Υπουργείο Εθνικής Παιδείας και Θρησκευμάτων Ειδική Υπηρεσία Διαχείρισης ΕΠΕΑΕΚ Κείμενο Θέσεων Υπ. Αρ. 5 Rates of return to dfferent levels of educaton: Recent evdence from Greece 2003-2005 Επιστημονικός Υπεύθυνος: Π.Ε. Πετράκης, Καθηγητής Τμήματος Οικονομικών Επιστημών Αθήνα, Οκτώβριος 2008 Το παρόν έχει παραχθεί στα πλαίσια της σύμβασης του ΕΚΠΑ (τεχνικός σύμβουλος) με το ΥπΕΠΘ/ΕΥΔ
Περιεχόμενα 1. Introducton...2 2. Lterature Revew...3 3. Data...5 4. Rates of return to educaton, 2003-2005...6 5. Conclusons...7 References:...14 1
1. Introducton The purpose of ths paper s to use short-cut methodologes n order to update the exstng estmatons of the prvate and socal rates of return to nvestment n each level of the Greek educatonal system. To do so we utlze a dataset consstng of average ncome levels for each educatonal level for years 2003 to 2005 and addtonal OECD estmatons for the educatonal expenses of every level of the Greek educatonal system. Petraks and Kotsos (2008) have, also, appled a smlar methodology to estmate the prvate rates of return and expected rates of return for the perod 1974 to 1998. Moreover, Prodromds and Prodromds (2008) explot a dataset consstng of three household surveys that took place n Greece and estmate a Mnceran earnngs functon. Our emprcal fndngs do not contradct these prevous observatons. However, contrary to the studes mentoned above, we -addtonally- attempt to calculate the socal rate of return to each educatonal level. Moreover, we estmate the expected return rates, takng nto consderaton unemployment rates n each level of educaton n order to depct an accurate pcture of the returns to educatonal nvestment. We document hgh and ncreasng prvate and publc rates of return accrung to tertary educaton. Moreover, we test the hypothess that these hgh rates of return, especally to tertary educaton are a result of the Greek unverstes poor fnancng. The Greek educatonal expenses are relatvely low compared to nternatonal standards; ths dfference ncreases wth the educatonal level, wth unversty educaton recevng less than half the fnancng of the OECD members average. One could argue that the hgh and ncreasng (wth the educatonal level) rates of return to educaton observed, result exactly from ths defct n publc educaton fnancng. We reject ths conjecture by estmatng hypothetcal socal rates of return once the overall educatonal expenses are set equal to OECD countres average. Therefore, keepng other parameters constant, we obtan an estmate of the magntude of the expected declne n socal return rates, due to the antcpated convergence of educatonal expenses wth the developed world standards. The hypothetcal (.e. assumng expenses level equal to OECD average) rates of return to the tertary educaton reman hgher than the true (.e. usng the actually observed cost) return rates of the secondary level. 2
2. Lterature Revew Rates of return to educaton are usually estmated as the dfference between the expected annual ncomes enjoyed by those that have fulflled dfferent educatonal levels; they can be estmated usng varous methodologes. The approach to be utlzed s usually drven by the avalable dataset. When age/earnngs profles can be constructed (usually from household survey data on ndvdual earnngs) the nternal rates of return to educaton can be estmated by equatng the present value of educatonal costs to the present value of expected earnngs by: N w t t = t t= 1 ( 1+ r) t= 1 ( 1+ r) where S and N, respectvely, correspond to the years of schoolng and the work years for each ndvdual, whle w t and c t stand for the average ncome and the cost to educaton for year t, respectvely. The purpose of ths method s to properly estmate the rate of return to the educatonal process, r. Moreover, when ndvdual data are avalable, one could estmate the basc Mnceran earnngs functon: ln 2 ( w) = b + bs+ b e+ b e S 0 1 2 Addtonally to ndvdual ncome, w, the number of schoolng years, s, and work experence, e, the related lterature has added other ndependent varables to the relaton above, n order to control for socal and economc background, ndvdual sklls, etc. Furthermore, the contnuous varable s, n the Mnceran equaton can be replaced by dummy varables representng the dfferent educatonal levels. In that case, the dfferences between the rates of return to dfferent levels of educaton become more straghtforward; they are calculated as the dfference between the coeffcents of the correspondng bnary varables. On the other hand, when the only avalable data concern average ncome levels per educatonal level, we can only estmate the returns to educaton usng short-cut methods to measure the addtonal average ncome attrbuted to educatonal level, compared to the average ncome n level -1, as: RORE 1 c W W = W S where RORE and S stand for the rate of return and the average years of schoolng for each educatonal level,. Moreover, we use the notaton W and W -1 for the average ncome for those who have graduated from educatonal levels and -1, 3 1 t e
respectvely. Usng the last two equatons we do not take nto account the educaton cost at each level; we only calculate a crude estmaton for the demand for graduates of each educatonal level,. We can measure net returns to nvestment n educaton by ncludng the correspondng cost, C, as: RORE = W W W C S ( ) 1 + Addtonally, Groot and Oosterbeek (1992) propose a smlar measure to proxy for the average dfference n ncome due to an addtonal educatonal level, as: RORE ln = 1 ( W ) ln( W ) However, the rates of return, calculated from the equatons above, do not depct the actual returns to nvestment n educaton, unless the employment rates per educatonal level are taken nto consderaton. We are able to estmate the expected returns to educatonal level I, as: S ( 1 ) 1 ERORE p RORE = where ERORE stands for the expected rate of return to educaton and RORE s the rate of return calculated by the methodologes descrbed above. The employment rate, 1-p, s measured usng the unemployment rate for each educatonal level, p. Dependng on whch correspondng rates of return we wsh to calculate, the cost varables represent the prvate or the socal cost to educaton. In ths paper we wll utlze short-cut methods to estmate both. We cannot, at ths pont, utlze econometrc technques to estmate a Mnceran earnngs equaton, snce the avalable dataset conssts of, solely, average ncomes for each educatonal level. A methodologcal approach smlar to the one utlzed below has been mplemented by Petraks and Kotsos (2008). The authors analyze the rates of educatonal returns n Greece for 1974 to 1998 perod; they have, addtonally, calculated rsk adjusted returns, usng dsperson metrcs per educatonal level. Cholezas and Tsakloglou (2001) explot a dataset of ndvdual surveys (1974, 1988 and 1994) and present statstcally sgnfcant evdence on return rates that ncrease wth the educatonal level. Addtonally, they document a declne n the rate of return to educaton between 1974 and 1988 and an ncrease n 1994. In the most recent emprcal work on Greek data, Prodromds and Prodromds (2008) nvestgate three Greek household surveys (1988, 1994, and 1999) and examne the trend of educatonal returns over tme. They fnd substantal return rates to hgher educaton, whch, moreover, grow over tme. Ths s consstent 4
wth our evdence of ncreasng rates of return for the hgh levels of secondary and tertary educaton. However, our fndngs contradct the emprcal results presented n Psacharopoulos (1993) and Psacharopoulos and Patrnos (2002). Addtonally, both studes of Cholezas and Tsakloglou (2001) and Prodromds and Prodromds (2008) evdence ncreased rates of return to educaton for females, as opposed to males whch comply wth nternatonal emprcal fndngs. Our work cannot offer such predctons, snce we have no data on male and female average ncomes separately. For a complete revew of the emprcal lterature on rates of return to educaton n Greece, see Cholezas and Tsakloglou (1999). 3. Data The data set for ths study conssts of the average ncome for the dfferent educatonal levels provded by the General Secretarat of Natonal Statstcal Servces (GNSS) for the years 2003 to 2005. The avalable levels of educaton nclude the prmary, secondary (lower and upper secondary educaton), tertary educaton and PhD holders. Addtonally, the average ncome levels for uneducated people for the three-year perod are, also, avalable. However, there s no dstncton between general and vocatonal upper secondary educaton or post-graduate (.e. master s degrees) and PhD levels. Moreover, ncome dsperson metrcs wthn each educatonal segment are not currently avalable and, hence, a rsk/return analyss cannot be mplemented. Furthermore, we wll utlze the unemployment rates (also provded by the GSNSS) for each educatonal level n order to calculate the expected rates of return, takng nto account the probablty of employment. However, as already mentoned above, the orgnal survey data are not currently avalable. Addtonally, we have used OECD data concernng the annual cost per student for every educatonal level (OECD, Educaton at a glance, 2007, Indcator B1.1). Unfortunately, there are no avalable data for the lower and upper secondary educaton, separately; we have used for both these educatonal levels the annual expenses per student at the secondary level, as a whole. Moreover, there s no dstncton between the cost of graduate and post-graduate tertary educaton. In order to resolve that, we set the cost of graduate educaton equal to the estmated cost of tertary educaton per student excludng R&D expenses, whle we proxy the cost per PhD canddate as the total (.e. ncludng R&D expendtures) annual expenses for tertary educaton per student. That the duraton of studes vares sgnfcantly wth the educatonal level, but also depends sgnfcantly on the tranng feld, s more than obvous. Therefore, what 5
s of nterest s the relatve addtonal ncome enjoyed per year of studes. Snce the years of schoolng for tertary educaton vary wth the educatonal feld, we wll assume an average of 5 years for undergraduate tertary educaton and an addtonal average of 4 years for the completon of a doctoral dssertaton. Settng the number of years, whch are necessary to fulfll graduate and post-graduate studes at lower levels, would only result n -even- hgher returns to tertary educaton. 4. Rates of return to educaton, 2003 2005 Even though the numercal results of the two utlzed methods ( short-cut method and the methodology of Groot & Oosterbeek) do not concde, the qualtatve results do not dffer. The most obvous pattern recognzed n the results obtaned, whch are presented n Table 1, concern the ncreasng returns to educaton for the secondary and tertary levels of the educatonal system. In other words the average ncome ncreases, wth hgher rates, as we move to the hgher levels of educaton. Addtonally, we confrm the hgh returns to prmary educaton, whch s evdent n the related lterature [Psacharopoulos (1993) and Psacharopoulos and Patrnos (2002)]; n fact, movng from prmary to lower secondary educaton s the only sgn of dmnshng returns to educaton! The hgh returns to prmary educaton can be attrbuted to the margnal stuaton t represents: the average ncome for uneducated people n Greece s so low that an nfntesmal ncrease to educaton (.e. graduatng from prmary school) results to a substantal percentage ncrease n average ncome. Ths ncrease s hgher than the one attrbuted to the two followng levels of secondary educaton. However, t s extremely nterestng to measure these rates of return to prmary educaton wth tme, snce the number of people that have fulflled prmary educaton only s expected to decrease wth tme. At ths pont, t should be mentoned that a fully segregated sample consstng of detaled data for the dstnct categores of each level of the educatonal system would result to a much more nterestng emprcal nvestgaton of the educatonal system and a better comprehenson of the demand for labor force n Greece. More specfcally, t s of extreme mportance to dentfy potental dfferences between the rates of return for the two categores of hgher secondary (general and vocatonal educaton) and tertary (techncal and unversty educaton) educaton that consttute avalable optons for Greek students. Addtonally, t would be very nterestng to examne the expected rates of return to post-secondary, non-tertary educaton. Unfortunately, the relevant data are not currently avalable and, hence, our paper cannot comment on ths ssue. Internatonal lterature documents hgher rates of return to the general rather than to techncal educaton. 6
Next, n order to take nto account the statstcally observed employment rates and calculate the expected rates of return to each level of educaton, we document the well known fact that unemployment n the Greek labor market does not decrease monotoncally wth the educatonal levels (columns (3), (6) and (10) n tables 3 and 4). Even when we take unemployment rates nto consderaton, the pcture does not change; once more, we confrm the emprcal fndngs of the nternatonal lterature [Psacharopoulos (1994), Psacharopoulos and Patrnos (2002)] of hgh returns to prmary educaton. However, we also document returns to secondary and tertary educaton, whch ncrease wth the educatonal level. Next, we present our fndngs on the correspondng socal rates of return to dfferent levels of educaton, estmatng the annual socal cost per student from OECD data for year 2004. The socal rates of return s lower than our bref estmatons of the prvate returns (not ncluded n the paper), whch s attrbuted to the state subsdzaton of hgher educaton. However, the pcture depcted prevously usng gross prvate rates of return, remans. One could argue that the hgh socal rates of return to hgher educaton can be attrbuted to the relatvely poor fnancng of the tertary level (column (5) of Table 4 ndcate the fundng shortage, relatvely to the average fnancng of OECD countres) of the Greek educatonal system; the percentage rates of return would be lower f the socal cost of graduate and post-graduate educaton were to be ncreased to OECD standards. In order to check the valdty of ths conjecture, we calculate hypothetcal socal rates of return to each educatonal level, settng the correspondng annual cost per student equal to the average OECD members annual expenses per student. In that case, the rates of return to each educatonal level decrease (compared to those estmated usng the actual cost per student documented by the OECD), but postgraduate and graduate studes stll offer hgher rates of return relatvely to secondary educaton. Therefore we conclude that the substantal returns to hgher educaton cannot be attrbuted to the poor fnancng of Greek unverstes. Even an ncrease n the per student expendtures to the correspondng average of the OECD countres, holdng everythng else constant, would not reverse the documented results, wth tertary educaton resultng to hgh rates of return. Both the actual and the hypothetcal socal rates of return to educaton are presented n Table 4. 5. Conclusons Ths work documents a number of nterestng observatons concernng the Greek educatonal system. Based on the above fndng, we are able to propose some 7
1. The rates of return to the dfferent levels of educaton follow a U-shaped curve wth the educatonal level, as already documented above. Prmary and tertary educaton yeld the most substantal returns, whle secondary level graduates (and especally those holdng only a secondary educaton degree) enjoy lower return rates. The dfference between the numercal results (Table 1) of the two methodologcal approaches (the usual short-cut method and the Groot and Oosterbeek methodology) utlzed s not sgnfcant, although t ncreases wth the educatonal level. More mportantly, the rankng of educatonal levels accordng to the correspondng return rates, reman unchanged ndependently of the methodology used. Our results are n accordance wth the emprcal nvestgaton of Greek household surveys n Cholezas and Tsakloglou (2001) and Prodromds and Prodromds (2008), where substantally hgh returns to tertary educaton (hgher than those of the secondary level) are also documented. Addtonally, our fndngs of hgh returns to hgher educaton comply wth Psacharopoulos (2008), ndcatng a sgnfcant demand for educated people n modern Greek economy. 2. The most well documented fndng of ths paper s the hgher rates of return to the tertary educaton. Moreover, we can compare these return rates wth the correspondng rsk-free nterest rate. We conclude that the rates of return to hgher educaton (both at graduate and post-graduate levels) are hgher than the rskless nterest rate, whle the return to secondary educaton lags the yelds of rsk-free nvestment. Hence, we expect the dropout rates to reman relatvely hgh at the secondary level of the educatonal system, whle the demand for graduate and post-graduate studes to grow further. 3. When adjustng for unemployment rsk, the calculated return rates are an average 0.5% lower than the ones estmated wthout takng nto account the probablty of employment. The greatest decrease n rates of return takes place n the secondary level, where the unemployment rates are hgher and the rates of return lower. However, unemployment remans at sgnfcant levels, even for hgher educaton graduates and, hence, decreases the rate of return to ths educatonal level, as well. Based on these fndngs, the quest for safer study felds (n the sense of reduced unemployment rsk) s well justfed, snce the contrbuton of unemployment levels to expected rates of return s sgnfcant, as descrbed. However, the value of the expected return rates s mostly nfluenced by the 8
relatve ncome levels. Therefore, what should be encouraged s seekng for studes wth hgh returns resultng from hgher expected ncome. 4. Moreover, the analyss predcts that a potental ncrease n unemployment levels or nterest rates would result n a totally dfferent stuaton for the decson to nvest n educaton, snce t would reduce the (adjusted for unemployment) expected rate of return and would make other nvestments, alternatve to nvestment n educaton (.e. relatvely rsk-less securtes wth yelds ted to the EURIBOR) seem more attractve. Currently the stuaton seems at a margnal pont; n the very lkely case of a further deteroraton of global economc crcumstances nvestment n educaton s expected to fade and certanly not to grow any further. 5. The calculated socal rates of return hghlghts two major polcy mplcatons: Frst, the utmost mportance of nvestment n prmary educaton, whch bears a return rate of about 8% and, second, the fact that socal rates of return to educaton are well below the correspondng prvates return rates. Both these observatons are well documented n the related lterature [see for example Psacharopoulos (1984), (1992)]. The fact that socal rates of return reman well below the prvate ones, ndcates that the potental nvestor (ether the prvate or the publc sector) that partcpates n the fnancng of nvestment n educaton, should expect relatvely low returns. These returns, ceters parbus, would become even lower n the event of an ncrease of educatonal fundng to OECD countres average. In other words, an ncrease n the level of nvestment would decrease the expected socal rates of return, unless the correspondng ncome levels ncrease to the same extent (as a result of ths hgher nvestment n educaton). Thus, an ncrease n educatonal nvestment, and the antcpated productvty gans resultng from t, should necessarly lead to rses n ncome levels; otherwse educaton as an nvestment opportunty wll seem less attractve than t s today. Addtonal nvestment should be headed for the felds (and, respectvely, professons), whch yeld the hghest returns. It s necessary to adopt a dfferent nvestment structure, and plan further nvestment n educaton accordng to the return of each tranng feld and each sector of the Greek economy related to t. 6. Can the above calculated rates of return cast any nformaton about the level of fundng of the dfferent levels of the Greek educatonal system? The only pece 9
of nformaton we can accurately export concerns the hghlghted dfference between the prvate and socal rates of return to educaton. 7. Last, we could underlne the relaton between the actual rates of return to educaton and the proper measurement of ncomes, due to the large scale of underground economy and tax avodance, for whch Greece s notorous about. If we could measure the actual ncome levels generated by the grey economy t s possble that the above calculated rates of return would dffer substantally. One could argue that the atttude towards taxes dffers wth the educatonal level. Addtonally, t s hghly possble that the professons of secondary educaton graduates facltate tax avodance and the grey economy. In other words, t s not unlkely that the true rates of return to lower levels of the educatonal system are not as low as documented n our study. Unfortunately, however, collectng the necessary dataset and measurng the true (.e. after adjustng for tax avodance or ncomes from underground economc actvty) average ncomes for each educatonal level, n order to properly adjust the above rates of return s not possble. Of course, ths measurement bas arses for every emprcal work on rates of return to educaton, whch takes place n countres wth hgh levels of grey economy. 10
Table 1: Gross rates of average return to each educatonal level per year of schoolng, for years 2003-2005 calculated usng the average yearly ncomes and utlzng the narrow method and the methodology of Groot and Oosterbeek (1992), descrbed above. (The averages ncomes n columns (3), (6) and (9) are provded by statstcal data of the GSNSS). 2003 2004 2005 Educatonal Level () Average years of RORE RORE RORE schoolng 1 (S ) Average Income ( ) Short-cut Method Groot & Oosterbeek Average Income ( ) Short-cut Method Groot & Oosterbeek Average Income ( ) Short-cut Method Groot & Oosterbeek (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) (11) No schoolng (1) 5.664 4.795 7.667 Prmary (2) 6 8.498 8,34% 6,76% 8.852 14,10% 10,22% 9.403 3,77% 3,40% Lower Secondary Upper Secondary (3) 3 9.080 2,28% 2,21% 9.574 2,72% 2,61% 9.720 1,12% 1,11% (4) 3 10.185 4,06% 3,83% 10.992 4,94% 4,60% 11.203 5,09% 4,73% Tertary (5) 5 15.119 9,69% 7,90% 15.271 7,79% 6,58% 16.213 8,94% 7,39% PhD (6) 4 26.301 18,49% 13,84% 22.772 12,28% 9,99% 24.076 12,12% 9,89% 1 The average years of schoolng for tertary educaton are assumed to be 5 years. The average years for fulfllng PhD studes are set to 4 years. 11
Table 2: Expected rates of return to each educatonal level, for years 2003-2005 calculated usng the average yearly ncomes and utlzng the methodology of Groot and Oosterbeek (1992), descrbed above (The ROREs n columns (4), (7) and (10) are calculated n columns (5), (8) and (11) of Table 1). Level of educaton () Average years of schoolng (S ) Probablty of employment 2003 2004 2005 Probablty of employment RORE ERORE RORE ERORE Probablty of employment RORE ERORE (1) (2) (3) (4) (5) = (3) x (4) (6) (7) (8) = (6) x (7) (9) (10) (11) = (9) x (10) No Schoolng (1) 95,70% 86.60% 90.70% Prmary (2) 6 93,60% 6,76% 6,33% 91.30% 10,22% 9,33% 92.70% 3,40% 3,15% Lower Secondary (3) 3 89,30% 2,21% 1,97% 88.80% 2,61% 2,32% 89.60% 1,11% 0,99% Upper Secondary (4) 3 88,20% 3,83% 3,38% 88.40% 4,60% 4,07% 88.90% 4,73% 4,20% Tertary (5) 5 93,30% 7,90% 7,37% 92.50% 6,58% 6,09% 92.90% 7,39% 6,87% PhD (6) 4 92,80% 13,84% 12,84% 94.10% 9,99% 9,40% 91.50% 9,89% 9,05% Table 3: Expected rates of return to each educatonal level, for years 2003-2005 calculated usng the average yearly ncomes and utlzng the short-cut method, descrbed above (The ROREs n columns (4), (7) and (10) are calculated n columns (5), (7) and (11) of Table 1). Level of educaton () Average years of schoolng (S ) Probablty of employment 2003 2004 2005 RORE ERORE Probablty of employment (1) (2) (3) (4) (5) = (3) x (4) (6) (7) (8) = (6) x (7) (9) (10) (11) = (9) x (10) No Schoolng (1) 95,70% 86.60% 90.70% Prmary (2) 6 93,60% 8,34% 7,81% 91.30% 14,10% 12,87% 92.70% 3,77% 3,49% Lower Secondary (3) 3 89,30% 2,28% 2,04% 88.80% 2,72% 2,42% 89.60% 1,12% 1,00% Upper Secondary (4) 3 88,20% 4,06% 3,58% 88.40% 4,94% 4,37% 88.90% 5,09% 4,53% Tertary (5) 5 93,30% 9,69% 9,04% 92.50% 7,79% 7,21% 92.90% 8,94% 8,31% RORE ERORE Probablty of employment RORE ERORE PhD (6) 4 92,80% 18,49% 17,16% 94.10% 12,28% 11,56% 91.50% 12,12% 11,09% 12
Table 4: Socal rates of return for each educatonal level, calculated usng the short-cut method. Columns (2) and (3) are converted from USD to Euros usng the OECD estmaton of PPP rate for year 2004. Column (6) s calculated usng the actual expenses per student, (2), whle column (7) uses the average annual cost per student of OECD members, column (3). (Source: OECD Educaton at a glance, 2007). Annual expendture per student SRORE SRORE Level of Educaton () Greece OECD AVG Dfference Dfference (%) (1) (2) (3) (4) =(3) - (2) (5) = (4) / (2) (6) (7) Prmary (2) 3,226.58 4,095.53 868.95 26.93% 8.43% 7.61% Lower Secondary (3) 3,660.83 5,108.98 1,448.15 39.56% 1.92% 1.72% Upper Secondary (4) 3,660.83 5,108.98 1,448.15 39.56% 3.57% 3.22% Tertary (5) 3,174.45 5,583.14 2,408.69 75.88% 6.04% 5.16% PhD (6) 3,927.18 7,794.43 3,867.25 98.47% 9.77% 8.13% Table 5: Socal rates of return for each educatonal level, calculated usng the short-cut method and adjustng for unemployment rsk by multplyng the socal rates of return calculated n Table 4 (columns (6) and (7) respectvely) by the employment probablty (Column (4) of the table). Level of Educaton Socal RORE Probablty of Socal ERORE () employment Greece OECD Greece OECD (1) (2) (3) (4) (5) = (2) x (4) (6) = (3) x (4) Prmary Lower Secondary Upper Secondary Tertary PhD 8.43% 7.61% 93.60% 7.89% 7.12% 1.92% 1.72% 89.30% 1.72% 1.54% 3.57% 3.22% 88.20% 3.15% 2.84% 6.04% 5.16% 93.30% 5.64% 4.82% 9.77% 8.13% 92.80% 9.06% 7.54% Table 6: One month and one year EURIBOR for years 1998 to 2007 (Source: ECB). 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 Monthly 3.97% 2.86% 4.24% 4.33% 3.30% 2.35% 2.08% 2.14% 2.94% 4.08% Yearly 3.91% 3.19% 4.78% 4.09% 3.49% 2.34% 2.27% 2.33% 3.44% 4.45% 13
References: Cholezas, I. and Tsakloglou, P. (1999), Prvate Returns to Educaton n Greece: A Revew of the Emprcal Lterature n Asplund, R. and Perera, T.P. (eds), Returns to Human Captal n Europe: A Lterature Revew, ETLA, The Research Insttute of the Fnnsh Economy, Helsnk. Cholezas, I. and Tsakloglou, P. (2001), Prvate Returns to Educaton n Greece, Workng Paper, Athens Unversty of Economcs and Busness, Department of Internatonal and European Economc Studes Groot, W. and Oosterbeek, H. (1992), Optmal nvestment n human captal under uncertanty, Economcs of Educaton Revew, Vol. 11, pp. 41-49 Mncer, J. (1974), Schoolng experence and earnngs, Columba Unversty Press Petraks, P and S. Kotsos, (2008), The Dynamc Evoluton of Educaton Investment, Rsk Return Relatonshp and Educaton Structural System Change Evaluaton, accepted for presentaton n 2 nd Internatonal Conference on Educatonal Economcs Prodromds, K. and P. Prodromds (2008), Returns to educatons: the Greek experence 1989-1999, Appled Economcs, Vol 4, pp. 1023-1030 Psacharopoulos, G. (1994), "Returns to nvestment n educaton: a global update,", Polcy Research Workng Paper Seres 1067, The World Bank Psacharopoulos, G. (2008), The returns to nvestment n hgher educaton: Methods, data and polcy mplcatons In Rates of return and fundng models n Europe, Fnal Report to the Drectorate-General for Educaton and Culture of the European Commsson Psacharopoulos, G. & Patrnos, H. (2002), "Returns to nvestment n educaton : a further update,", Polcy Research Workng Paper Seres 2881, The World Bank 14