Background paper prepared for the Education for All Global Monitoring Report Youth and skills: Putting education to work.

Transcription

1 2012/ED/EFA/MRT/PI/01 Background paper prepared for he Educaion for All Global Monioring Repor 2012 Youh and skills: Puing educaion o work GDP Projecions Eric A. Hanushek & Ludger Woessmann 2012 This paper was commissioned by he Educaion for All Global Monioring Repor as background informaion o assis in drafing he 2012 repor. I has no been edied by he eam. The views and opinions expressed in his paper are hose of he auhor(s) and should no be aribued o he EFA Global Monioring Repor or o UNESCO. The papers can be cied wih he following reference: Paper commissioned for he EFA Global Monioring Repor 2012, Youh and skills: Puing educaion o work For furher informaion, please conac efarepor@unesco.org

2 GDP PROJECTIONS Eric A. Hanushek Ludger Woessmann Inroducion We model a counry s growh rae as a funcion of he skills of workers and oher facors ha include iniial levels of income and echnology, economic insiuions, and oher sysemaic facors. Skills are frequenly referred o simply as he workers human capial sock. growh = α 1 human capial + α 2 oher facors + ε (1) This formulaion suggess ha naions wih more human capial end o coninue o make greaer produciviy gains han naions wih less human capial. The empirical macroeconomic lieraure focusing on cross-counry differences in economic growh has overwhelmingly employed measures relaed o school aainmen, or years of schooling, o es he human capial aspecs of growh models and has ended o find a significan posiive associaion beween quaniaive measures of schooling and economic growh. Average years of schooling is, however, a paricularly incomplee and poenially misleading measure of educaion for comparing he impacs of human capial on he economies of differen counries. I implicily assumes ha a year of schooling delivers he same increase in knowledge and skills regardless of he educaion sysem. For example, a year of schooling in Souh Africa is assumed o creae he same increase in producive human capial as a year of schooling in Korea. Addiionally, formulaions relying on his measure assume ha formal schooling is he primary (sole) source of educaion and ha variaions in non-school facors have negligible effecs on educaion oucomes and skills. This neglec of cross-counry differences in he qualiy of educaion and in he srengh of family, healh, and oher influences is probably he major drawback of such a quaniaive measure of schooling. Our analysis relies on he measures of cogniive skills developed in Hanushek and Woessmann (2009). Beween 1964 and 2003, welve differen inernaional ess of mah, science, or reading were adminisered o a volunarily paricipaing group of counries (see Hanushek and Woessmann (2011) for a review). These include 36 differen possible scores for year-age-es combinaions (e.g., science for sudens of grade 8 in 1972 as par of he Firs Inernaional Science Sudy or mah of 15-year-olds in 2000 as a par of he Programme on Inernaional Suden Assessmen). The assessmens are designed o idenify a common se of expeced skills, which were hen esed in he local language. Each es is newly consruced, unil recenly wih no effor o link o any of he oher ess. Hanushek and Woessmann (2009) describe he consrucion of consisen measures across ess a he naional level across counries hrough empirical calibraion of he differen ess. By ransforming he means and variances of he original counry scores (parly based on exernal longiudinal es score informaion

3 available for he Unied Saes), each is placed ino a common disribuion of oucomes. Each age group and subjec is normalized o he PISA sandard of mean 500 and individual sandard deviaion of 100 across OECD counries, and hen all available es scores are aggregaed o he counry level. We inerpre he es scores as an index of he human capial of he populaions (and workforce) of each counry. Our cross-counry regressions of economic growh follow an expanding lieraure which, over he pas en years, demonsraes ha consideraion of cogniive skills dramaically alers he assessmen of he role of educaion and knowledge in he process of economic developmen. Hanushek and Woessmann (2009) exend he empirical analysis o incorporae 50 counries ha have paricipaed in one or more inernaional esing occasions beween 1964 and 2003 and have aggregae economic daa for he period We esimae his model is wo differen forms. In he firs, we regress he average annual growh rae from for GDP per capia on iniial income level (GDP per capia in 1960), school aainmen in 1960, and cogniive skills measured by average inernaional es scores. In he second, insead of average es scores, we use measures of he ails of he disribuion he percenage of sudens scoring a leas 400 poins (one sandard deviaion below he OECD mean of 500) and he percenage scoring greaer han 600 poins (one sandard deviaion above he OECD mean of 500). The basic esimaes used for our projecions are shown in Table 1. Overview of Projecions The economic projecions involve several componens. Firs, we calculae he ime pah of he annual growh rae engendered by educaion reform designed o move sudens from heir curren performance o a given new level. This paern of economic oucomes represens he confluence of hree separae dynamic processes: (1) Changes in schools lead o he progressive improvemen in suden achievemen unil sudens fully reach he new seady-sae level of achievemen; (2) sudens wih beer skills move ino he labor force and he average skills of workers increase as new, higher achieving workers replace reiring workers; and (3) he economy responds o he progressive improvemen of he average skill level of he workforce. Second, based on he paern of prediced growh raes, we model he fuure developmen of GDP wih and wihou he educaion reform. Third, based on hese projecions, we calculae he oal value of he reform by aggregaing he discouned values of he annual differences beween he GDP wih reform and he GDP wihou reform. For he purposes here, reforms are assumed o ake 20 years o complee, and he pah of increased achievemen during he reform period is aken as linear. For example, an average improvemen of 25 poins on PISA is assumed o reflec a gain in he suden populaion of 1.25 poins per year. This migh be realisic, for example, when he reform relies upon a process of upgrading he skills of eachers eiher by raining for exising eachers or by changing he workforce hrough replacemen of exising eachers. This linear pah dicaes he qualiy of new cohors of workers a each poin in ime.

4 The expeced work life is assumed o be 40 years, which implies ha each new cohor of workers is 2.5 percen of he workforce. Thus, even afer an educaional reform is fully implemened, i akes 40 years unil he full labor force is a he new skill level. The benchmark here considers all economic reurns ha arise during he lifeime of a child ha is born a he beginning of he reform in According o he mos recen daa (ha refer o 2006), a simple average of male and female life expecancy a birh over all OECD counries is 79 years (Organisaion for Economic Co-operaion and Developmen (2009c)). Therefore, he baseline calculaions ake a ime horizon unil 2090, considering all fuure reurns ha accrue unil hen, bu neglecing any reurns ha accrue afer Finally, more immediae benefis are boh more valuable and more cerain han hose far in he fuure. In order o incorporae his, he enire sream is convered ino a presen discouned value. In simples erms, he presen discouned value is he curren dollar amoun ha would be equivalen o he fuure sream of reurns calculaed from he growh model. If we had ha amoun of funds and invesed i oday, i would be possible o reproduce he fuure sream of economic benefis from he principal amoun and he invesmen reurns. Thus, his calculaion of presen discoun value allows a relevan comparison for any oher curren policy acions. In doing so, he discoun rae a which o adjus fuure benefis becomes an imporan parameer. A sandard value of he social discoun rae used in long-erm projecions on he susainabiliy of pension sysems and public finance is 3 percen (e.g., Börsch-Supan (2000), Hagis, Klusen, Plae, and Raffelhüschen (2005)), a preceden ha is followed here. By conras, he influenial Sern Review repor ha esimaes he cos of climae change uses a discoun rae of only 1.4 percen, hereby giving a much higher value o fuure coss and benefis (Sern (2007)). A number of unesed assumpions go ino he projecions. Firs, hey assume ha skills play he same role in he fuure as hey have in he pas, so ha he evidence of pas resuls provides a direc way o projec he fuure. Second, while he saisical analysis did no look a how economies adjus o improved skills, he calculaions assume ha he experience of oher counries wih greaer cogniive skills provide he relevan insigh ino how he new skills will be absorbed ino he economy. Third, he projecion of simulaneous improvemen across counries presumes ha all counries can grow faser wihou deracing from (or benefiing) growh in oher counries. In oher words, he higher levels of human capial in each counry allow i o innovae, o improve is producion, and o impor new echnologies wihou deracing from he growh prospecs for oher counries. Furher, he esimaes ignore any oher aspecs of ineracions such as migraion of skilled labour across borders. (Of course, one way ha a counry could improve is human capial would be by arranging for is youh o obain schooling in anoher counry wih beer schools as long as he more educaed youh reurn o heir home counry o work). Fourh, all counries are assumed o have a saionary populaion wih a consan age disribuion. Finally, all calculaions are in real (inflaion-adjused) erms 2010 dollars under purchasing power pariy.

5 Deails of Projecions The economic impac of reform varies across four phases ha are defined by he average qualiy of he labor force. The reform plan we projec has several disinc phases defined by he composiion of he work force in each year over he 80 year period. a) Phase 1 ( ): During he 20 years of he educaion reform program, he addiional growh in GDP per capia due o he reform in year is given by: 1 working life growh coefficien * PISA * * (3) where he growh coefficien comes from he regression esimaions presened in he previous secions and ΔPISA is he increase in he average PISA es score due o he reform. (In he range of projecions we also consider a en-year reform program; i.e., a program ha reaches he new achievemen levels in jus 10 years insead of he 20 years described here). The working life erm indicaes ha each cohor of new, higher achieving sudens is only a fracion of he oal labor force. b) Phase 2 ( ): The educaion reform is now fully enaced, and achievemen of all subsequen sudens remains a he new level. Bu for he lengh of a work life from he sar of reform, which in he baseline simulaions is assumed o las 40 years, here are sill workers wih iniial levels of skills ha are being replaced in reiremen by higher achieving workers. During his phase, he addiional growh in GDP per capia in year due o he reform is given by: 1 working life 1 growh coefficien * PISA * (4) c) Phase 3 ( ): During his phase, he firs 20 labor-marke cohors which only parially profied from he educaion reform are replaced by cohors ha profied from he fully enaced educaion reform: growh coefficien * PISA * (5) working life d) Phase 4 (afer 2070): Finally, he whole workforce has gone hrough he reformed educaion sysem. The annual growh rae is now increased by he consan long-run growh effec Δ: The level of GDP wih and wihou reform growh coefficien * PISA (6) a) Wihou reform, he economy grows a he consan growh rae of poenial GDP: GDP no reform GDP 1 no reform * 1 poenial growh (7)

6 b) Wih reform, he annual growh rae is addiionally increased by he growh effec Δ : GDP reform Resuls of Projecions GDP 1 reform * 1 poenial growh The resuls are displayed in Table 2. We show he overall resuls for eigh separae reform pahs. Specifically we have a more modes and more aggressive reform plan for each of he goals (change in average performance and improvemen a he boom end). We also in each insance consider a 20-year and a 10-year reform pah. The mos modes reform plan would call for improving average performance by 25 poins (PISA equivalen) or one-quarer sandard deviaion, and i would do his improvemen over a 20 year period. As Table 2, his modes reform would yield addiions o GDP over he nex 80 years ha were worh over 300 percen of curren GDP. An aggressive program of 50 poin improvemen over 20 years would have a presen value of 664 percen of curren GDP. The programs of improvemen a he boom end of he achievemen disribuion also have large gains in he economy. A 20 percen improvemen in he proporion of sudens reaching level 1 (i.e., reaching 400 PISA poins) would yield higher GDP equal o 342 percen of curren GDP even wih a 20 year reform program. Wih a 10-year reform, he gains would be over four imes curren GDP. Table 2 also suggess how long run growh will change wih improved achievemen. A he low end of he able, a 25-poin improvemen in scores will lead o ½ percen higher annual growh enormous amoun when compounded over he lifeime of somebody born oday. Conclusions Pas hisory suggess he possibiliy of enormous gains in aggregae economic oucomes from improving suden learning. The improvemens in GDP come from he relaionship beween learning and growh of an economy. The simple idea embodied in he models is ha a beer educaed workforce is consisen wih produciviy gains and greaer innovaion, and hese have huge impacs on he fuure well-being of sociey. (8)

7 Table 1: Cogniive Skills as Deerminan of Growh of Income per Capia, (1) (2) GDP per capia (5.54) (5.12) Years of schooling (0.34) (0.28) Tes score (mean) Share of sudens above hreshold of 400 Share of sudens above hreshold of 600 (9.12) (3.61) (4.35) Consan (5.54) (2.97) N R 2 (adj.) Dependen variable: average annual growh rae in GDP per capia, saisics in parenheses.

8 Table 2. Presen Value of fuure gains relaive o curren GDP Speed of reform: Change in long-run growh rae 20 years 10 years (independen of duraion of reform) +25 poins on PISA 308% 375% 0.50% +50 poins on PISA 664% 815% 0.99% +20 percen more a 400 poins 342% 417% 0.55% +75 percen more a 400 poins 1634% 2048% 2.05%