International Data on Educational Attainment Updates and Implications

Transcription

1 Inernaional Daa on Educaional Aainmen Updaes and Implicaions Rober J. Barro Economics Deparmen, Harvard Universiy and Jong-Wha Lee Economics Deparmen, Korea Universiy Augus 2000 We are graeful o Francis Green, Hilary Seedman, and wo anonymous referees for helpful commens and suggesions. We also hank he paricipans of he "Skill Measuremen and Economic Analysis- An Inernaional Conference," a Universiy of Ken a Canerbury, England, March 2000, for heir commens. This work was prepared while Jong-Wha Lee was visiing he Cener for Inernaional Developmen a Harvard Universiy in He hanks he Cener for is hospialiy. This research has been suppored by he U.S. Naional Science Foundaion. The daa are available a he web sie of he Cener for Inernaional Developmen a Havard Universiy, hp://

2 ABSTRACT This paper presens a daa se ha improves he measuremen of educaional aainmen for a broad group of counries. We exend our previous esimaes of educaional aainmen for he populaion over age 15 and over age 25 up o 1995 and provide projecions for We discuss he esimaion mehod for he measures of educaional aainmen and relae our esimaes o alernaive inernaional measures of human capial socks. Rober J. Barro Deparmen of Economics Harvard Universiy Liauer 120, Cambridge MA USA ) rbarro@harvard.edu Jong-Wha Lee Economics Deparmen Korea Universiy Anam-Dong, Sungbuk-Ku Seoul , Korea )jongwha@korea.ac.kr JEL No. I20, J24, O15

3 1. Inroducion Human capial, paricularly ha aained hrough educaion, has been emphasized as a criical deerminan of economic progress. A greaer amoun of educaional aainmen indicaes more skilled and more producive workers, who in urn increase an economy's oupu of goods and services. An abundance of well-educaed human resources also helps o faciliae he absorpion of advanced echnology from developed counries. In addiion, he level and disribuion of educaional aainmen has a srong impac on social oucomes, such as child moraliy, feriliy, educaion of children, and income disribuion. In order o quanify he relaionship beween educaional aainmen and economic and social oucome variables, here have been a number of aemps o measure educaional aainmen across counries. Earlier empirical sudies used school enrollmen raios or lieracy raes. These daa are widely available bu do no adequaely measure he aggregae sock of human capial available conemporaneously as an inpu o producion. Our earlier sudy filled his gap in daa by measuring educaional aainmen for a broad group of counries. 1 The figures were consruced a five-year inervals from 1960 o 1985 (Barro and Lee, 1993). The daa showed he disribuion of educaional aainmen of he adul populaion aged 25 and over by sex a seven levels of schooling. In our subsequen work (Barro and Lee, 1996), we provided an updae on he daa for he populaion aged 25 and over up o 1990 and also consruced esimaes of aainmen for 1 Earlier aemps o consruc measure of educaional aainmen for inernaional comparisons include Psacharopoulos and Ariagada (1986), Lau e al. (1991), and Nehru e al. (1995). Their measures were hampered by more limied coverage of daa and by poenially larger measuremen errors han ours (see he discussion in foonoe 1 of Barro and Lee (1996)). 1

4 he populaion aged 15 and over. The over-15 age group corresponds beer o he labor force for many developing counries. In his paper, we provide a furher updae of educaional aainmen up o1995 and also consruc projecions o This daa se improves on our earlier esimaes in a number of respecs. Firs, our fill-in procedure for missing census/survey observaions now uses gross enrollmen raes, adjused for repeaers. This measure reflecs he inflows of new school graduaes o exising educaional socks more accuraely han he usual gross or ne enrollmen raios. Second, in he consrucion of average years of schooling, we now ake accoun of changes of school duraion over ime wihin counries. Our daa se on educaional aainmen provides good measures of he human capial sock for a broad number of counries. 2 The daa se comprises a leas one observaion for 142 economies, of which 107 have complee informaion a five-year inervals from 1960 o The percenage of he populaion who have successfully compleed a given level of schooling say secondary, eriary, or pos-primary schooling is a sraighforward way o show he populaion's aainmen of skills and knowledge associaed wih a paricular level of educaion. In pracice, however, each cycle of educaion has significan variaion in duraion across counries. We ake accoun of his variaion by using informaion on he ypical duraion of each level of schooling wihin counries. Wih hese daa we can also consruc measures of average years of schooling a all levels for each counry. 2 Our esimaes of educaional aainmen have been used in many previous sudies. Empirical work ha uses our daa se includes Rajan and Zingales (1998) and Ramey and Ramey (1995) in American Economic Review; Barro (1999) in Journal of Poliical Economy; Easerly and Levine (1998) and Hall and Jones (1999) in Quarerly Journal of Economics; and Sachs and Warner (1995) in Brookings Papers on Economic Aciviy. Up o February 2000, our papers on 2

5 Because human capial is mulifaceed and includes a complex se of human aribues, he sock of human capial held by individuals is hard o measure wih precision in a quaniaive form. Educaional aainmen is a bes a proxy for he componen of he human capial sock obained a schools. In secion 4, we relae our measure of educaional aainmen o alernaive measures of human capial, such as inernaional es scores, resuls from he inernaional adul lieracy survey, and esimaes of labor-marke oucomes. We also compare our numbers o he OECD's recen esimaes of educaional aainmen. 2. The Esimaion of Educaional Aainmen This secion summarizes he main procedure for consrucing he esimaes of human capial sock and discusses he modificaions ha have been made in he presen updae Esimaion Mehod The esimaion procedure is a perpeual invenory mehod, using he census/survey observaions on aainmen as benchmark socks and new school enrans as flows ha are added o he socks wih an appropriae ime lag. The benchmark figures on school aainmen have been colleced, as before, from census/survey informaion, as compiled by UNESCO and oher sources. The census/survey figures repor he disribuion of educaional aainmen in he populaion sraified by age and by sex in, for mos cases, six caegories: no formal educaion, incomplee primary, complee primary, firs cycle of secondary, secondary cycle of educaional aainmen daa (Barro and Lee,1993 and 1996) have been cied a leas 90 imes, according o he Social Science Ciaions Index. 3

6 secondary, and eriary. 3 The presen daa se adds more census informaion for 1990 and 1995 and updaes he esimaes of educaional aainmen o This daa se also includes new observaions for he ransiional economies ha became independen in he 1990s. Tables 1 and 2 summarize he coverage of he census daa for wo groups of he oal populaion aged 15 and over and aged 25 and over. Table 1 shows he disribuion of counries by he number of census/survey observaions available since For persons aged 15 and over, 141 counries have a leas 1 observaion, and 72 counries have 3 or more observaions. Table 2 shows he disribuion of numbers of observaions by census-survey year since 1950 (where he underlying figures are applied o he neares 5-year value). For he populaion over age 15, 59 observaions are available for 1960, 62 for 1970, 75 for 1980, and 56 for The oal number of observaions is 379, of which 354 are beween 1960 and These 354 daa poins are used for he benchmark figures on educaional aainmen. We fill in mos of he missing observaions on school aainmen by using informaion on school-enrollmen raios and he srucure of populaion by age groups. Following he perpeual invenory mehod, we consruc curren flows of adul populaion ha are added o he benchmark socks. The formulas for he various levels of schooling for he populaion aged 15 and over are as follows. 4 H H 1 δ ) + L15 *(1 PRI ) (1) 0. = 0, 5 ( 5 3 The classificaion of levels of educaion is based on he common crierion for inernaional comparison ha was adoped in 1976 as he Inernaional Sandard Classificaion of Educaion (ISCED, 1976). Counries may no always classify diplomas and qualificaions a he same ISCED levels, so he UNESCO classificaion may differ from he naional classificaion. 4

7 H 1. = H1, 5 ( 1 δ ) + L15 *( PRI 5 SEC ) (2) H = δ L20 * HIGH (3) 2. H 2, 5 (1 ) + L15 * SEC H = δ L20 * HIGH (4) 3, H 3, 5 (1 ) + where H j is he number of persons aged 15 and over for whom j is he highes level of schooling aained; j=0 for no school, 1 for primary, 2 for secondary, and 3 for higher. L15 denoes he number of persons aged and L20 denoes he number of persons aged The variables PRI, SEC, and HIGH are he enrollmen raios for primary, secondary, and higher schools, respecively. The enrollmen raios are he gross raios adjused for school repeaers, as we will explain in he nex secion. The assumpion in equaions (1)-(4) is ha persons aged would have received primary schooling 5 years earlier (if hey were enrolled) and would receive secondary educaion conemporaneously. Persons aged would receive higher schooling conemporaneously. The variable * is he moraliy rae for persons aged 15 and over and is esimaed from δ = ( L15 + L 5 L ) / L 5 (5) where L is he oal populaion aged 15 and over. The assumpion in equaions (1)-(4), which is no enirely accurae, is ha he moraliy rae is independen of he level of schooling aained. 5 4 For he procedure applied o he populaion aged 25 and over, see Barro and Lee (1993, par II). 5 In a ypical counry in which educaional aainmen is growing, moraliy would be higher for he older people who are less educaed. Then he assumpion of uniform moraliy can cause a downward bias in he esimaion of he oal educaional sock. Unforunaely, he limiaion of 5

8 Subsiuion from equaion (5) ino equaions (1)-(4) yields he formulas ha we use o fill in he cells for he aainmen raios, h j = H j /L : h H L = h [1 ( L15 / L )] + ( L15 / L ) *(1 PRI ) (6) 0, 0. / 0, 5 5 h 1, H1. / L = h1, 5 5[1 ( L15 / L )] + ( L15 / L )* ( PRI SEC ) (7) h H / L h [1 2, 2. = 2, 5 ( L15 / L )] + ( L15 / L ) * SEC ( L20 / L )* HIGH (8) h ( L20 / L ) * HIGH (9) 3, H 3, / L = h3, 5[1 ( L15 / L )] + The applicaion of equaions (6)-(9) generaes our panel daa se on educaional aainmen for he populaion aged 15 and over from 1960 o 2000 a he four broad levels of schooling: no school, some primary, some secondary, and some higher. 6 We hen break down he hree levels of schooling ino incomplee and complee educaion by using esimaes of compleion raios. 7 We use he same sources and mehodology o consruc a panel daa se on he educaional aainmen of females over age 15 and over age 25 a five-year inervals from 1960 o The daa on he disribuion of educaional aainmen among he populaion, combined wih he informaion for each counry on he duraion of school a each level, daa on age-specific educaion levels and moraliy raes by age group do no allow us o compue specific moraliy raes of populaion by levels of educaion. 6 Daa on enrollmen raes are available from UNESCO and daa on he srucure of populaion by age are from Unied Naions. For he projecions of educaional aainmen for 2000, we use he projecions on he populaion srucure for 2000 provided by he Unied Naions. 7 The compleion raios for he age 25 and over populaion group are described in Barro and Lee (1993, Tables 4, 5, and 7). The compleion raios for he age 15 and over group are reconsruced from hose for he age 25 and over group. The adjusmen is ha secondaryschool enrollees aged are reaed as incompleely educaed a he secondary level, and higher-school enrollees aged are reaed as incompleely educaed a he higher level. 6

9 generae he number of years of schooling achieved by he average person a he various levels and a all levels of schooling combined Modificaions The presen esimaes and projecions improve on earlier daa by making a few modificaions in he fill-in procedure of missing observaions and in he consrucion of average years of schooling Adjused gross enrollmen raios for he fill-in procedure Our procedure for filling in missing observaions requires school enrollmen raes. In he firs version of he daa se (Barro and Lee, 1993), we used he gross enrollmen raio, which is defined as he raio of all persons enrolled a a given level of schooling o he populaion of he age group ha naional regulaion or cusom dicaes should be enrolled a ha level. For example, he oal regisered sudens in primary school are ypically compared wih he populaion aged 6-1l years. However, he gross raios will oversae he accumulaion of human capial when sudens repea grades. In he nex revision (Barro and Lee, 1996), we used he ne enrollmen raio, which is defined as he raio of sudens a a given level of schooling in he designaed age group o he oal populaion of ha age group. For example, he regisered sudens aged 6-11 years in primary school are compared wih he oal populaion aged 6-11 years. Accordingly, his concep does no coun sudens in primary school who are younger han 6 or older han 11. 7

10 Ne enrollmen raios also inroduce measuremen error if here are under- or over-aged children saring each level of educaion. If a proporion of he populaion enrolls a a given level a an age consisenly earlier or laer han ha prescribed by regulaion and consequenly finishes school earlier or laer, hen ne enrollmen raios undersae he accumulaion of schooling capial. Sudens who sar school early or lae consiue a significan fracion of oal enrollmens in many counries. For insance, for 12 Sub-Saharan African counries, a large fracion beween 10 and 33%--of enrollmen in grade 1 of primary educaion is made up of sudens younger han he official enrance age. In 31 counries, a leas 50% of he sudens enrolled in grade 1 of primary school are older han he prescribed enry age (Gajraj, 1992). Alhough he over-aged sudens enrolled in his grade are due in par o repeiion, he incidence of a high proporion of over-aged sudens in counries where he proporion of repeaers is small demonsraes he imporance of lae enry. In response o his problem, we have revised our fill-in procedure o use 'adjused gross enrollmen raios,' insead of he usual gross or ne raios, for primary and secondary schools. (We use he gross raios for eriary schools.) We consruc he new raios by adjusing he gross enrollmen raios for repeiion raes Changes in he duraion of years of schooling In he previous daa se, we provided average years of schooling a each level and a all levels of schooling combined as an aggregae measure of he educaional sock. To 8 See Lee and Barro (1997) for a discussion of he repeiion daa. Alhough a proper accouning of he ne inflows o he human capial sock would also deduc dropous from gross enrollmens, he effec from dropous is already aken ino accoun by our measures of incomplee aainmen a primary and secondary schools. 8

11 consruc he average years of schooling, we used each counry s ypical duraion of years of schooling a each level. For example, we assumed ha he populaion who compleed primary school obained he counry s full duraion of primary schooling (Barro and Lee, 1993, p.387). However, since our procedure used each counry s duraion daa for 1965, we did no ake accoun of changes over ime in a counry s ypical duraion of schooling a each level. This approach creaes measuremen error, because a significan number of counries have alered he duraions of heir school sysems. Over he las hree decades, 32 counries have changed a leas once he ypical duraion of schooling a he primary or secondary levels (see UNESCO, Saisical Yearbook, various years). 9 In he presen revision, we ake accoun of changes in he duraion of schooling wihin counries. These changes would ake ime o become effecive because he pupils who have received educaion in he new sysem will be new enrans ino he adul populaion wih a long ime lag. We assume ha changes in he duraion of schooling a primary level apply o new enrans ino he populaion over age 15 in 10 years and ino he populaion over age 25 in 20 years, while he changes a secondary level become effecive for he populaion over age 15 in 5 years and for he populaion over age 25 in 15 years,. 3. The Complee Daa Se on Educaional Aainmen, The direcion of changes in duraion varies for individual counries. For some counries, duraion of schooling a he primary or secondary levels increased, while i decreased for ohers. For insance, El Salvador increased duraion of primary schooling from 6 o 9 years in In conras, Sri Lanka reduced duraion of primary schooling from 8 o 5 years in 1970 (and hen changed i again o 6 years in 1978). For mos cases, changes in duraion a he primary and secondary schooling occurred ogeher; for example, when primary duraion increased, secondary duraion decreased, and vice versa. 9

12 Appendix Tables A.1- A.2 conain he full daa se on aainmen a he seven levels of schooling for he overall populaion over age 25 and over age 15. Appendix Tables A.3- A.4 repor he same esimaes for he female populaion. The Tables also repor he average number of years of schooling achieved by he average person in each counry. 10 The daa se comprises a leas one observaion for 142 economies. Table 3 summarizes he schooling daa by region from 1960 o 2000 for 107 counries ha have complee informaion. 11 The able considers hree broad groups 23 advanced counries, 11 ransiional economies (TEs), and 73 developing counries. The developing group is furher broken down ino five regions: Middle Eas/Norh Africa (11 counries), Sub-Saharan Africa (22), Lain America/Caribbean (23), Eas Asia/Pacific (10), and Souh Asia (7). Regional averages are compued by weighing each counry's observaion by is share of he regional populaion aged 25 and over. Table 3 shows ha educaional progress has coninued in mos regions excep he ransiional economies in he 1990s. For he 23 advanced counries, he average years of schooling for persons aged 25 and over increased from 9.3 years in 1990 o 9.6 years in 1995 and are forecased o increase furher o 9.8 years in For he 73 developing counries, he average years of aainmen grew by 23% from 4.0 years in 1990 o 4.9 years in In he 11 TEs, he average years of schooling remained a he level of 10.0 years in he 1990s. Among he developing regions, Sub-Saharan Africa and Middle 10 The Appendix Tables and he complee panel daa se including he average number of years of schooling a he various levels are available from he web sie of he Cener for Inernaional Developmen a Harvard Universiy (hp://www2.cid.harvard.edu/ciddaa/). 11 New ransiional economies including Croaia, Czech Republic, Slovakia, and Slovenia are included as independen counries since 1995, replacing he former Yugoslavia and Czechoslovakia. The former Sovie Union is replaced by he Russian Republic since

13 Eas/Norh Africa showed he sronges progress, growing by abou 35% in he 1990s from 2.8 o 3.8 years and from 3.8 o 5.1 years, respecively. Alhough he educaional progress of he developing region as a whole has been relaively srong over he las decade, he educaion level in his region sill lags far behind he advanced and TE regions. For he developing counries, he projecions show ha in 2000, 37% of he populaion aged 25 and over sill had no formal schooling and only 27% had some secondary educaion. For he advanced and ransiion counries groups, in conras, abou wo-hirds of he populaion over age 25 had a leas some secondary schooling. Table 4 presens he ime-series daa on educaional aainmen by group for he populaion aged 15 and over. The regional paerns are similar o hose for he populaion aged 25 and over. For mos regions, he average years of schooling for he aged 15 and over populaion group are slighly higher han hose for he aged 25 and over group. For he developing counries as a whole in 2000, he populaion aged 15 and over is projeced o have an average school aainmen of 5.1 years, compared o 4.9 years for he populaion aged 25 and over. We have also consruced he daa on educaional aainmen for he female populaion over age 25 and over age 15. Table 5 uses hese resuls and compares he regional figures for females wih hose for males for he period from 1960 o 2000 by region. The able shows he "gender raio," defined as he raio of female o male aainmen (expressed as a percenage). In mos counry groups, here has been a subsanial rise in gender equaliy over he las decade. In he developing counries as a whole, he raio rose from 63% in 1990 o 67% in 1995 and 70% in 2000 for he 11

14 populaion aged 25 and over. The raio increased from 69% in 1990 o 73% in 2000 for he populaion aged 15 and over. During his period, he raios rose noably in Middle Eas/Norh Africa (57% o 68% for he populaion over age 25) and Souh Asia (44% o 53% for he same age group). Noe, however, ha he gender gap in educaion is sill large in hese regions, compared o hose in he advanced, TEs, and Lain American regions, where he raios are ypically well over 90%. Among all regions, he ransiional economies group is forecased o have he highes gender raio, reaching 105% in 2000 for he populaion aged 15 and over. Despie he upward rends of he raios in mos regions, he "world" average raio remained sable a around 85% over he 1990s. This apparen inconsisency is aribued o he fac ha since 1995 he "world" does no include he former Sovie Union counries excep Russia, which have relaively high gender raios Comparisons wih Alernaive Measures of he Human Capial Sock Our esimaes of educaional aainmen provide a reasonable proxy for he sock of schooling capial for a broad group of counries. However, hese daa have a number of shorcomings. Firs, he measure of educaional aainmen does no ake accoun of he skills and experience gained by individuals afer heir formal educaion. Second, he measure does no direcly measure he human skills obained a schools and, specifically, does no ake accoun of differences in he qualiy of schooling across counries. In his secion, we compare our concep of educaional aainmen wih alernaive measures of he sock of human capial. 12 When simple averages raher han populaion-weighed averages are used, he gender raios for he populaion aged 15 and over rose from 86.4 in 1990 o 88.1 in 1995 and hen slighly 12

15 4. 1. Inernaional Tes Scores by Sudens An alernaive measure of educaional sock, which is considered o reflec variaions of educaional qualiy across counry, is scores on inernaionally comparable ess of educaional achievemen. Concepually, he qualiy of educaion is refleced in he performance of sudens and graduaes. One indicaor of schooling qualiy is sudens scores on inernaionally comparable ess. Hanushek and Kimko (2000) consruced a cross-secion daa se of inernaional es scores in he subjecs of science and mahemaics for 39 counries. Lee and Barro (1997) have compiled es scores on he examinaions in science, mahemaics, and reading ha have been conduced in various years for up o 58 counries by he Inernaional Associaion for he Evaluaion of Educaional Achievemen (IEA) and he Inernaional Assessmen of Educaional Progress (IAEP). These sudies cover primary or secondary sudens of he same age or grade group, such as age 9, age 13, and pupils in he las year of secondary educaion. Since he ess are carefully designed o ensure inernaional comparabiliy, he es scores would capure he cross-counry variaions in cogniive skills of he sudens and hereby measure differences in he qualiy of he fuure labor force. Hanushek and Kimko (2000) and Barro (2000) find ha es scores are posiively relaed o growh raes of real per capia GDP in cross-counry regressions. One shorcoming of hese daa, however, is ha he observaions apply o differen years and are mos abundan for he 1990s. The sample sizes are also much smaller han hose for average years of schooling. Table 6 shows he average es scores on mahemaics and science for he sudens a he sevenh grade for 39 counries ha paricipaed in he Third Inernaional increased o 88.7 in

16 Mahemaics and Science Sudy (TIMSS) in 1994 and The scores are expressed on proficiency scales (scales ranging from 0 o 1000, wih a mean of 500 and a sandard deviaion of 100). Among he 39 paricipaing counries, sudens in he four Eas Asian economies, Singapore, Korea, Japan, and Hong Kong, showed he highes achievemen in Mahemaics. Their average scores, ranging from 564 for Hong Kong o 601 for Singapore, are well above he inernaional average of 484. In conras, Souh Africa and Colombia performed significanly less well han he oher counries, wih means below 400. The paern of he counry performances is similar for science achievemen. The correlaion beween he wo es scores is The correlaion beween hese sudens' es scores and average years of schooling is no very high. The correlaion of average years of schooling in 1995 wih he mahemaics or science es score is Boh average years of schooling and es-scores variables urn ou o be saisically significan in cross-counry growh regressions (Hanushek and Kimko, 2000). This resul suggess ha boh variables provide useful informaion abou a counry s educaion sock he es scores for qualiy of schooling and he average years of schooling for quaniy of schooling Inernaional Adul Lieracy Tes Alhough he es scores of sudens reflec he qualiy of schooling and, hence, indicae he qualiy of he labor force, hey do no direcly measure he educaional capial held by a counry s working-age populaion. Knowledge can be gained or los afer he compleion of formal educaion. Ideally, ess of cogniive abiliy would be adminisered o aduls, as well as o sudens. Recenly, he Inernaional Adul Lieracy Survey 14

17 (IALS) has, for he firs ime, provided a direc comparison of cerain work-relaed skills in he adul populaions of various counries. Alhough lieracy was previously reaed as a dichoomous concep eiher lierae or no he OECD sudy considered he mulifaceed naure of lieracy and made effors o measure direcly he skills of he adul populaion. 13 Lieracy was measured in hree domains prose lieracy, documen lieracy, and quaniaive lieracy. The lieracy skills were evaluaed in each counry hrough deailed inerviews wih a sample of 2,000 and 8,000 persons in he working-age populaion. Resuls were repored on a scale from 0 o 500 in each of he hree domains. 14 The IALS sudy is sill in early sages and covers only 12 OECD counries. Seven counries paricipaed in he firs Survey Canada, Germany, he Neherlands, Poland, Sweden, Swizerland (French-speaking and German-speaking, separaely), and he Unied Saes. Five addiional counries paricipaed laer Ausralia, Belgium (Flanders), Ireland, New Zealand, and he Unied Kingdom (see OECD and Saisics Canada, 1995, and OECD and Human Resources Developmen Canada, 1998). In addiion o he limied coverage of he daa, he IALS sudy has been criicized BECAUSE OF measuremen errors and problems in modeling echniques (see Blum, Goldsein, and Gu rin-pace, 2000). Table 6 shows he mean scores in each of he hree domains for he 12 OECD counries. Sweden was he bes performer in all domains, while Poland was he wors in 13 The IALS defines "lieracy" in erms of paricular skills of aduls "using prined and wrien informaion o funcion in sociey, o achieve one's goals, and o develop one's knowledge and poenial" (OECD and Saisics Canada, 1995, p.3). 14 For analyical purposes, he scores were also repored by five levels of lieracy: Level 1 (0 o 225 poins) indicaes persons wih very poor lieracy skills, while Levels 4-5 ( poins) 15

18 all domains. The scores among he hree domains are highly correlaed, wih correlaion coefficiens over One surprising observaion is he significan discrepancy in achievemen beween sudens and aduls. Figure 1 depics he TIMMS mahemaics score and he IALS quaniaive lieracy score. The correlaion beween he wo scores is only 0.07 in he common sample of 11 counries. In he scaer plo, however, Sweden seems o show an excepionally large gap beween he wo es scores. The Swedish aduls were he bes performers in he quaniaive lieracy es, bu he sevenh-grade sudens showed very poor performance in he mahemaics es in Sudies based on he IALS daa show ha hese measures of lieracy have a significan relaion wih gross earnings of individuals wihin a counry, independenly from he effecs of educaion (OECD and Human Resources Developmen Canada, 1998) Esimaes of he Marke Value of Human Capial Anoher approach o measuring he sock of human capial is o assess differences in labor-marke oucomes, such as earnings or employmen of educaed workers, ha are aribuable o he individuals' schooling. Along his line, here have been aemps o measure he human capial sock in erms of is marke value. A labor-income-based measure is derived by weighing differen segmens of he workforce by he raio of indicae ha persons have he capaciy o use higher-order hinking and informaion-processing skills. 15 The Swedish sudens a age 13 also showed below-average achievemen in he Second Inernaional Mahemaics Sudy ha was conduced for 20 counries in In conras, in he same es, he 12 h grade sudens did very well, wih he average score only lower han ha of 16

19 earnings a differen levels of educaion. Gollop and Jorgenson (1980), Jorgenson, Gollop, and Fraumeni (1987), and Mulligan and Sala-I-Marin (1995) used his mehod o esimae he human-capial sock of he Unied Saes. This measure, however, is no widely applicable o inernaional comparisons due o he limied availabiliy of deailed wage daa for mos counries. A relaed issue is wheher esimaed raes of reurn, raher han duraions of school, should be used as weighs in he consrucion of aggregae human capial. Collins and Bosworh (1996) and Priche (1996) employed his mehod o consruc measures of aggregae educaional aainmen. Since he esimaed raes of reurn measure produciviy of educaion a each level, his measure is concepually beer han average years of schooling. A measure of average years of schooling assumes unrealisically ha a year of educaion adds a consan quaniy of human capial, wheher underaken by a primary pupil or a college suden. One problem wih he measure aggregaed by raes of reurn, however, is ha he available esimaes of reurns o schooling are no always accurae. For insance, he esimaes may over-sae reurns because of he likely posiive correlaion beween schooling and unmeasured characerisics such as abiliy. The significan variaions in he available esimaes for an individual counry show ha he esimaes of reurns are very sensiive o assumpions abou social benefis, forgone earnings of sudens, and so on OECD Esimaes of Educaional Aainmen Finland among he OECD counries. This fac suggess ha sudens aain significan quaniaive skills a he secondary level. 17

20 The OECD has also compiled daa on educaional aainmen of he populaion in he 1990s. Mos daa come from labor-force surveys based on samples of households or individuals, in conras o he naional censuses in he UNESCO daabase (see OECD, 1998b, pp ). For he years around 1995, he OECD daa se covers 32 counries 24 OECD counries and 8 non-oecd counries. This daa se presens he percenages of he populaion or labor force aged beween 25 and 64 for whom he highes compleed level of educaion falls ino four caegories: below upper-secondary educaion, upper-secondary educaion, non-universiy eriary educaion, and universiy-level educaion. 16 One advanage of hese esimaes is ha he populaion group covered corresponds closely o he labor force in he ypical OECD counry. However, he age group is less appropriae for developing counries, in which a subsanial fracion of persons aged beween 15 and 24 are in he workforce. The OECD esimaes of educaional aainmen correspond reasonably well o our numbers. Table 7 compares he wo measures in erms of he percenage of he adul populaion aged 25 o 64 for he OECD and aged 25 and over in our measure by he level of schooling. The firs OECD caegory below upper secondary should correspond roughly o he sum of our firs hree caegories: no schooling, primary schooling, and parial secondary. In he common sample of 31 counries, he correlaion beween he wo esimaes is The OECD esimaes of he populaion aaining upper secondary or higher educaion are also highly correlaed wih our measure of average years of schooling. The correlaion is The OECD publicaion also provides esimaes for educaional aainmen of he populaion by sex and by disaggregaed age groups, 25-34, 35-44, 45-54, and

21 Noneheless, here are significan differences beween he OECD and our daa for some counries. For several counries, he OECD esimaes show ha he adul populaion had much higher aainmen a upper-secondary and above han our esimaes. For example, he OECD esimae shows ha abou 84% of he adul populaion in Germany had aained educaion a he upper-secondary or eriary level in 1995, while our esimae indicaes ha he corresponding number is only 48%. The discrepancy derives, in many cases, from he differen classificaion beween he OECD and he UNESCO sources. There is disagreemen beween inernaional organizaions abou he allocaion of differen educaion and raining o he caegories of he inernaional sandard classificaion (Seedman, 1999). Alhough he ISCED provides guidance on which qualificaions and sages of educaion should be assigned o he ISCED caegories, he ISCED classificaions do no fully reflec he heerogeneiy of educaional sysems, in paricular of vocaional and non-academic educaional raining, across counries. In addiion, when conflicing inerpreaions are possible, naional governmens can divide or reallocae qualificaion caegories based on poliical consideraions. Seedman (2000) poins ou ha he UK chooses a definiion of he ISCED levels ha gives he impression ha he educaion sysem produces a large supply of skills. Anoher source of he discrepancy is ha our daa refer o he populaion aged 25 and above, whareeas he OECD daa are for he populaion aged beween 25 and 64. Since he populaion aged 65 and above have in general lower aainmen han he younger populaion group, his difference would end o make he OECD figures on upper secondary or eriary aainmen higher han our corresponding numbers. Furher invesigaions on he exac 19

22 relaionship beween our esimaes and he OECD daa on educaional aainmen are sill in progress Conclusion This paper presens a daa se on educaional aainmen for he populaion over age 15 and over age 25 a five-year inervals beween 1960 and 2000 for a broad number of counries. The daa provide he disribuion of educaional aainmen of he populaion a seven levels of schooling and show a breakdown by sex. This improved daa se on educaional aainmen will coninue o conribue o empirical research in many areas. Our daa have been widely used o invesigae he effecs on economic growh of educaional aainmen, measured by overall years of schooling as well as by he composiion of aainmen a various levels of educaion. The daa se has also been useful for sudying he linkages across counries beween educaion and oher imporan economic and social variables, such as expor compeiiveness, feriliy, income inequaliy, and poliical freedom. There has been considerably more work in improving inernaional measures of human capial. Inernaional es scores of sudens a he primary and secondary levels provide useful informaion on he qualiy of educaion. The Inernaional Adul Lieracy Survey is a very promising aemp o measure direcly he skills of he workforce for inernaional comparison. However, hese measures are a presen resriced by he limied sample, which consiss mosly of OECD counries. In his conex, he daa on 17 Since he OECD and he UNESCO sources use differen populaion groups and differen school classificaions, we have no uilized he OECD daabase for he esimaion of our measure of educaional aainmen. The OECD sources are available only for he 1990s. Hence, if we mix 20

23 educaional aainmen sill provide he bes available informaion abou he amoun of human capial sock for a broad number of counries. Furher improvemens of he exising human capial esimaes including he exension of he coverage of he TIMMS and IALS daa o more developing counries mus be imporan. This exension would make i possible o assess he ineracions beween he quaniy and qualiy of educaion for a broad panel of counries. I would also be possible o sudy furher he effecs of human capial on economic performance and o see how hese effecs inerac wih he level of economic developmen. hem wih he earlier UNESCO census daa, i can cause inconsisency over ime in an individual counry's daa. 21

24 References Barro R.J. (1999).'Deerminans of Democracy,' Journal of Poliical Economy 107, Supplemen, S158-S183. Barro, R.J. (2000). 'Educaion and Economic Growh,' Working Paper, Harvard Universiy. Barro, R.J. and Lee. J.W. (1993). 'Inernaional Comparisons of Educaional Aainmen,' Journal of Moneary Economics, 32, Barro, R.J. and Lee, J.W. (1996). 'Inernaional Measures of Schooling Years and Schooling Qualiy,' American Economic Review, 86, Blum, A., Goldsein, H., and Gu rin-pace, F. (2000), 'An Analysis of Inernaional Comparisons of Adul Lieracy,' Working Paper, Insiue of Educaion. Collins, S. and Bosworh, B. (1996). 'Economic Growh in Eas Asia: Accumulaion versus Assimilaion,' Brookings Papers on Economic Aciviy 2, Easerly, W, and Levine, R. (1998). 'Africa's Growh Tragedy: Policies and Ehnic Divisions,' Quarerly Journal of Economics 112, Gajraj, S. (1992). 'Primary Educaion- Age in Grade 1,' Working Paper, Division of Saisics, UNESCO. Gollop, F.M. and Jorgenson, D.W. (1980). 'U.S. Produciviy Growh by Indusry, , in New Developmens in Produciviy Measuremen,' in J. Kendrick and B. Vaccara (eds), The Universiy of Chicago Press. Jorgenson, D.W., Gollop F.M.and Fraumeni, B.M. (1987). Produciviy and U.S. Economic Growh, Harvard Universiy Press. Hall, R.E, and Jones, C.I. (1999). 'Why Do Some Counries Produce So Much More Oupu Per Worker Than Ohers?', Quarerly Journal of Economics, 114, Hanushek, E. and Kimko, D. (2000). 'Schooling, Labor Force Qualiy, and he Growh of Naions,' Working Paper,, Rocheser Cener for Economic Research. Lau, L., Jamison, D. and Loua, F. (1991). 'Educaion and Produciviy in Developing Counries: an Aggregae Producion Funcion Approach,' Repor No. WPS 612, The World Bank. Lee, J.W. and Barro, R.J. (1997). 'Schooling Qualiy in a Cross-secion of Counries,' NBER Working Paper No

25 Mulligan, C. B. and Sala-I-Marin, X. (1995). 'A Labor-Income Based Measure of he Value of Human Capial: An Applicaion o he Saes of he Unied Saes,' NBER Working Paper No Nehru, V., Swanson, E. and Dubey, A. (1995). 'A New Daa Base on Human Capial Sock: Sources, Mehodology, and Resuls,' Journal of Developmen Economics, 46, OECD (1998a). Human Capial Invesmen: An Inernaional Comparison, Cenre for Educaional Research and Innovaion, OECD (1998b). Educaion A A Glance: OECD Indicaors, Cenre for Educaional Research and Innovaion, OECD and Saisics Canada (1995). Lieracy, Economy and Sociey- Resuls of he Firs Inernaional Adul Lieracy Survey, Paris and Oawa. OECD and Human Resources Developmen Canada (1998). Lieracy Skills for he Knowledge Sociey: Furher Resuls from he Inernaional Adul Lieracy Survey, Paris and Oawa. Priche, L. (1996). 'Where Has All he Educaion Gone?' Working Paper No. 1581, World Bank. Psacharopoulos, G. and Ariagada, A.M. (1986). 'The Educaional Composiion of he Labor Force: an Inernaional Comparison,' Inernaional Labor Review, 125, Rajan, RG and Zingales, L. (1998). 'Financial Dependence and Growh,' American Economic Review 88, Ramey, G, and Ramey, V.A. (1995).'Cross-counry Evidence on he Link Beween Volailiy and Growh,' American Economic Review 85, Sachs, J.D., and Warner, A. (1995). 'Economic Reform and The Process of Global Inegraion,' Brookings Papers on Economic Aciviy, Seedman, H. (1999). 'Measuring he Qualiy of Educaional Oupus: Some Unresolved Problems,' forhcoming in Alexander R., Broadfoo P, and Phillips D. (eds) Learning from Comparing: New Direcions in Comparaive Educaion Research, Triangle Books. Seedman, H. (2000). 'Looking Ino he Qualificaions 'Black Box': Wha Can Inernaional Surveys Tell Us Abou Basic Compeence?' Working Paper, Cenre For Economic Performance, London School of Economics and Poliical Science. 23

26 UNESCO, Saisical Yearbook, various years, Paris. 24

27 Table 1. Breakdown of number of counries by number of census-survey observaions Number of observaions Number of counries ( ) Populaion Populaion aged 25& over aged 15& over Toal Noe: The daa refer o census-survey observaions for educaional Aainmen for he oal and female populaions in each age caegory. 25

28 Table 2. Breakdown of number of counries by number of census-survey year Census-survey year Number of counries (o he neares 5-year value) Populaion Populaion aged 25& over aged 15& over Toal Toal Noe: The daa refer o census-survey observaions for educaional aainmen for he oal and female populaions in each age caegory. 26

29 Table 3. Trends of educaional aainmen of he oal populaion aged 25 and over by region Highes level aained Average Region Year Pop. No Primary Secondary Higher years (no. of over 25 school oal full oal full oal full of counries) (mill.) (Percenage of he populaion aged 25 and over) school World (107) All Developing (73) Middle Eas/ Norh Africa (11) Sub-Saharan Africa (22) Lain Ameri./ Caribbean (23)

30 Table 3--coninued Highes level aained Average Region Year Pop. No Primary Secondary Higher years (no. of over 25 school oal full oal full oal full of counries) (mill.) (Percenage of he populaion aged 25 and over) school Eas Asia/ Pacific (10) Souh Asia (7) Advanced Counries (23) Transiional economies (11) Noes: Regional averages are weighed by each counry's populaion. See Appendix Tables for he lis of counries included in each region/group. Counries ha do no have complee daa a five-year inervals from 1960 o 2000 are excluded. Former Germany, Eas is classified in he Transiional economies group unil 1990 and since hen o he advanced counries group as he unied Germany. Former Sovie Union counries excep Russian Republic are excluded since Croaia, Czech Republic, Slovakia, and Slovenia are included in he group of ransiional economies since 1995, replacing he former Yugoslavia and Czechoslovakia. "Toal" refers o he percenage of he populaion for whom he indicaed level is he highes aained. "Full" refers o hose who had he compleion of he indicaed level as he highes aained. 28

31 Table 4. Trends of educaional aainmen of he oal populaion aged 15 and over by region Highes level aained Average Region Year Pop. No Primary Secondary Higher years (no. of over 15 school oal full oal full oal full of counries) (mill.) (Percenage of he populaion aged 15 and over) school World (107) All Developing (73) Middle Eas/ Norh Africa (11) Sub-Saharan Africa (22) Lain Ameri./ Caribbean (23)

32 Table 4--coninued Highes level aained Average Region Year Pop. No Primary Secondary Higher years (no. of over 15 school oal full oal full oal full of counries) (mill.) (Percenage of he populaion aged 15 and over) school Eas Asia/ Pacific (10) Souh Asia (7) Advanced Counries (23) Transiional economies (11) Noe: See able 3. 30

33 Table 5. Educaional aainmen by sex Region Populaion age 25 and Over Populaion age 15 and over (no. of Year Average school year Average school year counries) Females (A) Males (B) Gender raio (A/B,%) Females (A) Males (B) Gender raio (A/B,%) World (107) All Developing (73) Middle Eas/ Norh Africa (11) Sub-Saharan Africa (22) Lain Ameri./ Caribbean (23)

34 Table 5--coninued Region Populaion age 25 and Over Populaion age 15 and over (no. of Year Average school year Average school year counries) Females (A) Males (B) Gender raio (A/B,%) Females (A) Males (B) Gender raio (A/B,%) Eas Asia/ Pacific (10) Souh Asia (7) Advanced Counries (23) Transiional economies (11) Noe: The average gender raios for he whole "world" are downward-biased in 1995 and 2000 because since 1995 he "world" does no include former Sovie Union counries, which have relaively high gender raios, excep Russia. See Noes o Table 3. 32