Why school choice policy has failed: EU comparative study focusing on Estonia (Motivation for a Matching and Mechanism Design) Kaire Põder Project team: Andre Veski, Triin Lauri and Tarmo Veskioja Contents • Good policy: – Equity (background or heterogenity within the school) – Efficiency • Causes of failure: decomposition • Description of data (PISA and secondary): – Mechanism Effect on PISA results • Choice policy configurations and their effect on – Efficiency and equity • Estonian case – regulation design Tallinn University of Technology -- ESCM project Policy platform • NPM and OECD comparative turn • Battle: – Choice-proponents: • choice will bring along both equity and efficiency (complements) – Choice opponents: • there is high trade-off between these two (substitutes) • Causes of failure (literature): – Users ability to execute choice has not been increased • Lower SES are less able to participate in choice (for many reasons) – Schools are allowed to select students (cream-skimming) • Historical reputation effects, aspects of topped-off voucers etc. – Competition must be real (but many cases is not) • league tables usually reflect the results that are also dependent on peer-effects and teacher-effects, etc. • Our arguments why choice ‘reforms’ have not been too successful: – Choice is one policy component in configuration with other institutional rules – Bad choice design (that allows manipulation or magnifies segregation) Tallinn University of Technology -- ESCM project Education policy outcomes • Efficiency of educational outcomes – We don’t assess cost efficiency – Mean results of the education system (e.g. PISA test scores; TIMSS, PIRLS, etc) • Or in school level analysis mean school (PISA) score • Equity – Minimal family background effects • Difficult to operationalise: – Secondary literature: Ferreira, F.H.G., & Ginoux, J. 2011 or – Schuetz, G.; Ursprung, H.W.; Woessmann, L. 2008 – School level analysis: mean tested S.D. of PISA scores within school Tallinn University of Technology -- ESCM project Decomposition Π‘urrent policy outcomes: (residential, ethnic, religious) segregation, low mean results, …. Policy goals and resources: Influenced by theoretical trends (NPM) and international organisations such as OECD + generosity and priorities of the welfare regime Metrics for school priorities (autonomous or centralised) Student preferences: what kind of variability within the system Matching mechanism: allocation In the case of over-demand Tallinn University of Technology -- ESCM project Data: • PISA for measuring Efficiency and Equity: – Collected in many countries – Conducted every 3 year • Here we use data from 2009 – Problems: • In the case of single structure education children were matched long time ago (unables to measure the effect of recent policy amendments) • Typology of ‘school choice regimes’: – Põder et al 2013 • • • • Catchment (zoned); Regulated (initiated by parents); Controlled (central priority list based) Others (unable to classify) – Problem two main choice ‘locuses’: • Primary school choice • Lower-secondary school choice • Explanatory policy variables: – Eurydice, PISA and also secondary literature Tallinn University of Technology -- ESCM project Equity: small background effect (primary school choice, system-level) Tallinn University of Technology -- ESCM project Equity as a mean tested variance (primary school choice, system-level) Tallinn University of Technology -- ESCM project Efficiency and inequality (primary and lower-secondary school choice, school-level) Tallinn University of Technology -- ESCM project Policy configuration: • We search for “Best practices“ – Combination of institutions: • • • • CHOICE = C: availability of parents’ choice at the system level; TRACKING = T: in primary or lower secondary stage of education; VARIABILITY = V: availability of school types and private participation; EMPOWERMENT = E: information about schools (accountability measures like central tests), free transport; • FINANCING = F: incentive schemes for schools and teachers created by financial instruments. – Method: • QCA fuzzy set Tallinn University of Technology -- ESCM project Policy effect on efficiency/equity: • We are after feasible combinations of institutions (C, T, V, E, F) that: – satisfy the minimum consistency requirement (0,75) – separately for ‘choice’ and ‘no-choice’ cases. • Necessary conditions of ‘no-choice’ combination: π β π‘ → πΈπΉ π΄ππ· πΈπ • Necessary conditions of ‘choice’ combination: πΆ β π → πΈπΉ π΄ππ· πΈπ Tallinn University of Technology -- ESCM project Choice is neither necessary nor sufficient for EF&EQ Illustration: Estonian case • 2008-2010 (semi-regulated choice) vs 2011 (free market) – Case specificities: • • • • • Choice within public sector (minimal private participation) Primary school choice (elite and non-elite schools) Aptitude tests in entrance School autonomy (no central priorities or matching) Prep-schooling • Is such ‘total’ market harmful for equity: – Surprisingly de-segregating effect (next slide) Tallinn University of Technology -- ESCM project Tallinn University of Technology -- ESCM project .6 .5 .4 .1 .1 .2 .2 .3 .3 .4 .5 .6 Pr(elite_school) .7 .7 .8 .8 .9 .9 1 1 Example of the middle class child, who doesn’t live in the centre, (family with 2 children, not single parent, average income 500-800 eur, parent reads cultural journals and goes to the cultural events with the child(ren)) 0 1 2 3 4 5 6 0 Mother's_education 1 2 3 4 5 6 Mother's_education Pr(elite_school) 95% CI Pr(elite_school) 95% CI Panel (b): 2008-2010, child attended prep-school .6 .5 .4 .1 .1 .2 .2 .3 .3 .4 .5 .6 Pr(elite_school) .7 .7 .8 .8 .9 .9 1 1 Panel (a): 2011, child attended prep-school 0 1 2 3 4 5 Mother's_education Pr(elite_school) 6 0 1 2 3 4 5 Mother's_education 95% CI Pr(elite_school) 95% CI Tallinn University of Technology -- ESCM Panel (c): 2011, child didn’t attend prep-school project Panel (d): 2008-2010, child didn’t attend prep-school 6 Conclusions • Choice policy is complex – Matching matters if: • there is over-demanded and under-demanded schools • Where is variability between schools (we don’t mean reputable – not reputable kind of variability) – Cases differ (not all cases need matching): • E.g. Swedish reform in 1990es – no need for central priority lists, schools must accept everybody, no topped-off vouchers, previous policy failure – residential segregation • E.g. Estonian case 2011 – Total market (even without central priorities or matching) seems to be more efficient than ‘chaos' by semi-regulated choice which creates perverse incentives (manipulation, prep-schooling, etc.) – But can be similar by most common problems (before choice) • Residential segregation • Private vs public schools • Early tracking Tallinn University of Technology -- ESCM project Thank You Kaire.poder@ttu.ee