Early Selection in Hungary A P ossible Cause of High Educational Inequality
-
Upload
lillith-best -
Category
Documents
-
view
18 -
download
1
description
Transcript of Early Selection in Hungary A P ossible Cause of High Educational Inequality
Early Selection in HungaryA Possible Cause of High Educational Inequality
Daniel Hornresearch fellow
IE-HAS and [email protected]
Motivation• Age of selection is likely the best proxy for comprehensive
schooling (The later the first age of selection the longer students will study in heterogeneous groups)
• Comprehensive schooling is said to be decreasing inequality (Meghir and Palme 2005, Pekkarinen, Uusilato and Kerr 2007)– and tracking correlates with higher inequality across countries
(e.g. Hanushek and Woessmann 2005, Pfeffer 2008, Horn 2009)• Why?
– Teachers matter (Hanushek, Rivkin and Kain 2005; McKinsey 2007)
– Peers matter (Sacerdote 2001, Hanushek et al 2003.)– …
• So the longer one studies in selected groups the larger the difference will be
• The Hungarian system is an „ideal” case for testing early selection
• Inequality in the Hungarian education system is especially high
Source: OECD PISA 2009 Vol. II, p 44.
The system of education is certainly not comprehensive
The Hungarian public education system Early-selective tracks
8-yr-ac and 6-yr-ac
Track type combinations within sites
Number of sites
general 8-yr-ac 6-yr-ac 4-yr-ac technikum voc. train.
2463 1 0 0 0 0 0228 0 0 0 0 1 1162 0 0 0 0 1 095 0 0 0 1 1 091 0 0 0 1 0 082 1 0 1 1 0 078 0 0 0 0 0 172 1 0 0 1 0 052 1 1 0 1 0 050 0 0 0 1 1 143 1 0 0 0 0 123 1 1 0 0 0 023 1 0 1 0 0 023 1 0 1 1 1 015 1 1 0 1 1 014 1 0 0 0 1 08 0 0 1 1 0 08 1 0 0 0 1 18 1 0 0 1 1 07 1 0 0 1 0 17 1 0 0 1 1 15 0 0 0 1 0 13 1 0 1 0 1 03 1 1 1 1 0 02 0 1 0 1 0 02 0 0 1 1 1 02 1 0 1 1 1 11 0 0 1 1 1 11 1 1 0 0 1 01 1 1 1 0 0 01 1 1 0 1 0 11 1 1 0 1 1 11 1 1 1 1 1 0
Questions
• Are early-selective tracks status selective?• Do early-selective tracks have a higher value-
added?If answers are affirmative the early selective
tracks increase initial differences between students based on their status. Thus, the Hungarian system is more unequal (ceteris paribus) because of the early selection.
Data
• National Assessment of Basic Competencies (NABC)– Annually collected since 2006, PISA-like survey of all
6th, 8th and 10th grade students (see below)– Reading and mathematical literacy– Approximate mean: 1500, sd: 200 (before 2010: 500/100)
• Cross cohort and cross year comparable. Mean 1500 and sd 100 is only for 2008. 6th grade
– Detailed background questionnaires (generated Socio-economic status (SES) index)
– Individual panel since 2008
The NABC cohorts 4th grade 6th grade 8th grade 10th grade
2003 020 students from every
school 020 students from each track
from each school
2004 020 students from every
school20 students from
every school20 students from each track
from each school
2006 full cohortevery student from a
sample of 195 schools full cohort30 students from each track
from each teaching site
2007 full cohortevery student from a
sample of 200 schools full cohort30 students from each track
from each teaching site
2008*every student from a
sample of 200 schools full cohort full cohort full cohort
2009*every student from a
sample of 200 schools full cohort full cohort full cohort
2010*every student from a
sample of 200 schools full cohort full cohort full cohort
* Individual identification numbers available
Descriptive statistics
Number and percentage of students in different tracks6th to 8th grade panel
Number and percentage of students in different tracks8th to 10th grade panel
Average literacy scoresMath 2008/6th grade 2010/8th grade 2008/8th grade 2010/10th
gradeGeneral 1491 1615 1604 1615*8-yr-ac 1673 1782 1772 18216-yr-ac 1675 1756 1755 18074-yr-ac 1693* 1704technikum 1598* 1608voc. train. 1450* 1452
Read 2008/6th grade 2010/8th grade 2008/8th grade 2010/10th grade
General 1494 1576 1583 1628*8-yr-ac 1667 1746 1754 18176-yr-ac 1661 1726 1739 18064-yr-ac 1682* 1744technikum 1578* 1622voc. train. 1408* 1410
Average SES
Socio-economic Status (SES)
2008/6th grade 2008/8th grade
General (other) -0,059
8-yr-ac 0,85 0,86-yr-ac 0,82 0,84-yr-ac 0,39technikum -0,1voc. train. -0,81
ARE EARLY-SELECTIVE TRACKS STATUS SELECTIVE?1st question
Problems
Pr(track)=a+b*SES+c*Score+e• Test score point estimate for 8-yr-ac is
endogenousTrack->score->track choice
• 6-yr-ac is unbiased– 6th grade score is a unaffected by 6-yr-ac track
effect
Logit estimateDV=6-yr-long academic tracks
6-yr-long academic (1) (2) (3) (4) (5)
SES, std 2.961*** 2.041*** 1.870*** 2.075***
(0.161) (0.0919) (0.0947) (0.0680)
Math score, 6th grade, std 2.122*** 1.859*** 1.889*** 2.498***
(0.120) (0.102) (0.125) (0.0958)
Reading score, 6th grade, std 1.818*** 1.534*** 1.480*** 1.717***
(0.0759) (0.0624) (0.0743) (0.0701)
Female 1.189*** 1.063 1.124* 1.113 1.124**
(0.0758) (0.0693) (0.0722) (0.0796) (0.0535)
Constant 0.0339*** 0.0314*** 0.0254*** 0.0405***
(0.00359) (0.00351) (0.00293) (0.00522)
Observations 94,024 93,799 89,448 34,884 26,713
6-yr-ac closer than 3km n n n y y
Previous school FE n n n n y
Robust, site-clusterer standard errors in parentheses, ORs reported; std - 0 mean/ 1-sd
*** p<0.01, ** p<0.05, * p<0.1
DO EARLY-SELECTIVE TRACKS HAVE A HIGHER VALUE-ADDED?(6TH TO 8TH GRADE)
2nd question
OLS estimateDV= literacy score 2008/8th grade
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)VARIABLES Read Math Read Math Read Math Read Math Read Math General track -150.6*** -141.6*** -28.17*** -1.659 -6.411** 10.07** -7.513*** 4.499 -39.32*** -26.12
(6.917) (10.37) (2.980) (4.378) (2.716) (4.291) (2.841) (4.228) (7.808) (18.04)8-yr-ac 19.84** 26.08* 15.18*** 28.19*** 14.87*** 27.64*** 14.35*** 28.04*** 0.897 32.13
(9.604) (15.34) (3.955) (6.527) (3.652) (6.345) (3.681) (6.180) (11.73) (20.32)Reading, 2008/6 0.748*** 0.543*** 0.165*** 0.543*** 0.165*** 0.533*** 0.174***
(0.00397) (0.00477) (0.00555) (0.00477) (0.00554) (0.00398) (0.00401)Math, 2008/6 0.772*** 0.222*** 0.633*** 0.222*** 0.632*** 0.241*** 0.626***
(0.00561) (0.00475) (0.00646) (0.00477) (0.00647) (0.00355) (0.00490)SES 16.16*** 11.79*** 16.30*** 12.16*** 13.23*** 13.11***
(0.744) (0.899) (0.743) (0.899) (0.542) (0.574)female
43.45*** -26.57*** 43.43*** -26.66*** 43.54*** -27.65***
(0.967) (1.094) (0.967) (1.094) (0.922) (0.960)Ed. Provider FE n n n n n n y y n n
Site FE n n n n n n n n y yConstant 1,726*** 1,756*** 483.6*** 461.3*** 414.7*** 426.3*** 416.3*** 433.3*** 432.7*** 455.1***
(6.703) (10.19) (7.294) (10.13) (7.453) (10.38) (7.573) (10.53) (9.285) (18.42)
Observations 85,550 85,549 85,550 85,549 82,211 82,210 82,211 82,210 82,211 82,210R-squared 0.062 0.051 0.598 0.577 0.638 0.595 0.638 0.596 0.696 0.711
Reference 6-yr-ac, Robust site clustered standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1
Problems
• Endogenity – students start general tracks six years, and 8-yr-ac tracks two years earier than we measure them. Thus 6th grade test scores reflect the quality of the given track (6-yr-ac value-added is unbiased)
• Instrumental variable estimation– instrument: distance from nearest 6-yr-ac and 8-
yr-ac
Instrumental variable estimation OLS IV - First stage IV- Second StageVARIABLES Read Math alt gimn8 Read Math alt -7.513*** 4.499 -97.80** -7.669
(2.841) (4.228) (42.33) (60.49)8-yr-long 14.35*** 28.04*** -6.899 -58.48
(3.681) (6.180) (48.67) (71.32)Reading, 2008/6
0.543*** 0.165*** -7.91e-05*** 3.45e-05*** 0.536*** 0.167***
(0.00477) (0.00554) (1.08e-05) (7.93e-06) (0.00574) (0.00710)Math, 2008/6 0.222*** 0.632*** -0.000198*** 7.43e-05*** 0.205*** 0.636***
(0.00477) (0.00647) (1.98e-05) (1.43e-05) (0.00918) (0.0128)SES 16.30*** 12.16*** -0.0297*** 0.0113*** 13.46*** 12.97***
(0.743) (0.899) (0.00258) (0.00173) (1.375) (1.906)female 43.43*** -26.66*** -0.0136*** 0.00878*** 42.48*** -26.09***
(0.967) (1.094) (0.00281) (0.00178) (1.139) (1.370)Distance from 8-yr-ac
0.000833** -0.00187***
(0.000370) (0.000226)Distance from 6-yr-ac
0.00174*** 0.000670*
(0.000409) (0.000355)Ed. Provider FE y y y y y y
Constant 416.3*** 433.3*** 1.340*** -0.129*** 538.5*** 436.5***(7.573) (10.53) (0.0294) (0.0207) (55.98) (79.61)
Observations 82,211 82,210 85,812 85,812 82,079 82,078
R-squared 0.638 0.596 0.251 0.121 0.626 0.590Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1
DO EARLY-SELECTIVE TRACKS HAVE A HIGHER VALUE-ADDED?(8TH TO 10TH GRADE)
2nd question
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)VARIABLES Read Math Read Math Read Math Read Math Read Math
8-yr-long 11.80 14.65 1.215 1.837 -0.696 1.666 -1.325 1.574 -0.195 -1.311(10.06) (15.32) (5.343) (6.949) (4.632) (6.719) (4.586) (6.398) (6.568) (7.946)
4-yr-long -61.41*** -102.6*** -23.83*** -57.09*** -18.29*** -47.35*** -17.83*** -49.09*** -16.98*** -47.11***(7.405) (10.36) (3.967) (4.714) (3.470) (4.333) (3.469) (4.254) (4.223) (4.519)
Reading, 2008/8 0.682*** 0.549*** 0.194*** 0.549*** 0.193*** 0.512*** 0.167***
(0.00538) (0.00533) (0.00553) (0.00534) (0.00551) (0.00473) (0.00478)Math, 2008/8
0.732*** 0.181*** 0.578*** 0.179*** 0.575*** 0.146*** 0.532***
(0.00878) (0.00507) (0.00891) (0.00505) (0.00884) (0.00424) (0.00811)SES
11.37*** 11.23*** 11.60*** 11.51*** 5.565*** 6.734***(0.744) (0.810) (0.738) (0.801) (0.543) (0.590)
female
31.69*** -73.75*** 31.51*** -73.93*** 26.63*** -72.09***(1.396) (1.520) (1.389) (1.514) (1.116) (1.264)
szakk
-183.6*** -198.4*** -74.17*** -83.47*** -55.90*** -72.16*** -53.69*** -73.87*** -50.66*** -68.80***(8.064) (11.65) (4.366) (5.439) (3.882) (4.891) (3.925) (4.879) (5.305) (5.958)
szaki
-396.1*** -354.9*** -172.1*** -132.2*** -138.4*** -109.8*** -136.9*** -112.4*** -128.5*** -112.3***(8.463) (11.37) (5.194) (5.982) (4.866) (5.593) (4.896) (5.592) (6.059) (6.934)
Ed. Provider FE n n n n n n y y n n
Site FE n n n n n n n n y y
Constant1,806*** 1,807*** 619.0*** 520.1*** 507.7*** 485.7*** 514.2*** 499.5*** 622.5*** 601.7***(7.200) (10.84) (10.17) (15.41) (10.20) (14.60) (10.41) (14.37) (9.424) (13.58)
Observations 76,530 76,521 76,530 76,521 73,256 73,253 73,256 73,253 73,256 73,253
R-squared 0.375 0.255 0.639 0.592 0.659 0.633 0.660 0.634 0.699 0.677
Robust standard errors in parentheses, reference: 6-yr-ac
Instrumental variable estimation OLS IV - First stage IV- Second StageVARIABLES Read Math alt gimn8 Read Math 8-yr-long 13.74*** 40.79*** 13.22 49.55*
(3.517) (4.614) (25.31) (29.26)6-yr-long 16.00*** 42.19*** -15.17 44.15**
(3.446) (3.734) (21.65) (21.00)Reading, 2008/8 0.546*** 0.180*** 8.77e-05*** 0.000116*** 0.549*** 0.178***
(0.00808) (0.00838) (2.43e-05) (2.49e-05) (0.00866) (0.00871)Math, 2008/8 0.213*** 0.708*** 8.48e-05*** 7.13e-05** 0.215*** 0.707***
(0.00636) (0.00991) (2.90e-05) (3.33e-05) (0.00720) (0.0103)SES 10.24*** 9.553*** 0.0110*** 0.0305*** 11.43*** 9.333***
(0.933) (1.033) (0.00325) (0.00422) (1.099) (1.160)female 29.39*** -70.96*** -0.00383 -0.0131** 28.89*** -70.91***
(1.508) (1.647) (0.00372) (0.00661) (1.600) (1.687)Distance from 8-yr-ac
-0.00423*** 0.00101*
(0.000492) (0.000612)Distance from 6-yr-ac
0.00123* -0.00558***
(0.000708) (0.000527)Ed. Provider FE y y y y y y
Constant 445.7*** 245.3*** -0.190*** -0.180*** 437.9*** 248.5***(12.89) (14.22) (0.0530) (0.0593) (14.93) (15.54)
Observations 30,889 30,889 32,396 32,396 30,859 30,859
R-squared 0.537 0.643 0.069 0.065 0.533 0.643Robust standard errors in parentheses, *** p<0.01, ** p<0.05, * p<0.1
Conclusion
Conclusion
• Early selective tracks are status selective (6-yr-ac tested)
• Early selective tracks have a higher value added – even if biased estimates are controlled for– Between 6th and 8th grade in reading, and– Between 8th adn 10th grade in math
• Early selection increases the gap between students of different social background
Thank you for your attention!
Comments [email protected]
Spatial distribution of early selective tracks