COEFF. ACROSS PREDICTED
ADMISSION RANGE
20
LAST-NAME INITIAL STEP-FUNCTION SPECIFICATION for
the 40-60 range
21
Size of Estimated Effects
  • In a very selective secondary school (50
    admission rate), moving from A to Z
    corresponds to a move from the median mathematics
    test score to the 60th percentile,
  • about 10 of standard deviation in academic
    programs.
  • So, among marginal college applicants (a group
    with indistinguishable admission test scores),
    Zs are smarter. In presence of noisy entrance
    exams, colleges should choose Z over A among
    marginal applicants.
  • But college admission estimates suggest moving
    from A to Z reduces admission chances by 2
    percent. Reflects mix of schools with and without
    alphabet admissions.

22
Calibrating the Efficiency Loss
  • Simulate population graduating from secondary
    schools
  • Generate 9th graders ability, take them through
    our model admission procedure (70 admission
    rate).
  • Assume schools dont change ability gap.
  • Run test score regressions -- they match Table 2.
  • Next, simulate college admissions (50 admission
    rate).
  • gt Should colleges apply reverse alphabetical
    order, they would improve matches for 52 of
    marginal applicants from academic secondary
    programs.
  • Accounting for the share of this group on all
    admitted, the repeated use of the alphabetical
    order may lead to inefficient school-student
    matches for about 5 of students admitted to
    Czech universities.
  • In reality, those not admitted to one school may
    enroll in another. Czech universities have some
    of the lowest program completion rates in the EU
    CR 63 , OECD 70, Germany 75, Portugal 66,
    Turkey 76 .

23
Could There Be a Wage Effect?
  • Ability-alphabet sorting suggest a wage effect
    within school type, to the extent that wages
    reflect ability.
  • If wages rise with ability the same way for
    workers with different education, there would be
    no effect of the alphabet on wages on average.
  • Use a 1996 household sample and look at male
    wages.

24
.
25
Conclusion
  • Sorting may affect allocation of rationed public
    services. We have a unique opportunity to study a
    sorting name-based mechanism affecting entire
    cohorts.
  • We never document how treatment works and whether
    it exists at all, but support for the hypothesis
    comes from different data and school types.
  • It is also reassuring that first name initial is
    not siggy.
  • The use of the seemingly non-discriminatory
    alphabetical order may have not only
    distributional but also efficiency consequences.
  • FRIN on the use of alphabet in public decision
    making.

26
Gender Gap in Performance under Competitive
Pressure
  • Gneezy et al. (2003) experimental evidence that
    women are less effective in competitive
    environments, even if they perform similarly well
    in non-competitive settings.
  • Field tests using competitive nature of education
    process Örs et al. (2008) applicants to a
    French business school. Show that within this
    group, women outperform their male colleagues in
    non-competitive comprehensive tests, but lag
    behind men in the competitive admission process.
  • We perform similar analysis using entire cohort
    of Czech secondary-school graduates applying to
    all universities.
  • ? admission chances at a given university of men
    and women with similar non-competitive test
    scores
  • ? gender gap in performance under a varying
    degree of competition
  • can control for field-specific unobservable
    ability.

27
.
28
.
Write a Comment
User Comments (0)
About PowerShow.com