101563 Slovak Republic: Skilling up the next generation An analysis of Slovak Republic’s performance in the Program for International Student Assessment Slovak Republic: Skilling up the next generation An analysis of Slovak Republic’s performance in the Program for International Student Assessment Contents Figure 8. PISA performance over me, all subjects, Slovak Republic ................................................... 23 Acknowledgments .................................................................................................................................. 6 Figure 9. Distribu on of students by proficiency level in mathema cs and reading, 2003 and 2012.. 24 Abbrevia ons and Acronyms.................................................................................................................. 6 Figure 10. Standard devia on of PISA scores by discipline, 2003, 2006, and 2012 .............................. 24 Execu ve Summary ................................................................................................................................ 7 Figure 11. PISA problem-solving and mathema cs scores, Slovak Republic and regional Why skills ma er for Slovakia............................................................................................................... 11 neighbors, 2012 .................................................................................................................................... 25 The importance of cogni ve skills ........................................................................................................ 14 Figure 12. Equity performance: Slovak Republic, peer countries, and regional neighbors, 2012 ........ 26 Slovakia’s educa on system ................................................................................................................. 16 Figure 13. ESCS condi ons and aggregate effect, 2003-12................................................................... 27 Cogni ve skills of 15-year-old Slovak students ..................................................................................... 21 Figure 14. Difference between 2012 and 2003 PISA performance, by ESCS and subject ..................... 28 Snapshot of Slovakia’s performance in PISA......................................................................................... 22 Figure 15. PISA performance and socioeconomic composi on of school type, 2012 .......................... 29 Performance in PISA and student background ..................................................................................... 26 Figure 16. Varia on in PISA student mathematucs performance between and within schools, Understanding drivers of performance decline.................................................................................... 32 Slovakia and regional neighbors, 2012 ................................................................................................. 30 Repe on Rates ................................................................................................................................... 33 Figure 17. Urban-rural performance gaps in PISA, 2003-12 ................................................................. 31 Social School Stra fica on ................................................................................................................... 33 Tables Policy Implica ons ................................................................................................................................ 39 Table 1. Annual public expenditure on educa on in Slovakia, 2004–11 .............................................. 18 Promo ng quality and equity in the educa on system........................................................................ 40 Promo ng access to quality early childhood educa on....................................................................... 41 References ............................................................................................................................................ 44 Annex 1: Educa on a ainment across years ........................................................................................ 47 Annex 2. Decomposi on methodology ................................................................................................ 47 Boxes Box 1. How has Slovakia’s educa on system been reformed? ............................................................. 20 Box 2. PISA’s Index of Economic, Social, and Cultural Status ................................................................ 27 Box 3. Roma Communi es in Slovakia ................................................................................................. 32 Figures Figure 1. PISA mathema cs performance by school types and socioeconomic status, 2012 ................ 9 Figure 2. Slovakia’s income convergence: GDP per capita, Slovakia and nearby countries, 1995 - 2013.......................................................................................................................... 12 Figure 3. Slovakia’s popula on age distribu on and projected popula on growth, 2010 - 2050 ........ 13 Figure 4. Three dimensions of skills ..................................................................................................... 15 Figure 5. The structure of Slovakia’s educa on system ........................................................................ 17 Figure 6. PISA mathema cs scores and public expenditures per student, Slovak Republic and PISA countries wordlwide............................................................................................................................. 19 Figure 7. PISA scores, all subjects, Slovakia and neighboring countries, 2012 ..................................... 22 Acknowledgments This report was prepared by a World Bank team consis ng of Ka a Herrera-Sosa, Lucas Gortazar, Jeremie Amoroso, Chris an Bodewig, Sandor Karacsony and Mar n Moreno under the overall guidance of Mamta Murthi, Country Director, Central Europe and the Bal cs, and Cris an Aedo, Educa on Global Prac ce Manager. The note benefited from peer review comments from Rafael De Hoyos. Marc De Francis edited the report. Executive Abbrevia ons and Acronyms ESCS Economic, Social, and Cultural Status ECA Europe and Central Asia Summary ECE Early Childhood Educa on EU European Union GDP Gross Domes c Product OECD Organiza on for Economic Co-opera on and Development OLS Ordinary Least Squares PIRLS Progress in Interna onal Reading Literacy Study PISA Programme for Interna onal Student Assessment RIF Re-centered Influence Func ons TIMSS Trends in Interna onal Mathema cs and Science Study UN United Na ons UNESCO United Na ons Educa onal, Scien fic and Cultural Organiza on VET Voca onal Educa on and Training page 8 page 9 Executive Summary Figure 1. PISA mathema cs performance by school types and socioeconomic status, 2012 Facing the prospects of rapid aging and shrinking popula on over the coming decades, Slovakia needs a highly skilled workforce to help generate the produc vity growth that it needs to fuel its con nued convergence of living standards with its West European neighbors. Skilling up the workforce starts with equipping youth with the right cogni ve and socio-emo onal founda on skills. Interna onal research has iden fied three dimensions of skills that ma er for good employment outcomes and economic growth: cogni ve skills, such as literacy, numeracy, crea ve and cri cal thinking or problem-solving; socio-emo onal skills and behavioral traits, such as conscien ousness, grit or openness to experience; and job- or occupa on-specific technical skills, such as the ability to work as an engineer. Cogni ve and socio-emo onal skills forma on starts early in a person’s life. Good cogni ve and socio-emo onal skills provide a necessary founda on for the subsequent acquisi on of technical skills. Put differently, poor literacy and numeracy skills severely undermine a person’s ability to benefit from further training and lifelong learning. Source: World Bank Staff es mates using PISA 2012 data. 40 points is the equivalent to what an average students learns in a school year. Slovakia can do significantly be er in preparing its next genera on with the right cogni ve founda on skills. OECD averages of ESCS index and PISA mathema cs score are 0 and 500, respec vely. The ESCS school average is calculated by compu ng the weighted average of student’s ESCS at each school. Secondary college refers to secondary voca onal schools. This report focuses on cogni ve skills and examines results for Slovakia from the Program for Interna onal Student Assessment (PISA), which assesses the mathema cs, reading, and science competencies of 15-year-olds. Its findings The explana ons for the significant variance in student achievement seem to lie in the significant social suggest that Slovakia can do significantly be er in helping students develop cogni ve founda on skills. Slovakia’s stra fica on of schools, which goes hand-in-hand with early ability-based selec on into gymnáziums and aggregate mathema cs, reading, science, and problem-solving scores have remained below OECD averages, while voca onal tracks; insufficient efforts to tackle inequity in learning condi ons faced by Slovakian students, and an its 15-year-olds perform on par with peers of several neighboring countries, like Poland, in problem-solving. increase in repe on rates among 15-year-olds. Finally, recent evidence (OECD 2015) points to an increased effect from the selec vity prac ced by schools may exacerbate the segrega on of students by socioeconomic status.  Slovakia’s educa on system is one of the most inequitable educa on systems in the European Union (EU). This is the most worrying finding of this report. Performance in PISA varies significantly by the student’s so- Policy recommenda ons. Slovakia has the dual challenge of ensuring equality of opportuni es and quality across cioeconomic background. While it is not surprising to find dispari es in countries between skills of students its en re educa on system. Countries like Poland have shown that with the introduc on of relevant educa on from rich and poor households, the gap between students from the top and bo om quin le in Slovakia is policies, both objec ves can be achieved. Moreover, interna onal evidence suggests that accomplishing both the equivalent to almost four years of schooling, significantly more than elsewhere in the EU. Performance goals has a posi ve effect on economic growth. This note proposes several policy recommenda ons, based on also varies by type of school – between general secondary schools (gymnáziums) and voca onal secondary an analysis of Slovakia’s PISA data as well as interna onal evidence, to make Slovakia’s educa on system both schools. Slovakia’s 15 year-olds from socioeconomically disadvantaged background are dispropor onately stronger and more inclusive, including: represented in voca onal schools, where aggregate performance is significantly behind that in the gym- náziums. While PISA data do not allow for a disaggregated analysis by ethnic background of students, Roma  Delaying the age of selec on of students into gymnáziums and voca onal schools. Delayed selec on, as youth are likely to suffer dispropor onately from the inequi es in the educa on system stemming from so- in the case of Poland and other countries, has been shown to have a posi ve effect on aggregate student cioeconomic disadvantages. Lastly, while repe on rates are small, they appear to be on the rise and are a achievement and on equity in learning outcomes. major factor contribu ng to the decline of PISA mathema cs scores between 2003 and 2012.  Improving the cogni ve skills of students in voca onal schools by introducing systemic measures that Students from different socioeconomic strata are highly segregated into different types of schools and achieve enhance learning. Since most students in voca onal schools come from less advantaged backgrounds, the widely differing levels of cogni ve skills. This is illustrated in Figure 1, which summarizes the gaps--represen ng purpose of this policy would be to try to offset the influence of peer effects through be er-quality voca- the equivalent of mul ple years of schooling-as measured by PISA mathema cs scores. More than a quarter of onal schools and more general curriculum content. Slovakian 15 year-olds perform at the bo om level in PISA’s mathema cs test and risk leaving school without the  Examining the causes of the rise of repe on rates and finding ways to reduce them. Increasing repe - minimum literacy and numeracy skills needed to succeed in obtaining a produc ve job, in subsequent training, on rates are an important factor that helps explain the decline in learning, par cularly among students and in lifelong learning. In fact, the propor on of students who do not reach the basic proficiency in mathema cs, who belong to the bo om quin les. reading, and science levels has increased since 2003. page 10 page 11  Promo ng universal coverage of quality preschool educa on for children between ages three and five. Inequali es can start early in life, and good quality preschool educa on can help students from all socioeco- nomic backgrounds lay a founda on to be er develop cogni ve and socio-emo onal skills. Chapter 1 Why Skills Matter for Slovakia page 12 page 13 Why skills matter for Slovakia whether the educa on system is providing youth with the skills required to enhance their produc vity and labor Slovakia, like many of its European peers, faces the dual demographic challenge of an aging and shrinking market integra on. popula on. Slovakia has enjoyed the advantage of robust income growth during the past two decades. In 1995, Figure 3. Slovakia’s popula on age distribu on and projected popula on growth, 2010 - 2050 Slovakia’s GDP per capita was slightly more than 40 percent of the EU15 average. By 2013, its GDP per capita rose to 74 percent of that average (Figure 2), narrowing the gap between its living standards and those of the EU15, cemen ng its “Tatra Tiger” moniker in the process. However, looking ahead, its demographic challenge of an aging and shrinking popula on could prevent that convergence in living standards from con nuing. Slovakia faces a popula on decline of almost 10 percent between 2010 and 2050 and median ages are expected to increase from 37.2 to 48.2 during the same years (Figure 3). Figure 2. Slovakia’s income convergence: GDP per capita, Slovakia and nearby countries, 1995 - 2013 Source: World Bank Staff es mates using Eurostat data Leveraging its human capital will be a key strategy if Slovakia is to sustain the gains made in its living standards. Source: World Bank staff es mates using UN Popula on Prospects. Medium variant. Overall, policymakers can mi gate the threats to economic growth due to popula on aging and decline. Recommended policies usually balance quan ty and quality approaches. A quan ty-focused approach centers on expanding the number of workers through tradi onal means such as increasing the employment rate and encouraging immigra on. A quality approach requires, among other things, enhancing labor produc vity by raising the skills of the current and future workforce. Slovakia has significant room for improving its employment rate (currently at 65 percent for the popula on ages 20–64, according to the 2013 Labor Force Survey). Moreover, young Slovaks experience a slow school-to-work transi on. According to Eurostat, since 2011, just under 15 percent of 15- to 24-year-olds have remained idle and are not in employment, educa on, or training (NEET). In addi on to the lost income, poor labor market outcomes at the beginning of the professional life may have a long-las ng, nega ve impact over long-term labor market outcomes, limi ng the possibili es of young people (Schmillen and Umkehrer, 2013; Kahn, 2010; Gregg and Tominey, 2005). As such, skills development will play an important role in harnessing the labor poten al of both the employed and the idle popula on. This report places the spotlight on Slovakia’s next genera on and explores page 14 page 15 The importance of cognitive skills Figure 4. Three dimensions of skills Interna onal evidence shows how much the skills of a country’s workforce ma er for economic growth and shared prosperity. Interna onal evidence suggests that quality of educa on is one of the most important determinants of long-term economic growth.1 Research has explored student assessment surveys from 1960 onward (Hanushek and Woessmann, 2007 and 2012), es ma ng that a score improvement of 50 points in PISA would imply an increase of 1 percentage point in the annual growth rate of GDP per capita.2 Both the share of students achieving basic literacy and the share of top-performing students ma er for growth (Hanushek and Woessmann, 2007; OECD, 2010). A recent OECD (2015) report presents economic returns to universal basic skills, defined as all students enrolled in secondary schooling and the performance of those young people currently not in school raised to achieving level 1 skills (420 points) in PISA by 2030. While low- income countries with lagging educa on systems stand to gain the most, advanced middle-income and high- income countries can expect a significant boost for long-run economic growth (un l 2095) simply by making their educa on systems deliver be er for the weakest students. The report finds that on average, high-income countries could gain a 3.5 percent higher discounted average GDP over the next 80 years if they were to ensure that all students achieved basic skills, defined as level 1 in PISA. As will be presented in this report, a significant and growing share of Slovak 15-year-olds currently perform below level 1 of PISA. Ensuring universal basic skills in Slovakia would add 5.4 percent to discounted future GDP. Ensuring that all have basic cogni ve skills also helps make growth inclusive. Beyond aggregate growth, educa on improves the living standards of individuals. Individuals who are more educated can acquire even more Source: Bodewig and Badiani-Magnusson (2014) skills – and higher-order skills as well – making them more produc ve and employable. Through the acquisi on of “Skills” can be differen ated into separate, mutually reinforcing dimensions along cogni ve, socio-emo onal, more higher-order skills, they can extend their labor market par cipa on over their life mes, which in turn leads and technical skills. Figure 4 presents the differen a on across the different skills dimensions. CogniƟve skills to higher earnings and be er quality of life.3 Educa on is an engine of social mobility: Human capital is a key asset include literacy and numeracy, such as measured in PISA, but also competencies like cri cal thinking and problem- in income genera on and hence cri cal to reducing poverty and increasing shared prosperity (Bussolo and Lopez- solving. Socio-emoƟonal skills, also known as non-cogni ve skills, capture one’s ability to interact with others Calva, 2014). as well as determina on and focus on ge ng a job done. Technical (job-relevant) skills in turn capture one’s ability to perform technical tasks in any occupa on, such as work as a plumber or engineer. Measuring the level of educa onal a ainment does not automa cally mean measuring actual skills. While many countries in Central and Eastern Europe have seen educa onal a ainment (years of educa on, level of educa on completed) expand since the start of the economic transi on, they have not necessarily seen improvements in their performance in interna onal student assessments that measure cogni ve skills, such as PISA (Sondergaard and Murthi, 2012). Cogni ve skills built in childhood and youth are a necessary founda on for successful acquisi on of technical and job-specific skills later in life. The founda ons of cogni ve and behavioral skills are formed early and are the pla orm upon which later skills are built. The most sensi ve periods for building a skill vary across the three dimensions of skills, and skill forma on benefits from previous investments and is cumula ve. Technical and job relevant skills – o en acquired last, through technical and voca onal educa on and training (TVET), higher educa on, and on-the-job learning – benefit from strong cogni ve and behavioral skills acquired earlier in the 1 See Sala-i–Mar n, Doppelhofer, and Miller (2004). 2 See Hanushek and Woessmann (2007) and Hanushek (2010). Using these tests as measures of cogni ve skills of the popula on, they show that countries educa on system. In other words, the cogni ve skills acquired in childhood and youth, such as those measured by that had be er quality of educa on in the 1960s experienced faster economic growth during the years 1960-2000, controlling for other factors. PISA, will help workers to con nuously update their technical skills during their working lives. This is of par cular 3 See Hanushek (2013). page 16 page 17 importance in aging economies such as Slovakia’s where workers need to adapt to technological progress during Figure 5. The structure of Slovakia’s educa on system their longer working lives. This report focuses on cogni ve skills and examines evidence from the performance of 15-year-old Slovakian students on the PISA assessment of mathema cs, reading, science and problem-solving competencies. Introduced in 2000 by the Organiza on for Economic Co-opera on and Development (OECD), PISA is a worldwide study of 15-year-old school students’ performance in three different disciplines: mathema cs, science, and reading. PISA 2012 also included a problem-solving assessment. PISA focuses on the competence of students and their ability to tackle real-life problems in these disciplines. The assessment emphasizes cri cal skills for personal and professional development. The objec ve of the PISA tes ng is to determine how well students are prepared to meet some of the challenges of their future lives. A sample ques on from mathema cs illustrates the applied nature of the PISA tests: “Nick wants to pave the rectangular pa o of his new house. The pa o has length 5.25 meters and width 3.00 meters. He needs 81 bricks per square meter. Calculate how many bricks Nick needs for the whole pa o”.1 In assessing the performance of 15 year-olds, it captures largely those Slovakian students who are in any one track in upper-secondary educa on and a small share of students who are s ll in lower-secondary. Since skills forma on is cumula ve, PISA reflects not just competencies acquired in those schools but competencies acquired even earlier in the educa on system as well. PISA’s scoring system is standardized so that the mean score for each discipline among OECD countries in the year 2000 is 500 points, with a standard devia on of 100 points. According to OECD, 40 points in PISA is equivalent to what students learn in one year of schooling.2 Slovakia has par cipated in PISA rounds since 2003. Slovakia’s education system Slovakia’s educa on system is aligned with the skills forma on process, star ng with early childhood educa on, and focuses on cogni ve skills-oriented learning content. Slovakia’s compulsory educa on system begins with Source: ONISEP, 2015. primary school, where students enter at age six (see Figure 5). While early childhood educa on (ECE) is not compulsory, the government is responsible for its provision from ages three to six (Materská škola). There is a As in many other European countries, Slovakia’s educa on system allows students to pursue different substan al shortage of kindergarten places in Slovakia due in part to the increasing number of births between educa onal tracks at the start of upper secondary educa on. Upper secondary educa on (ISCED 3, typically 2002 and 2011 (OECD, 2015). The popula on of children ages three to five increased from 154,000 in 2006 to for students ages 15 to 19) covers the equivalent of grades 10–12 and is provided in three parallel tracks: general 168,000 in 2012. Between 2007 and 2013 the number of rejected applica ons to kindergarten increased more secondary schools (Gymnáziums), voca onal secondary schools (Stredná odborná škola), and conservatories than five-fold, and the number is expected to con nue increasing (OECD, 2015). In 2015, the government pledged (Konzervatórium3). General secondary schools provide general educa on and prepare students for the school an addi onal ECE investment of €10 million to reduce this shortage (Eurypedia, European Encyclopedia on Na onal leaving examina on, the prerequisite for admission to higher educa on. Secondary voca onal schools provide Educa on Systems, Slovakia). Primary (ISCED 1) and lower-secondary educa on (ISCED 2) form a single structure general and pre-voca onal educa on, prepare students for the secondary school-leaving examina on required and last for nine years. In this single structure, the first stage of primary school is organized as grades one to four, for higher educa on, and offer voca onal post-secondary programs. However, it is possible for students to enter and the second stage as grades five to nine. In the final year of compulsory schooling (year 10), students a end the general secondary schools or conservatories as early as age 11 for a period of eight years. first year of upper secondary educa on. 1 While the skills percentages are lower for the Poland data compared to those observed in the original paper of Autor, Levy, and Murnane (for instance, the original ar cle shows about 12 percentage points for non-rou ne cogni ve skills analy cal for the USA, compared with an increase of less than 2 percentage points in Poland), the original paper covered 40 years of data (between 1960 and 2000). The data used to construct the Poland analysis covers just eight years, 3While the skills percentages are lower for the Poland data compared to those observed in the original paper of Autor, Levy, and Murnane (for instance, the between 2002 and 2010. 2 While the skills percentages are lower for the Poland data compared to those observed in the original paper of Autor, Levy, and Murnane (for instance, the original ar cle shows about 12 percentage points for non-rou ne cogni ve skills analy cal for the USA, compared with an increase of less than 2 percentage points in Poland), the original paper covered 40 years of data (between 1960 and 2000). The data used to construct the Poland analysis covers just eight years, original ar cle shows about 12 percentage points for non-rou ne cogni ve skills analy cal for the USA, compared with an increase of less than 2 percentage between 2002 and 2010. points in Poland), the original paper covered 40 years of data (between 1960 and 2000). The data used to construct the Poland analysis covers just eight years, between 2002 and 2010. page 18 page 19 Slovakia’s public spending on educa on is significantly below OECD averages. The OECD public spending Figure 6. PISA mathema cs scores and public expenditures per student, Slovak Republic and PISA countries on educa on averages 5.3 percent of GDP (and 6.1 percent total, public and private, spending). In contrast, wordlwide Slovakia spent about 4.1 percent of GDP (and 10.4 percent of total government spending) on educa on in 2011. The propor on of Slovakia’s public investment has remained at similar levels since 2004, although it was marginally lower between 2005 and 2008 (Figure 3). Slovakia’s performance is marginally above what should be expected given its current level of public expenditure per student and similar to that of neighboring economies. Nonetheless, some countries with similar levels of investment, like Poland and Latvia, have much higher cogni ve outcomes (Figure 3). Table 1. Annual public expenditure on educa on in Slovakia, 2004–11 2004 2005 2006 2007 2008 2009 2010 2011 Total expenditure As percent of 4.2 3.8 3.8 3.6 3.6 4.1 4.2 4.1 GDP As percent of total 11.1 10.1 10.4 10.6 10.3 9.8 10.6 10.4 government expenditure Expenditure per student (in PPP$) Primary 1,737.7 2,371.1 2,817.8 3,261.8 3,629.2 4,260.3 5,301.3 5,010.4 educa on Secondary 2,472.6 2,475 2,732.6 3,079.1 3,503.8 4,175.8 4,655.3 4,627.1 Source: World Bank staff es ma ons using PISA 2012 data and UNESCO 2012 data. Note: The curve represents educa on a logarithmic approxima on of the sca er plots. Ter ary 4,700.6 3,882.7 4,507.5 4,077.2 4,254.5 4,245 4,562.7 5,683 educa on Source: UNESCO, 2015. page 20 page 21 Chapter 2 Cognitive skills of 15-year-old Slovak students page 22 page 23 Figure 8. PISA performance over me, all subjects, Slovak Republic Cognitive skills of 15-year-old Slovak students Snapshot of Slovakia’s performance in PISA Slovakia’s 15-year-olds performed modestly in PISA 2012. In context, Slovakia’s PISA 2012 scores lagged the OECD average in all subjects.4 Of the 65 par cipa ng countries, Slovakia ranked 34th in math, 40th in science, and 43rd in reading. The difference between Slovakia and the OECD average is widest in reading and narrowest in math, with a differen al of 33 points in reading and 13 points in math. Overall, its 15-year-olds performed worse in PISA 2012 than all peer countries in the Visegrad group and in the Bal c States (Figure 7), sugges ng that the Slovak educa on community is failing to provide students with adequate skills for their future. Figure 7. PISA scores, all subjects, Slovakia and neighboring countries, 2012 Source: World Bank staff es mates using PISA data. Source: World Bank staff es mates using PISA data. Linear interpola on was used to es mate the PISA 2009 scores given the declining performance since 2003 (Panel A). Slovakia has experienced a decline in the three disciplines—math, reading and science—since its inaugural par cipa on in PISA in 2003, if its 2009 performance is excluded.5 The drop in test scores between 2003 and 2012 was Slovakia also saw an increase in the share of innumerate and illiterate students in the 2012 assessment. PISA greater in science (24 points) but was also large in mathema cs (16 points) (Figure 8, panel A). Reading scores remained categorizes scores in six levels of proficiency; students who score below level 2 in reading and mathema cs about the same, 469 in 2012 and 463 in 20036. While performance declined across almost the en re distribu on in tests are considered func onally illiterate and innumerate, respec vely. These defini ons imply that students mathema cs and the bo om half in reading, low achievers lost more than others. Overall, Slovakia’s high-achievers below this level are unable to understand and solve simple problems, severely limi ng their development and increased their performance in reading but decreased in mathema cs and science (Figure 8, panels B and C). subsequent cogni ve and technical skill acquisi on process. 4 While the skills percentages are lower for the Poland data compared to those observed in the original paper of Autor, Levy, and Murnane (for instance, the original ar cle shows about 12 percentage points for non-rou ne cogni ve skills analy cal for the USA, compared with an increase of less than 2 percentage Figure 9 shows how almost 30 percent of 15-year-olds scored below level two in reading and more than one- points in Poland), the original paper covered 40 years of data (between 1960 and 2000). The data used to construct the Poland analysis covers just eight years, between 2002 and 2010. quarter scored this low in math, leaving Slovakia far below the OECD average. In the performance distribu on, 5 While the skills percentages are lower for the Poland data compared to those observed in the original paper of Autor, Levy, and Murnane (for instance, the between 2003 and 2012 more 15-year-olds shi ed toward levels 1 and 2 than shi ed toward levels 5 and 6. original ar cle shows about 12 percentage points for non-rou ne cogni ve skills analy cal for the USA, compared with an increase of less than 2 percentage points in Poland), the original paper covered 40 years of data (between 1960 and 2000). The data used to construct the Poland analysis covers just eight years, As such, the gains by high performers in math and science in 2006-12 were negated by the performance declines between 2002 and 2010. 6 While the skills percentages are lower for the Poland data compared to those observed in the original paper of Autor, Levy, and Murnane (for instance, the among low performers widening the performance gap. original ar cle shows about 12 percentage points for non-rou ne cogni ve skills analy cal for the USA, compared with an increase of less than 2 percentage points in Poland), the original paper covered 40 years of data (between 1960 and 2000). The data used to construct the Poland analysis covers just eight years, between 2002 and 2010. page 24 page 25 Figure 9. Distribu on of students by proficiency level in mathema cs and reading, 2003 and 2012 above comparator countries like Poland and Hungary, but behind others such as Austria, Germany, and the Czech Republic (Figure 11). This new dimension allows one to analyze performance gaps in compared to the mathema cs assessment and, given that the delivery mode of the test is computer-based, illustrates the influence of computer skills on rela ve performance in problem solving. Slovakia’s small varia on of 6 points in this test as compared to the paper-based math test can be a ributed to other factors related to acquisi on of cogni ve skills and not to a lack of computer skills. In fact, the delivery mode (computer-based) of the problem-solving test had a slight posi ve effect on performance if it had any effect at all. Figure 11. PISA problem-solving and mathema cs scores, Slovak Republic and regional neighbors, 2012 Source: PISA 2003 and 2012. In addi on to the decline in aggregate learning outcomes, Slovakia saw an increase in test score inequality. There has been an increased dispersion in the test score distribu on for the three subjects, sugges ng increasing inequali es in learning. For instance, while learning outcomes in reading did not decrease much, the dispersion in scores increased quite a bit. The standard devia on increased from 86 points in 2003 to 101 points in 2012. This makes the reading test score distribu on quite unequal. The dispersion in test scores also increased for mathema cs, although much less than it did for reading, from 90 to 98 points between 2003 and 2012. Inequality for science only marginally increased, remaining broadly at similar levels as in 2003 (the standard devia on was 95 points in 2003 and 98 points in 2012) (Figure 10). Figure 10. Standard devia on of PISA scores by discipline, 2003, 2006, and 2012 Source: World Bank staff es ma ons using PISA 2012 data. Source: PISA 2003, 2006, and 2012. In the new PISA problem-solving assessment, the performance of 15-year-olds in Slovakia was above that of their peers in many neighboring countries, although it was below the OECD average. Conducted for the first me in 2012, the new PISA crea ve problem-solving assessment measures the capacity of students to respond to non- rou ne analy cal problems in a digital environment.7 Results in the crea ve problem-solving test place Slovakia 7While the skills percentages are lower for the Poland data compared to those observed in the original paper of Autor, Levy, and Murnane (for instance, the points in Poland), the original paper covered 40 years of data (between 1960 and 2000). The data used to construct the Poland analysis covers just eight years, original ar cle shows about 12 percentage points for non-rou ne cogni ve skills analy cal for the USA, compared with an increase of less than 2 percentage between 2002 and 2010. page 26 page 27 Performance in PISA and student background Substan al dispari es in cogni ve skills emerge between students from rich and poor households. PISA allows researchers to assess equity in educa on systems and its rela on to socioeconomic status, because it collects in- forma on on performance and student background characteris cs (Box 2). Two measures are used in this report to examine equity in educa on: (i) How much of the varia on in outcomes can be explained by socioeconomic status and (ii) how much of the inequality in scores can be a ributed to ESCS (measured as the PISA score gap between top and bo om ESCS quin les).8 Figure 12 summarizes the evidence from those two measures. Roughly speaking, one quarter of Slovakia’s performance variance can be explained by the student ESCS Index, significantly more than for the performance of its regional neighbors (Figure 12, panel A). The difference in Slovakia’s PISA scores in reading and math between its top and bo om ESCS quin les is approximately 150 points – the equivalent of al- most four years of schooling (Figure 12, panel B). This dura on is also equivalent to the length of upper-secondary educa on in many countries. Both indicators suggest that a student’s household characteris cs dispropor onately determine cogni ve skill acquisi on. Between 2003 and 2006, the socioeconomic condi ons substan ally improved for the bo om 40 percent of the Figure 12. Equity performance: Slovak Republic, peer countries, and regional neighbors, 2012 student popula on, while the ESCS condi ons remained mostly the same for the rest of the students. Between 2006 and 2012, ESCS condi ons remained constant for most of the student popula on (Figure 13, panel A). The overall effect of changes in the ESCS condi ons between 2003 and 2012 on the student popula on can be seen in Figure 13 (panel B). The improved condi ons in socioeconomic status are consistent with the reduced income inequality that occurred in Slovakia during this period. In many countries, an improvement of the socioeconomic condi ons of the household is associated with an increase in students test scores (OECD, 2012). This is because families with higher socioeconomic status are more able to devote more resources or are more mo vated to support their children’s learning. In fact, many countries work to break this strong link between socioeconomic condi ons and student outcomes by building good educa on systems that compensate for different socioeconomic condi ons. Figure 13. ESCS condi ons and aggregate effect, 2003-12 Source: World Bank staff es mates using PISA 2012 data. Note: The Index of Equality of Opportunity is the percent of the variance in reading scores explained by the main predetermined socioeconomic characteris cs in a linear regression (Ferreira and Gignoux 2011). 8While the skills percentages are lower for the Poland data compared to those observed in the original paper of Autor, Levy, and Murnane (for instance, the original ar cle shows about 12 percentage points for non-rou ne cogni ve skills analy cal for the USA, compared with an increase of less than 2 percentage points in Poland), the original paper covered 40 years of data (between 1960 and 2000). The data used to construct the Poland analysis covers just eight years, between 2002 and 2010. page 28 page 29 Students who are 15 years old are distributed among three types of school. Around 43 percent of students remain in basic school, 34 percent a end the secondary voca onal stream (secondary college), and 23 percent a end gymnáziums, the general secondary academic stream. The fact that such a high percentage of the sample of students is s ll in basic school can (mostly) be explained by the school entrance criteria. The PISA sample is constructed for 15-year-olds, which for the 2012 assessment consisted of those born in 1996. In Slovakia, this group, as well as earlier and subsequent cohorts, is exogenously divided into two different grades, nine and 10, because school entrance depends on month of birth within the same year. For instance, more than 80 percent of students born before September are able to enroll in basic school, while those born from September onwards have to wait one more year to join basic school. Thus, students who were born before September join either gymnáziums or voca onal schools by grade 10; while those who were born later remain in basic school. The Source: World Bank staff es mates using PISA 2012 data. sample of students in basic schools also includes students who repeated grades in addi on to those who were born later in the year. Unfortunately, the socioeconomic improvements in Slovakia did not translate into achievement improvements There are significant performance differences by school types. The difference in performance between students in PISA scores.9 In fact, between 2003 and 2012, students from the lowest socioeconomic quin le, as well as in gymnáziums and those in voca onal schools (secondary college) is par cularly striking, accoun ng for almost those from third and fourth quin les experienced larger performance declines than their peers in other quin les. three years of schooling. Figure 15 (panel A) illustrates the differences in performance across the three streams. The largest drop in scores was in science, followed by math (Figure 14). Students belonging to the bo om As seen in Figure 15 (panel B), social stra fica on plays an important role, with many students from the richer socioeconomic quin le experienced severe declines in scores—by 30 points in science, 20 points in math, and 15 socioeconomic strata a ending gymnáziums, while less affluent students go to voca onal schools (secondary points in reading. In contrast, reading performance showed more varia on, as it did not decline for all students. colleges). In addi on, the difference between gymnáziums and basic schools performance is as great as the For instance, students from the top socioeconomic quin le improved their reading scores by 9 points during the difference in performance between gymnáziums and voca onal schools (secondary colleges). period, while performance in reading decreased for the lowest socioeconomic quin le by more than 20 points. Performance also decreased for students who spoke Romani at home, which, for the most part, characterized the Figure 15. PISA performance and socioeconomic composi on of school type, 2012 more socioeconomically disadvantaged households. Figure 14. Difference between 2012 and 2003 PISA performance, by ESCS and subject Source: World Bank staff es mates using PISA 2012 data. Source: World Bank staff es mates using PISA 2012 data. 9 The parameter rela ng socioeconomic status with learning outcomes in the produc on func on was zero, as it was sta s cally insignificant. page 30 page 31 Between-school varia ons in test results are much stronger than within-school differences, sugges ng an Figure 17. Urban-rural performance gaps in PISA, 2003-12 inequitable school system. The varia on in performance between schools is a measure of how big the “school effects” are, and they are closely related to how students are allocated or selected into schools. It can be argued that the lower the between-school variance, the more equitable the educa on system is. Between-school differences in mathema cs performance in 2012 accounted for close to 60 percent of the varia on in student performance in Slovakia. This is significantly above the OECD average (see Figure 16) and that of neighboring countries such as Czech Republic, Germany, and Bulgaria. This reinforces the point that the school that students a end predicts their learning outcomes (OECD, 2014b). Figure 16. Varia on in PISA student mathematucs performance between and within schools, Slovakia and regional neighbors, 2012 Source: OECD (2014b) PISA performance also varies across urban and rural loca ons as well as by regions in Slovakia. First, there is a clear urban-rural gap in student achievement that has increased over the years (Figure 17, panel A). The urban- rural gap in Slovakia is among the largest in the region and larger than that in countries such as Slovenia, Latvia, Romania, and Poland. About 95 percent of students in the PISA 2012 sample who speak Romani at home live in rural areas; and over 80 percent belong to the bo om quin le (see also Box 3). Second, while in urban areas PISA mathema cs performance decreased marginally (by less than 10 points) among low achievers and increased for the top achievers, in rural areas it dropped substan ally for both middle and low achievers (Figure 17, panel B). Third, performance in 2012 varied significantly across Slovak regions (Figure 17, panel C). Source: World Bank staff es mates using PISA 2003, 2006 and 2012 data. page 32 page 33 Figure 18. Decomposi on of changes in PISA reading scores between 2003 and 2012 into factors and by student achievement group Source: World Bank staff calcula ons based on PISA 2003 and 2012 data. Note: Results decomposi on was done using an Oaxaca-Blinder method on RIF-regressions for each quin le of the distribu on of performance (Firpo, For n, and Lemieux, 2009). Low, middle, and high achievers are students in the 20th, 50th, and 80th percen le, respec vely.11 The teacher par cipa on index measures the degree of par cipa on that teachers have in school decision-making (OECD, 2005). Note that when decomposing differences, one explanatory factor is o en found to be nega ve or higher than the actual difference, meaning that other factors outweighed its impact. For more informa on, see Annex 2, Table A2. Understanding drivers of performance decline Repetition Rates Decomposing the aggregate performance change by student type allows one to examine changes in underlying While grade repe on has not been common in Slovakia, it seems to be on the rise and is associated with factors.10 As shown in Figure 8 (panel B), PISA reading scores declined at the bo om of the performance student performance. There was a significant increase in the repe on rate reported by students in Slovakia distribu on, while they increased at the top. Figure 18 depicts the actual performance gap between 2012 and between 2003 and 2012, from 2.5 percent to 7.6 percent (OECD, 2013c). Most affected are students who belong 2003 as well as factors that contributed to the changes in reading test scores between those years by different to the bo om ESCS quin le. In 2003, for example, 9 percent of students who belonged to the bo om quin le performance groups (low, middle, and high achievers). The analysis shows that while socioeconomic peer effects were repeaters, but by 2012 that number had increased to 25 percent. This contrasts with an overall drop in grade (measured with the average ESCS at the school level) had a small and posi ve impact on scores between 2003 repe on among OECD countries during the same period. The gap between repeaters and non-repeaters is the and 2012, the gains were offset by repe on, quality of educa onal resources, a higher propor on of minority largest in the OECD. students and students speaking a different language at home than Slovak. Low achievers (students in the 20th percen le of the reading distribu on) are the most affected by these factors. In contrast, high achievers (those Social School Stratiϐication Slovakia has high levels of social school stra fica on. Social school stra fica on is measured with an index that belonging to the 80th percen le) substan ally benefited from peer effects, while the influence of repe on or represents the correla on between the PISA student’s socioeconomic status and the school’s average socioeconomic other differen a on policies was minor. In mathema cs and science these factors are also sta s cally significant status. The index runs from zero to 1. Countries without any social stra fica on, where children from different but with much smaller impact, with most of the decline remaining unexplained. socioeconomic backgrounds were studying together in all schools, would have a value of zero in the stra fica on index. In such a case, the schools would exactly reflect the diversity of the society. For the most part, countries that 10 The analysis in this sec on is based on a two-fold Oaxaca–Blinder (OB) decomposi on analysis based on a tradi onal educa on produc on func on. The perform be er in PISA have low levels of social school stra fica on (Figure 19 ). In fact, there are only a handful variables used in the analysis include age, gender, ESCS student, ESCS school (peer-effects), grade, quality of educa onal resources (index), % of minority students at school, Romani speaker (dummy), and repe on (dummy). Further, the Firpo, For n, and Lemieux (2009) methodology allows one to apply the of countries, like Germany or Austria, which perform above the OECD average and also have high levels of social OB decomposi on to the 20th and 80th percen le observa ons, and not just to the means alone (see Annex 2). stra fica on. Social stra fica on in Slovakia has marginally increased from 0.58 in 2003 to 0.63 in 2012. page 34 page 35 Figure 19. Social stra fica on: PISA math scores by degree of school stra fica on, Slovak Republic and PISA Figure 20. Simulated PISA mathema cs gains (Slovakia) based on emula on of other na ons’ social stra fica on countries worldwide indices Source: World Bank staff calcula ons using PISA 2012 data. Note: The simulated mathema cs gains here represent the expected scores for Slovakia if its social stra fica on were reduced to the levels of each of the countries shown. Gains are based on a cross-country linear regression. Source: World Bank staff es mates using PISA 2012 data. Note: PISA mathema cs scores on ver cal axis. Index of School Social Stra fica on on horizontal axis. The index ranges from 0 to 1. A higher index indicates a higher correla on between students’ and schools’ socioeconomic Slovakia’s educa on system selects students between general and voca onal tracks at a young age. While status. OECD mathema cs score average 500 points. OECD average Index of School Social Stra fica on is 0.525. most students get selected into different types of school at age 15, the system also allows for an earlier selec on of students, star ng from age 11, into gymnasiums (Figure 21). Interna onal evidence suggests that such early Reducing social stra fica on could raise performance. Social stra fica on heavily depends on parental decisions tracking has a nega ve impact on equity and does not improve average performance. Hanushek and Woessman and on policies governing the selec on of students into different educa on tracks. Simula ons using the 2012 PISA (2006) used previous PISA data to show how early tracking systems lead to a systema c increase in inequality of data suggest that PISA scores could increase if the level of social segrega on were reduced to match that of some student performance without affec ng average performance levels. At the na onal level, similar evidence has comparator countries (Figure 20.) For instance, PISA mathema cs scores in Slovakia would be almost 40 points been found in Poland (see Jakubowski et al., 2010) and Germany (Piopiunik, 2013). The findings suggest that no higher – the equivalent of one year of schooling – if its school social stra fica on levels were like those in Finland. efficiency is gained by streaming students early and that, in fact, delayed tracking can promote be er performance among all students. page 36 page 37 Figure 21. Age of first selec on between general and voca onal tracks, Slovakia and PISA countries worldwide Figure 22. PISA mathema cs scores by ESCS index and school type, 2012 Source: World Bank Staff es mates using PISA 2012 data. Source: OECD (2010a) Note: 40 points is the equivalent to what an average students learns in a school year. OECD averages of ESCS index and PISA mathema cs score are 0 and 500, respec vely. The ESCS school average is calculated by compu ng the weighted average of student’s ESCS at each school. The overall message from the analysis is that educa on in Slovakia does not act as an engine of social mobility, Secondary college refers to voca onal secondary school. but appears to further deepen socioeconomic disadvantage. There are three key messages that emerge from this analysis. First, performance in PISA varies significantly by the student’s socioeconomic background. Second, performance varies by type of school. Third, 15-year-olds from socioeconomically disadvantaged backgrounds are dispropor onately represented in voca onal schools and voca onal secondary schools (or indeed are s ll in basic schools, where aggregate performance is significantly lower than that in general secondary schools12). Figure 22 summarizes the above evidence in a single picture showing how students from different socioeconomic strata distribute into different educa onal tracks and achieve widely differing levels of cogni ve skills as measured by PISA. 12 In Slovakia, the 15-year-old group is divided into two different grades, 9 and 10, exogenously. This is because the school entrance depends on the month of birth. The majority of students who were born September onwards are s ll enrolled in basic schools (although students who are repea ng grade enter in this group too), while those who were born on or before August have already selected between different types of schools. Tracking takes place between grades 9 and 10. page 38 page 39 Chapter 3 Policy Implications page 40 page 41 Policy Implications delaying the age of selec on of students into voca onal and general tracks, for example un l a er the end of Slovakia is facing the dual challenge of a fast aging and shrinking popula on and significant cogni ve skill compulsory educa on, may promote quality and equity. The example of Poland is instruc ve: Poland delayed the shortages among many youth. The principal response to ensure growth and convergence in mes of demographic selec on into voca onal and general tracks by one year un l the age of 15 as part of its comprehensive educa on decline is to raise produc vity. Doing this requires raising the skills base of the popula on. Yet Slovakia is not reform package in 2000, thereby extending all students’ exposure to general curriculum content. Poland has seen doing as well as it needs to in equipping all its youth for produc ve employment along longer working lives. It significant improvements in both aggregate PISA scores and equity indicators since then (Jakubowski et al., 2010). is already experiencing the nega ve consequences, evident in the rela vely high share of its youth who are not Moreover, the incen ves that schools are faced with in Slovakia need to be reviewed. For instance, given the high in employment, educa on or training (NEET), a share that was measured in 2013 as 13.7 percent of its 15- to degree of decentraliza on in admission decision-making, schools appear to have incen ves to select students 24- year-olds (Eurostat data). A principal challenge in skilling up its next genera on is to reduce the impact of based either on socioeconomic or ability characteris cs, reinforcing peer effects and reducing school effort. socioeconomic disadvantage on access to quality educa on and on learning outcomes. Raise the quality of educa on in voca onal secondary schools. PISA scores in voca onal schools significantly Addressing equity issues in educa on is crucial for reducing poverty, boos ng shared prosperity, and fostering lag those in gymnasiums. While some of these differences may be expected, the gap in performance between sustained economic growth and convergence in living standards. In the case of Slovakia, students’ improved students in gymnasiums and those in voca onal and basic schools is unacceptably large, the equivalent of mul ple socioeconomic condi ons of students, par cularly among those at the bo om of the ESCS distribu on, do not years of schooling. It raises the ques on whether the curriculum in voca onal schools devotes sufficient a en on appear be associated with improved test scores. In fact, the data show a decline in student performance since to general, cogni ve content and whether teachers are equipped with the tools to impart those skills effec vely 2003. While household socioeconomic condi ons are important for student achievement, so are many other to students who need addi onal support. If the voca onal curriculum and teachers are not so equipped, there is a policies in the educa on system. It is important to note that since the PISA test is conducted at age 15, most need to increase educa onal quality in these schools, even if, as recommended in this report, the age of selec on students taking it are in their first year of secondary educa on, sugges ng that not all of their varying performance between voca onal and general secondary schools is raised. can be a ributed to the type of secondary school they a end. However, the analysis also points toward the fact Reconsider repe on as a differen a on policy. With an increased body of evidence showing the failure of that, for reasons not captured by the PISA data (and sugges ng the need for more analysis of socioeconomic reten on, many OECD countries are either banning or reducing repe on in schools. However, this is not the stra fica on of schools in basic educa on), students get selected into different tracks in such a way that enrolment case in Slovakia, where repe on by 15-year-olds has increased from 2 percent to 7 percent in the last decade. and performance across secondary school types is stra fied along socioeconomic background lines. This suggests Moreover, repe on is dispropor onately common among the most vulnerable students. This report shows that that factors related to school segrega on based on socioeconomic condi ons, even during basic educa on, an increase in repe on is associated with a decline in performance, especially among low-achieving students. and repe on rates all play a role in aggregate performance and the inequity in outcomes. Moreover, recent Thus, it is appropriate to explore in more detail the causes of the increasing repe on rates and measures that evidence (Šiškovič & Toman, 2015) also suggest that the increase in schools’ selec vity may be an important factor could help address those causes. contribu ng to school segrega on that would benefit from examina on in more detail. This report suggest recommenda ons to help Slovakia increase its quality of educa on and to make the system Promoting access to quality early childhood education more equitable. The recommenda ons proposed below rely on the analy cal findings of this note as well as Expand access to quality preschool educa on. Expanding access to preschool can help improve the opportuni es interna onal experience. Given that public spending on educa on in Slovakia is low rela ve to the OECD spending of children from disadvantaged backgrounds. Global evidence shows that providing quality preschool educa on is average, there appears to be some fiscal space to make investments in quality and equity-enhancing policies. important for promo ng children’s social, emo onal, physical, and cogni ve development; it also increases school readiness, which helps learning (Heckman and LaFontaine 2010; Heckman 2008; Engle et al. 2011). Cogni ve skills Promoting quality and equity in the education system gaps start opening during early life, and inequali es in access to early childhood perpetuate learning gaps. Given To ensure greater equity in Slovakia’s educa on system, coordinated policies need to be introduced to promote that a endance in early childhood programs is correlated with higher educa onal a ainment and higher socio- inclusiveness and quality for the most vulnerable within the system. These measures should be coupled emo onal skills, policies to improve access to and the quality of early childhood educa on in Slovakia would help with adequate monitoring and accountability mechanisms that prevent local policies of any type from causing improve the skills for new labor market entrants, transla ng into higher human capital and produc vity and likely segrega on at the school or classroom level to proliferate. Raising the quality of educa on for all can be achieved contribu ng to an overall reduc on in learning inequality. through three channels: (i) by delaying the selec on of students between voca onal and general educa on tracks, (ii) by raising the quality of educa on in voca onal and voca onal secondary schools, and (iii) by reconsidering While general access to preschool educa on is rela vely high in Slovakia--over 80 percent of the students in the policies related to repe on. PISA sample had more than one year of pre-primary educa on--inequali es in access remain. For instance, only 68 of students who belong to the bo om quan le a ended more than one year of preschool educa on, compared Delay the age of selec on of students into voca onal and general tracks. Interna onal evidence suggests that page 42 page 43 with 89 percent of those in the top quin le. About 14 percent of students belonging to the bo om quin le never a ended preschool, compared with only 2 percent of students in the top quin le. An analysis of determinants of student achievement does not show preschool educa on as a sta s cally significant variable. This can have many causes, including the possible lack of quality of the preschool services offered when students a ended these programs during the late 1990s and early 2000s. However, evidence does show that Roma children who completed ECD are less likely to be enrolled in special schools (by about 7 percent) and are more likely to complete secondary school and less likely to be on social assistance (World Bank, 2012). Universal preschool educa on for children aged 3–5 in Slovakia would provide a great opportunity to effec vely narrow the skills gap from the early stages of children’s lives. References Annex page 44 page 45 References Hanushek, E. 2013. “Economic Growth in Developing Countries: The Role of Human Capital.” Economics of EducaƟon Review, vol. 37: 204 – 12. Amermueller, A. 2004. “PISA: What Makes the Difference? Explaining the Gap in Pisa Test Scores Between Finland and Germany.” Discussion Paper No. 04-004, ZEW Center for European Economic Research, Mannheim, Germany. Hanushek, E., and L. Woessmann. 2006. “Does Educa onal Tracking Affect Performance and Inequality? Differences-in-Differences Evidence Across Countries.” The Economic Journal, vol. 116, no. 510: C63-C76. Autor, D., F. Levy, and R. Murnane. 2003. The Skill Content of Recent Technological Change: An Empirical Explora on. Quarterly Journal of Economics, 118(4), November 2003, 1279-1334. Hanushek, E., and L. Woessmann. 2007. “The Role of Educa on Quality in Economic Growth.” Policy Research Working Paper 4122. Washington, DC: World Bank. Barro, R., and J. Lee. 2013. “A New Data Set of Educa onal A ainment in the World, 1950-2010.” Journal of Development Economics, vol. 104: 184 – 98. Hanushek, E., and L. Woessmann L. 2012. “Do Be er Schools Lead To More Growth? Cogni ve Skills, Economic Outcomes, and Causa on.” Journal of Economic Growth, vol. 17: 267-321. Barrera-Osorio, F., V. Garcia-Moreno, H. A. Patrinos, and E. Porta. 2011. “Using the Oaxaca-Blinder Decomposi on Technique to Analyze Learning Outcomes Changes Over Time: An Applica on to Indonesia’s Results in PISA Math.” Hanushek, E., G. Schwerdt, S. Wiederhold, and L. Woessmann. 2013. “Returns to Skills Around the World: Evidence Working Paper 5584. Washington, DC: World Bank. From PIAAC.” Working Paper 199762. Cambridge, Mass.: Na onal Bureau of Economic Research. Bloom, N., Lemos, R., Sadun, R., and J., Van Reenen. 2014. “Does Management Ma er in School?” Na onal Bureau Heckman, J. 2008. “Schools, skills, and synapses”. Economic Inquiry, 46(3): 289-32. of Economic Research (NBER) Working Paper. Heckman, J., & LaFontaine. 2010. “The American High School Gradua on Rate: Trends and Levels”. Review of Bodewig, C. and R. Badiani Magnusson. 2014. “Skilling Up Vietnam: Preparing the Workforce for a Modern Market Economics and StaƟsƟcs, 92(2): 244–262, (2010). Economy.” Direc ons in Development Series. Washington, DC: World Bank. Jakubowski, M., H. A. Patrinos, E. E. Porta, and j. Wisniewski. 2010. “The Impact of the 1999 Educa on Reform in Bussolo, M. and L. F. Lopez-Calva. 2014. “Shared Prosperity: Paving the Way in Europe and Central Asia.” Poland.” Policy Research Working Paper 5263. Washington DC: World Bank. Washington, DC: The World Bank. Kahn, L. B. 2010. “The Long-Term Labor Market Consequences of Gradua ng from College in a Bad Engle, p. Fernald, L., Alderman, H., Behrman, J., O’Gara, C., Yousafzai, A., Cabral de Mello, M., Hidrobo, M., Ulkuer, Economy.” Labour Economics, 17: 303 - 16. N., Ertem, I., Iltus, S., & Global Child Development Steering Group. 2011. “Strategies for Reducing Inequali es and OECD. 2005. PISA 2003 Technical Report. Paris. Improving Developmental Outcomes for Young Children in Low-Income and Middle-Income Countries”. The Lancet, October (Vol. 378, Issue 9799, Pages 1339-1353). DOI: 10.1016/S0140-6736(11)60889-1. OECD. 2010. “The High Cost of low Educa onal Performance: The Long-Run Impact of Improving PISA Outcomes.” OECD Publishing. Retrieved May 12, 2015 from h p://www.oecd.org/pisa/44417824.pdf Eurostat: Regional Sta s cs. 2011. Retrieved July 2015 from h p://ec.europa.eu/eurostat/ OECD. 2011. “PISA 2009 Results: Learning to Learn,” vol. 3. Paris. Ferreira, H. G., and J. Gignoux. 2011. “The Measurement of Educa onal Inequality: Achievement and Opportunity.” IZA Discussion Paper No. 6161. OECD. 2012. “PISA 2009 Technical Report.” Paris. Retrieved April 10, 2014 from h p://www.oecd.org/pisa/ pisaproducts/50036771.pdf Firpo, S., N. For n, and T. Lemieux. 2009. “Uncondi onal Quan le Regressions.” Econometrica, vol. 7, no 3: 953 – 73. OECD. 2014a. “PISA 2012 Results: What Students Know and Can Do,” vol. 1. Paris. Retrieved April 10, 2014 from h p://www.oecd.org/pisa/keyfindings/pisa-2012-results-volume-I.pdf Gregg and Tominey, 2005. “The Wage Scar from Male Youth Unemployment”. Labour Economics. 12: 87-509. OECD. 2014b. “PISA 2012 Results: Excellence Through Equity,” vol. 2. Paris. Hanushek, E. 2010. “The High Cost of Low Educa onal Performance. The long-run economic impact of improving PISA outcomes.” OECD Publica ons. OECD. 2014c. “PISA 2012 Results: Crea ve Problem Solving,” vol. 5. Paris. Hanushek, E. 2009. “School Policy: Implica ons of Recent Research for Human Capital Investments in South Asia OECD. 2014d. “Reviews of Evalua on and Assessment in Educa on.” Paris. and Other Developing Countries.” EducaƟon Economics, vol. 17, no. 3: 291 - 313. OECD. 2014e. “TALIS 2013 Results: An Interna onal Perspec ve on Teaching and Learning.” Paris. page 46 page 47 OECD. 2015a. Universal Basic Skills: What Countries Stand to Gain. Paris. Annex 1: Education attainment across years Šiškovič, M., & J. Toman. 2015. “OECD Review of Policies to Improve the Effec veness of Resource Use in Schools. Country Background Report for the Slovak Republic.” Educa onal Policy Ins tute, Ministry of Educa on, Science, Research and Sport of the Slovak Republic. Table A1. Percent of popula on aged 15+ by highest level of schooling a ained and average years of schooling in Slovak Republic, 1990-2010 Piopiunik, M. 2013. “The Effects of Early Tracking on Student Performance: Evidence from a School Reform in Bavaria”, Ifo Working Paper No. 153. Munich: Germany. Highest level a ained Average Years of Schooling Year Primary Secondary Ter ary Total Primary Secondary Ter ary Patrinos, H., Garcia-Moreno, V., & De Hoyos, R. 2015. “The Impact of an Accountability Interven on with Total Completed Total Completed Total Completed Diagnos c Feedback. Evidence from Mexico”. Policy Research Working Paper, no. 7393. Washington, DC: World 1990 38.2 28.9 52.3 32.6 7.9 4.0 10.69 8.44 2.01 0.24 Bank. 1995 28.4 22.2 59.5 38.4 10.3 5.0 11.24 8.56 2.37 0.31 2000 27.5 21.7 60.8 42.6 9.7 4.7 11.20 8.46 2.46 0.29 Sala-i-Mar n, X., G. Doppelhofer, and R. I. Miller. 2004. “Determinants of Long-Term Growth: A Bayesian Averaging 2005 9.4 7.5 74.3 55.7 15.2 7.3 12.22 8.27 3.50 0.45 of Classical Es mates (BACE) Approach.” American Economic Review vol. 94, no. 4: 813–35. 2010 1.3 1.0 80.1 61.8 18.3 8.8 12.82 8.04 4.24 0.54 Source: Barro-Lee Dataset (2012). Schmillen, A., and M. Umkehrer. 2013. “The Scars of Youth: Effects of Early-Career Unemployment on Future Unemployment Experience.” IAB Discussion Paper No. 6/2013. Shewbridge, C., et al. 2014.” OECD Reviews of Evalua on and Assessment in Educa on: Slovak Republic 2014.” Annex 2. Decomposition methodology OECD Publishing. h p://dx.doi.org/10.1787/9789264117044-en The analy cal approach used in this report is based on the Firpo, For n, and Lemieux (2009) methodology. Sondergaard, L. and M. Murthi. 2011. Skills Not Just Diplomas. World Bank, Washington DC. Typically, the literature on decomposi on of student scores in PISA through groups (Amermueller, 2004) and years (Barrera et al., 2011) has focused on the mean differences, with li le a en on to what happens at the tails of the United Na ons Educa onal, Scien fic, and Cultural Organiza on (UNESCO). 2015. Slovakia Country Profile. UNESCO distribu on. The Firpo, For n, and Lemieux (FFL) method allows one to decompose gaps in student performance Ins tute for Sta s cs (UIS). Montreal. not only for the mean but also for other sta s cs of the distribu on. Tradi onally, the problem with quan le Willms, J. D. 2006. “Learning Divides: Ten Policy Ques ons about the Performance and Equity of Schools and regressions has been that the law of iterated expecta ons does not apply, thus making it impossible to interpret Schooling Systems.” UNESCO Ins tute for Sta s cs Working Paper. the uncondi onal marginal effect of each independent variable on a student’s performance. However, recent econometric techniques, such as the one proposed by FFL, have solved this methodological difficulty. The FFL World Bank. 2012. “Diagnos cs and Policy Advice on the Integra on of Roma in the Slovak Republic.” Washington, technique is based on the construc on of re-centered influence func ons (RIF) of a quan le of interest as a DC: World Bank. dependent variable in a regression: World Bank. 2007. “Educa on Quality and Economic Growth.” Washington, DC: World Bank. (1) where is an indicator func on and is the density of the marginal distribu on of scores. A crucial characteris c of this technique is that it provides a simple way of interpre ng the marginal impact of an addi onal unit of a certain factor on students’ PISA scores. Once the uncondi onal quan le regression has been computed for different quan les of the distribu on, the results can be decomposed following the Oaxaca-Blinder approach. page 48 page 49 Table 2A. Decomposi on of reading score gaps between 2012 and 2003 by student achievement group Percen le 20 Percen le 50 Percen le 80 VARIABLES Low Achiever Middle Achiever High Achiever Year 2012 378.4*** 473.1*** 551.7*** -9.991 -4.817 -6.797 Year 2003 396.9*** 469.6*** 536.1*** -4.149 -3.394 -4.316 Difference -18.43* 3.461 15.67** -10.7 -5.837 -7.981 Explained -16.18** 2.955 9.096 -7.78 -4.767 -6.024 Unexplained -2.245 0.506 6.571 -7.279 -3.689 -5.558 Age -0.214 -0.0647 -0.0382 -0.945 -0.293 -0.187 Gender 0.00342 0.148 0.073 -0.837 -0.492 -0.504 ESCS Index (student) 0.808 0.931 1.193 -0.997 -0.844 -0.926 ESCS Index (school) 3.321 3.971 6.442 -3.867 -3.604 -4.986 Romani Speaker -4.857*** 0.599* 1.733*** -1.486 -0.331 -0.591 Repeater -9.261*** -0.788*** 1.230*** -1.919 -0.276 -0.401 Grade -2.400* -1.079* -1.793* -1.333 -0.606 -0.981 Quality of School Resources Index -0.265 -0.879 -1.16 -1.314 -0.702 -0.994 % of Minority Students (School) -3.319* 0.116 1.418* -1.694 -0.402 -0.79 Constant 363.7 -112.7 -365.8** -257.8 -120.3 -181.1 Observa ons 11,753 11,594 11,696 Robust standard errors in parentheses *** p<0.01, ** p<0.05, * p<0.1