14th european workshop on efficiency and productivity analysis (ewepa)
TRANSCRIPT
Locus of decision-making and theefficiency of the education sector
Marie Le Mouel, Alexander Schiersch, Marting Gornig
June 17, 2015
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Motivation
Education spending in the EU averages 6% of GDP, 90% of whichis public. How efficiently are these funds being spent?
In sum, the general pattern of the cross-country analyses suggeststhat quantitative measures of school inputs such as expenditureand class size cannot account for the cross-country variation in
educational achievement. By contrast, several studies tend to findpositive associations of student achievement with the quality of
instructional material and the quality of the teaching force.
Hanushek & Woessmann (2011)
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Research question and Methodology
Research questions
How are resources related to outputs in the education sector?
Is centralisation of decision-making related to the efficiency ofthe education sector?
2-stage procedure a la Simar and Wilson (2007)
Stage 1 : Non-parametric estimation (DEA) of the productionfunction of education
2-input, 2-output framework for education sector as a wholeInput-oriented Sheppard inefficiency scores
Stage 2 : Truncated regression:
Dependent variable: inefficiency scoresExplanatory variables: proxies for teacher quality and for locusof decision-making (Woessmann, 2003)
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Summary of results
Proxy for teacher quality is consistently associated with lowerinefficiency in the education sector
Indicator of overall centralisation shows no significantrelationship to inefficiency
Breakdown by type of decision reveals that centralisation ofdecisions relating to a) personnel management are associatedwith higher inefficiency and to b) planning and structures (e.g.programme design and accreditation) with lower inefficiency.
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Outline
1 IntroductionMotivationResearch question and MethodologySummary of results
2 DataEducation outputEducation inputs and environmental variables
3 Data Envelopment AnalysisInput-oriented Sheppard inefficiency score
4 Truncated regressionEstimation equationsPreliminary Results
5 ConclusionSummary of the resultsRobustness checks and next steps
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Overview
We construct a cross-country unbalanced panel:
23 countries: selection of EU countries, United States &Japan
Period 2002-2010
Number of observations:
First stage: 158 observationsSecond stage: 95 observations
Sources: Eurostat, Education at a Glance (OECD, 2011,2012), World Input-Output Database (WIOD), OECD STANDatabase
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Education output
Education output
Quality-adjusted measure of the number of students going throughthe education system
O’Mahony & Stevens (2009), Schreyer (2010), INDICSER (2012)
Compulsory Education (ISCED 0 to 2):
YCE = PISA ∗ (N0 + N1 + N2) (1)
Post-compulsory Education (ISCED 3 to 6):
YPCE =ISCED6∑ISCED3
Ni ∗Wi (2)
whereWi = pGi p
Ei Wi + (1 − pGi )pEi−1Wi−1 (3)
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Education inputs and environmental variables
Education inputs
1 Number of employees in the Education sector
2 Capital stock in the Education sector
Environmental variables
Teacher quality: relative teacher pay compared to GDP percapita
Locus of decision making: % of decisions taken at each level
Central Intermediate School Sum
Personnel xp ... ... 100%Resource xr 100%Org. of instruction xi 100%Planning & structures xs 100%
TOTAL(xp+...+xs)
4 100%
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Input-oriented Sheppard inefficiency score
Bias-corrected input inefficiency scores, average over 2002-2010
0.5
11.
52
Inef
ficie
ncy
scor
es
SV
NS
VK
US
AS
WE
JPN
IRL
ES
PP
OL
ITA
GR
CN
OR
BE
LF
ING
BR
PR
TD
NK
AU
TN
LDC
ZE
HU
ND
EU
FR
AE
ST
Inefficiency scores Bias-corrected inefficiency scores
Source: Authors calculations based on data from Eurostat, OECD and WIOD
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Estimation equation
Truncated Regression
θit = α + β1TeachSalariesit + β2jDecisionsjit + uit (4)
θit ≥ 1 (5)
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Summary Statistics
Variable Mean Std. Dev. Min. Max. NEduc3to6 2427.9 765.3 1011.2 4142.8 158Educ0to2 68.4 10.5 46.7 85.7 158Employment 30.3 8.3 10 49.9 158CapitalStock 1696.7 1058.6 174.5 4146 158
Inefficiency 1.2 0.2 1 2 158BCInefficiency 1.3 0.3 1 2.4 158
Planning 40.9 30.9 0 100 114Instruction 6.1 7.1 0 25 114Resources 7.5 18.6 0 66.7 114PersManag 18.1 19.5 0 70.8 114Decisions central 18.2 14.5 0 61.1 114Decisions school 47.4 19.6 21.4 96.4 114Teacher salaries 1.2 0.2 0.6 1.7 132
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Preliminary ResultsSummary decision-making index
Inefficiency BCInefficiency(1) (2)
Teacher-salaries -.980 -1.341(.308)∗∗∗ (.535)∗∗
Decisions-central .004 -.001(.006) (.009)
Decisions-school .007 .009(.004)∗ (.006)
cons 1.659 1.717(.284)∗∗∗ (.475)∗∗∗
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Preliminary ResultsLooking at decision areas
Decisions PersManag Res. Inst. Plan.(1) (2) (3) (4) (5)
Teacher-salaries -1.484 -1.782 -1.528 -1.596 -1.152(.620)∗∗ (.649)∗∗∗ (.650)∗∗ (.674)∗∗ (.413)∗∗∗
Decisions-central -.009(.009)
PersManag .014(.006)∗∗
Resources -.012(.012)
Instruction .006(.017)
Planning -.009(.004)∗∗∗
cons 2.430 2.481 2.358 2.336 2.511(.441)∗∗∗ (.499)∗∗∗ (.488)∗∗∗ (.509)∗∗∗ (.349)∗∗∗
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Preliminary ResultsLooking at decision areas
PersPlan ResPlan InstPlan All(1) (2) (3) (4)
Teacher-salaries -1.125 -1.152 -1.003 -1.030(.341)∗∗∗ (.415)∗∗∗ (.408)∗∗ (.379)∗∗∗
PersManag .016 .016(.004)∗∗∗ (.005)∗∗∗
Resources -.002 -.010(.009) (.006)∗
Instruction .022 .004(.012)∗ (.012)
Planning -.012 -.009 -.011 -.010(.003)∗∗∗ (.003)∗∗∗ (.004)∗∗∗ (.003)∗∗∗
cons 2.485 2.507 2.307 2.360(.312)∗∗∗ (.351)∗∗∗ (.373)∗∗∗ (.381)∗∗∗
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Summary of the results
Relative teacher pay is consistently associated with lowerschool inefficiency
No significant relationship between inefficiency and aggregatecentralised decision-making
Breakdown by type of decisions supports microeconomictheory: making personnel management decisions atdecentralised level improves performance if there are externalstandards to assess the quality of instruction (e.g. externalexams and curricula).
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
Robustness checks and next steps
1 Bootstrapped confidence intervals - not reported here
2 Test for frontier shift
Conditional DEASemi-parametric approaches (e.g. SFA)
3 Include Intangible capital stock as a third input
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.
Thank you for your attention
DIW Berlin – Deutsches Institutfur Wirtschaftsforschung e.V.Mohrenstraße 58, 10117 Berlinwww.diw.de
Introduction Data Data Envelopment Analysis Truncated regression Conclusion
References
Barslund, M. & M. O’Mahony (2012). Output and Productivity in the Education Sector. INDICSER Discussionpaper 39
Hanushek, E. & L. Woessmann (2011). The Economics of International Differences in Educational Achievement. InEric A. Hanushek, Stephen Machin, and Ludger Woessmann, Handbooks in Economics, Vol. 3, The Netherlands.
Maimaiti, Y.& M. O’Mahony (2012). Quality adjusted education output in the EU. INDICSER Discussion paper 37
O’Mahony, M., J. Pastor, F. Peng, L. Serrano & L. Hernandez (2012). Output growth in the post-compulsoryeducation sector: the European Experience. INDICSER Discussion paper 32
O’Mahony, M. & P. Stevens (2009). Output and productivity growth in the education sector: comparison for theUS and the UK. Journal of Productivity Analysis, 31:177-194
Schreyer, P (2010). Towards Measuring the Volume Output of Education and Health Services: A Handbook.OECD Statistics Working Papers 2010/02
Simar, L. & P. Wilson (2007). Estimation and inference in two-stage, semi-parametric models of productionprocesses. Journal of Econometrics 136(2007) 31-64
Woessmann, L (2003). Schooling Resources, Educational Institutions and Student Performance: the International
Evidence. Oxford Bulletin of Economics and Statistics, 65:2
SPINTAN Project: funded from the EU 7th Framework Programme. Grant agreement no: 612774.