Does Higher Educated People Earn More Money in the Labor Market in China?

Transcription

1 Does Hgher Educated People Earn More Money n the Labor Market n Chna? Abstract Ths study uses the most recent Chnese Urban Household Survey data whch cover all provnces except Tbet n 1991 and 2000 to estmate returns to educaton n urban Chna. Most exstng studes n the lterature are based on data from the 1980s and coverng only part provnces of Chna and fnd that the rates of return to educaton n Chna are remarkably low. Ths study nvestgates econometrcally whether the returns to educaton has a sgnfcant rsng trend n the 1990s based on Mncer-type equaton, and estmate statstcally whether the expected net revenues of educatons at varous levels are postve sgnfcantly when educaton costs are consdered. The regresson results and statstc estmates show a few fndngs: Frst, the Mnceran rate of return to educaton s postve sgnfcantly and ndeed has a remarkable rsng trend n the 1990s. Second, the expected net revenues at all educaton levels are postve, but the dfference among varous educaton levels s very bg. Fnally, the expected net revenues for hgher educaton s bggest so that we can explan the reason why the hgher educaton expanson s so rapd n recent years. JEL classfcaton: 120; 121 Keywords: Return to educaton; Cost recovery; Chna Changjun YUE Assocate Professor of Economcs The Insttute of Economcs of Educaton Pekng Unversty Bejng , P. R. Chna Tel: (0086) (H); (O) Fax: (0086) E-mal: cjyue@gse.pku.edu.cn 1

2 Does Hgher Educated People Earn More Money n the Labor Market n Chna? 1. Introducton Snce the begnnng of 1990s, Chna has made remarkable growth n educaton at varous levels. In partcular, the scale of Chnese hgher educaton has undergone a very rapd expanson. In 1991, there were only 0.62 mllon new students admtted to regular nsttutons of hgher educaton, whle n the year of 2003, the new ntakes of students ncreased to 3.4 mllons. The average rate of growth s as hgh as 15.2%. Along wth the hgher educaton expanson, Chnese unverstes started to charge tuton and fees as a cost sharng and cost recovery, and the tuton and fees have been rsng much quckly than the scale expanson. In 2002, the average tuton and fees has reached 5000 Yuan, whch s bgger than 60% of GDP per capta n the same year. However, t seems that the hgh tuton and fees have lttle mpact on reducng the demand for hgher educaton. Why the demand for hgher educaton s so great n 1990s n Chna? Is t worthy to pay so much tuton and fees for hgher educaton? The reasonable explanaton s that the return to educaton must be hgh enough to cover all educatonal costs. As a matter of fact, almost all exstng studes fnd very low returns of ncome to educaton n Chna n 1980s and n the begnnng of 1990s, based on Mncer-type earnngs regresson equatons. Byron and Manaloto (1990) found a very low fgure of only about 1.4% for each addtonal year of schoolng n Chna, usng a survey of 800 state ndustral workers n Nanjng n Maurer-Fazo (1999) used the urban sample of 1988 Chnese Household Income Project (CHIP) data and found the schoolng coeffcents n smple Mncer equaton regresson result are only 2.9%. Many other researches obtaned the smlar results. Nevertheless, the returns to educaton n Chna are found to have a rsng trend n 1990s. Zhang and Zhao (2002) used annual data from Urban Households Surveys whch cover 6 provnces and ctes and found that there had been a dramatc and consstent ncrease n return n urban Chna. Durng the perod of 1988 and 1999, the rate of return to educaton rose from 4.7% to 11.5%. L (2003) used the urban sample of 1995 Chnese Household Income Project data whch cover 11 provnces and ctes and found the overall return for an addtonal year of 2

3 schoolng was 5.4%, whch was much larger than former estmates. 3

4 Unfortunately, most exstng studed n the lterature are based on regonal data, whch cover only part provnces and ctes, and the former studes used the same method of Mnceran rate of return to educaton and used the data before the mddle of 1990s durng whch tme college students were free to receve hgher educaton. However, snce 1997, each college students n Chna must pay tuton and fees for hgher educaton. Therefore, the objectves of ths artcle are to estmate econometrcally whether there s a rsng trend n the change of returns to educaton n urban Chna usng most recent Chnese urban households survey data of 1991 and 2000 whch cover all provnces and ctes except Tbet, and to estmate statstcally whether the expected net revenues of educatons at varous levels are ndeed postve sgnfcantly when educaton costs are consdered, whch ncludng both drect educatonal cost and opportunty cost. Ths artcle s dvded nto four sectons. In secton 2, we test econometrcally the rsng trend of Mnceran rate of return to educaton between 1991 and 2000, based on Mncer-type equaton. In secton 3, we estmate statstcally the expected net revenues of educatons at varous levels based on an ndex of prvate rate of return to educaton. Secton 4 concludes. 2. Mnceran rate of return to educaton 2.1. Data The data used n ths paper come from annual data from Chnese Urban Households Surveys n 1991 and 2000, conducted by the Urban Socoeconomc Survey Teams of the State Statstcs Bureau. Defntons of varables and descrptve statstcs are lsted n Table 1. After droppng the observatons wth mssng values on educaton, experence, etc., the numbers of vald observatons are and for 1991 and 2000, respectvely. The educaton measure n the surveys ncludes seven categores based on degrees for the year 2000, and sx categores for To estmate overall return to schoolng, a partcular number of years should be estmated for each educaton category. Followng other studes, the number of years s assgned as follows: Unversty (UNIV, Daxue, 16 years), Non-unversty Tertary (NUT, Dazhuan, three-year college, 15 years), Specalzed Secondary School (SSS, 4

5 Zhongzhuan, 12 years), Senor Hgh School (SHS, 12 years), Junor Hgh School (JHS, 9 years), Prmary School (PS, 6 years), and below prmary school (0 years). The average number of educaton years has rsen from n 1991 to n

6 The ndustry measure n the surveys ncludes 14 categores n 1991 and 16 categores n To make them unanmous, the 16 ndustres n 2000 are aggregated nto 5 groups: Industry1 comprses 1 ndustry (Farmng, Forestry, Anmal Husbandry and Fshery); Industry2 comprses 2 ndustres (Manufacturng; Constructon); Industry31 comprses 4 ndustres (Health Care, Sports & Socal Welfare; Educaton, Culture and Arts, Rado, Flm and Televson; Scentfc Research and Polytechnc Servces; Government Agences Party Agences and Socal Organzatons); Industry32 comprses 3 ndustres (Fnance and Insurance; Real Estate; Socal Servces); Industry4 comprses 6 ndustres (Mnng and Quarryng; Producton and Supply of Electrcty Gas and Water; Geologcal Prospectng and Water Conservancy; Transport, Storage, Post & Telecommuncaton Servces; Wholesale and Retal Trade & Caterng Servces; Others). In 1991 sample, the two ndustres of Real Estate and Socal Servces are unted as one, and there s no others ndustry. So the 14 ndustres n sample 1991 are aggregated nto the same groups as for sample The samples both n 1991 and n 2000 cover 29 provnces and ctes except Tbet. Accordng to the geographcal and economc characterstcs, the provnces and ctes are aggregated nto three groups: Regon1 comprses three ctes (ncludng Bejng, Tanjn, and Shangha); Regon2 covers seven coastal provnces (ncludng Guangdong, Zhejang, Jangsu, Shandong, Fujan, Laonng, and Hebe); Regon3 s composed of nneteen nland provnces and Hanan (ncludng Shanx, Inner Mongola, Helongjang, Jln, Anhu, Jangx, Henan, Hube, Hunan, Hanan, Schuan, Guzhou, Yunnan, Guangx, Shaanx, Gansu, Qngha, Nngxa, and Xnjang). The ownershp measure n the surveys ncludes 3 groups: Ownershp1 comprses State Owned Untes; Ownershp2 s Collectvely Owned Enterprses; Ownershp3 s composed of all other owned enterprses The regresson functon of Mnceran Rate of Return to Educaton The most wdely used form for statstcal ncome functon s the followng Mncer-type econometrc regresson functon (Mncer, 1974), LnW = β Experence + u (1) 0 + β1schoolng + β2 Experence + β3 2 6

7 Where W s the annual earnngs; Schoolng s a measure of educatonal attanment; Experence s years of labor market experence. The schoolng coeffcents can be nterpreted as the average (prvate) rate of return to an addtonal year (or level) of schoolng. Based on human captal theory, wages are determned by nvestng n human captal. Schoolng and onthe job tranng are major types of nvestment. Generally, experence s used as a proxy for job-tranng nvestment. We wll use the natonal wde data to estmate the return to educaton. Snce Chna s both a bg country and a transton economy, there are gender dscrmnaton, ndustry monopoly, uneven development among provnces, and uneven dstrbuton of labors among ownershps. Hence, n addton to human captal theory, the wages are also affected by other condtons. So some dummy varable are added nto the smple Mncer equaton as the followng form: LnW = β 0 + β + β 1 5 Schoolng Industry + β + β 6 2 Experence Re gon + β 7 + β 3 Experence Ownershp + u 2 + β 4 Gender (2) Where, Gender s a dummy varable capturng the ncome dfferental between genders. Industry dfferences n earnngs are controlled for by ntroducng four dummy varables: Industry1, Industry31, Industry32, and Industry4; the benchmark ndustry s ndustry2. Introducng two regonal dummy varables controls for regonal dfferences n earnngs. Regon1 ncludes 3 ctes; Regon2 covers 7 coastal provnces, whch are relatve rch economes; the benchmark regon s Regon3, and ths regon comprses 19 relatve poor and less developed provnces. Ownershp dfferences n earnngs are controlled for by ntroducng two ownershp dummy varables: Ownershp2 and Ownershp3. The benchmark ownershp s Ownershp1. Based on human captal theory, ncomes are determned by nvestment n human captal. Schoolng and on-the-job tranng are major types of nvestng. Hence, the schoolng coeffcents, whch can be nterpreted as the average rate of return to an addtonal year of schoolng, and the experence coeffcent, whch can be nterpreted as the average rate of return to an addtonal year of experence, are expected to be postve. Snce the gender dscrmnaton or dfference s a unversal phenomenon n many countres n the world, so the coeffcent of gender s expected to be postve. Although ndustry1 s composed of Farmng, Forestry, Anmal Husbandry and Fshery, the ndvduals 7

8 n ths ndustry n the sample are workers n urban area, not nclude rural farmers. In addton, the percentage of ndustry n sample s only 1.1%. So the sgn of coeffcent of ndustry1 dummy varable s hard to predct. Snce Industry31 comprse four monopolstc ndustres, ts coeffcent must be sgnfcantly postve. The benchmark ndustry s ndustry2 and t covers manufacturng and constructon, whch s an almost complete compettve ndustry and labor s easy to move to ths ndustry. Whle ndustry32 and ndustry4 are less compettve, compared to ndustry2. Therefore, the coeffcents of ndustry31 and ndustry4 are expected to be postve. Snce the regon1 s the rchest area n Chna and regon 2 s also rcher than regon3, so the coeffcents of regon1 and regon2 should be postve. The ownershp1 ncludes many monopolstc sectors and ownershp2 s relatve compettve sector so the coeffcent of ownershp2 should be negatve. Ownershp3 comprses both foregn funded enterprses and prvate enterprses. The wages n foregn funded enterprses could be very hgh for staffs and could be low for general workers. So the coeffcent of ownershp3 s also hard to predct Econometrc results Table 2 reports estmaton results. Regresson 1 and 3 are run usng the smple Mncer equaton for the years of 1991 and 2000, respectvely. In regresson 2 and 4, a bunch of control varables are added n order to test for the robustness of the changng trend of Mnceran rate of return to educaton. These control varables nclude gender, ndustry, regon, and ownershp. All coeffcents of schoolng are postve and statstcally sgnfcant at the 1% level of sgnfcance. In partcular, the Mnceran rate of return to educaton rose from 3.03% n 1991 to 8.47% n 2000 based on smple Mncer equaton. When dummy varables are added, the returns to educaton n regresson 2 and 4 are lower than the specfcaton n a smpler model n regresson 1 and 3, respectvely. But the rsng trends are consstent. The Mnceran rate of return to educaton rose from 2.07% n 1991 to 6.22% n All coeffcents of experence are postve and statstcally sgnfcant at the 1% level of sgnfcance. However, the rate of returns to experence decreased from 5.78% n 1991 to 8

9 3.37% n 2000, or from 5.90% n 1991 to 4.45% n 2000 when control varables are added. It s nterestng to compare the rate of return to educaton wth the rate of return to experence. In 1991, the rate of return to experence was bgger than the rate of return to educaton, whch means that the people had more earnngs wth an addtonal experence than wth an addtonal educaton. In 2000, on the contrary, the rate of return to educaton was bgger than the rate of return to experence. So the educaton has become a key varable to explan the ncome dfference. The coeffcents of gender dummy varable are postve and statstcally sgnfcant at the 1% level of sgnfcance. The coeffcent rose from 8.42% n 1991 to 15.8% n The value of gender dummy varable equals to 1 for the male, and equals 0 for the female. Obvously, the gender gap of ncome exsts and has been enlargng n recent years. The coeffcents of ndustry except ndustry1 are postve and statstcally sgnfcant at the 1% level of sgnfcance, the coeffcent of ndustry1 for 1991 s negatve sgnfcantly. The coeffcents for 1991 are very low, whch mples the ncome gap across ndustres s small. However, all coeffcents for 2000 are pretty bg, and ths shows the gap across ndustres has become large. The benchmark ndustres are manufacturng and constructon, and these two ndustres have the lowest wages n that all the coeffcents of ndustry dummy varables are postve. In fact, based on the data from Chna Statstcal Yearbook (2002), the rato of the hghest wage to the lowest wage among 16 ndustres rose from 1.8 n 1991 to 2.9 n Industry2 s a most compettve ndustry and s a labor-ntensve ndustry so no wonder ts wages on average s lowest. The coeffcents of ndusty31 and ndustry32 dummy varables are larger than 20% n 2000, whch mples the ncome gap among ndustres s remarkable. Table 3 reveals that ndustry compostons of urban employed persons by educatonal attanment are qute dfferent, and they are closely related to the ndustry patterns of average wage of staff and workers. The average wage for all ndustres s 10,870 Yuan. There are 5 ndustres n whch the wages are lower than the average wage (lower-wage ndustres). The percentages of staff and workers who work n the lower wage ndustres are 76.1%, 64.4%, 45.8%, 23.1%, 19.7% for prmary school level, junor hgh school level, senor hgh school level, college level, unversty level, respectvely. It s easy to see that the hgher educated 9

10 people have more opportunty to leave lower wage ndustres and move to hgher wage ndustres. Accordng to the human captal theory, the hgher educated people have hgher productvtes so that they deserve hgher wages. Hence, the uneven dstrbutons of labors at varous educaton levels further enhanced the ncome gap among ndustres. The coeffcents of regon are postve and statstcally sgnfcant at the 1% level of sgnfcance. The benchmark regon comprses nland provnces; Regon1 ncludes three huge ctes: Bejng, Shangha, and Tanjn (BST); Regon2 s composed of seven rcher coastal provnces. The regresson results show that the ncome gap among three regons s bg and sgnfcant. Moreover, the gap has become bgger from 1991 to The ncome gap between BTS and nland provnce rose from 17.6% n 1991 to 40.6% n 2000, whle the gap between coastal provnces and nland provnces rose from 18.2% n 1991 to 29.3% n Calculated from the data of Chna Statstcal Yearbook (2002), the average wages n 1991 for BST, coastal provnces, and nland provnces are 2708 Yuan, 2261 Yuan, and 1873 Yuan, respectvely. The wage of BST s 45% bgger than the wage of nland provnces. In 2000, the average wages for BST, coastal provnces, and nland provnces are Yuan, Yuan, and 7372 Yuan, respectvely. The wage of BST s 121% bgger than the wage of nland provnces. Snce the uneven development across provnces and ctes, the regon varable s a necessary explanng varable n Mncer equaton when used for Chna. The coeffcents of ownershp2 are negatve and statstcally sgnfcant at the 1% level of sgnfcance. Compared to the benchmark ownershp of state-owned enterprses, the collectve owned enterprses are unsklled labor-ntensve so ts average wage should be lower. Calculated from the 1991 sample data, 66% workers schoolng attanments n collectve owned enterprses are junor hgh school and below, whle only 38% workers schoolng attanments n state owned unts are junor hgh school and below. In 2000, 49.7% workers schoolng attanments n collectve owned enterprses are junor hgh school and below, whle only 24.5% workers schoolng attanments n state owned unts are junor hgh school and below. The ncome gap between ownershp2 and ownershp1 rose from 17.0% n 1991 to 31.6% n The coeffcents of ownershp3 are postve and statstcally sgnfcant at the 1% level of sgnfcance. Compared to the benchmark ownershp of state-owned enterprses, ownershp3 ncludes foregn funded enterprses and prvate enterprses, whch are most 10

11 vgorous n Chnese market economy. The wages of these enterprses mght be hgher. So ts coeffcents are postve and rose from 7.1% n 1991 to 11.25% n In the above regressons, the educatonal attanment s measured by years of schoolng, whch means the schoolng coeffcent s derved from a contnuous schoolng varable, so that the coeffcent captures the average effect of schoolng assumng constant margnal effects. Table 4 shows the margnal return from each addtonal ncrement n dscrete educatonal attanment. It shows that returns to each level of educaton have rsen substantally from 1991 to Compared to senor hgh school, the annual return to educaton for unversty level ncreased from 4.24% n 1991 to 13.1% n 2000, and the annual return to educaton for non-unversty tertary level ncreased from 4.24% n 1991 to 10.2% n Compared to junor hgh school, the annual return to educaton for senor hgh school level ncreased from 2.77% n 1991 to 6.57% n Compared to prmary school, the annual return to educaton for junor hgh school level ncreased from 1.41% n 1991 to 3.99% n Snce n the data 1991, the unversty level and non-unversty tertary are combned together, so the return to educaton of ths group compared to senor hgh school s an average fgure. 11

12 In what follows, we wll use the smple Mncer equaton to examne returns to educaton by gender, ndustry, regon, and ownershp. The analyss wll further ndcate whether the rsng trend of returns to educaton s robust. Table 5 reports the 26 regressons, whch are run separately. All regressons show that trends of returns to educaton are rsng from 1991 to It s nterestng to see that returns to schoolng are consstently hgher for female workers than for male worker ether n 1991 or n Many exstng studes had the smlar results (e.g., Psacharopoulos, 1994). Though the average wage of female workers s lower than the male workers, the gap of wages among female workers maybe larger than the gap of wages among male workers. There exst small dfferences n returns to educaton across ndustres. All the returns to educaton n the 5 aggregated ndustres rose remarkably from 1991 to Especally, the return to educaton n ndustry32, whch s composed of most monopolstc and knowledgentensve ndustres, has become the largest one n More than half labors n ndustry32 have obtaned hgher educaton, whle there are only 4.66% labors n total ndustry are at hgher educaton level. The dfferences n returns to educaton across regons are also small. It s worthy to note that the order arranged by the value of rates of return to educaton of the three regons has reversed from 1991 to In 1991, the bggest rate of return to educaton of 3.41% belongs to regon3, whle the smallest one of 2.03% belongs to regon1. However, n 2000, the rate of return to educaton of regon1 becomes the bggest and reaches 9.14%, whle the rate of return to educaton of regon3 becomes the smallest and reaches 8.29%. The dfferences n returns to educaton across ownershps are also very small. Among three ownershps, the rate of return to educaton of ownershp3 s largest, and t rose from 3.80% n 1991 to 9.30% n Generally speakng, the ownershp3 ncludes foregn funded enterprses and prvate enterprses and market force has more mpacts on the wage dfferences n these sectors. It s also nterestng to note that returns to educaton wthn the tradtonal state and urban collectve sectors have also gone up qute dramatcally. Returns to educaton n the state sector went up from 2.45% n 1991 to 7.70% n For urban collectves, the 12

13 numbers are 2.46% and 7.99%, respectvely. 13

14 3. The estmaton of expected net revenues 3.1. The formula of expected net revenues In ths part, we wll compare statstcally the dfference of revenues among varous educaton levels, usng three ndcators ncludng expected total revenue (ETR) n lfetme, educaton cost (EC), and expected net revenue (ENR). The ENR of educaton level k s defned as ts ETR mnus ETR of educaton level k-1, and mnus the total educaton cost of educaton level k (nclude both educaton drect cost and educaton opportunty cost). The formula s as the followng: NR k = R R 1 C (3) k k k Where n R k = R tk t= 1 t ( 1 + λ ), m = t= 1 C k C tk m t ( 1 + λ ) ; λ s the dscount rate of future revenue and cost; n s the expected work years after the person fnshes educaton of level k; m s the years for the person to fnsh educaton of level k. The future revenue means the total revenue from the tme when the person fnshes educaton to the tme when the person retres. Snce the dscount rate λ s very hard to get, n order to make the above formula easy to estmate, cross-sectonal data are used wdely n exstng researches. For example, the educaton total revenue of a certan level s estmated as the sum of averaged annual ncome of by dfferent work-year. Snce the dstrbuton of the number of dfferent work-year n sample data s uneven, ths paper used the weghted averaged annual ncome (weght by frequency). The educaton cost of level k equals to ts drect cost and opportunty cost. The opportunty cost of educaton level k s calculated by the weghted annual ncomes of educaton level k-1 tmes the educaton years of educaton level k. The drect cost of educaton level k s estmated by tuton and fees per student. The tuton and fees per student are calculated by educaton expendture per student tmes the percent of tuton and fees n total educaton expendture. The data used to calculate the drect educaton cost are from the book of Strde from a Country of Tremendous Populaton to a Country of Profound Human Resources (Yuan etc., 2003). 14

15 15

16 3.2. Comparsons of averaged annual ncomes among varous educaton levels Table 6 shows that educatons ndeed have sgnfcant mpact on the annual ncome. In the year of 2000, ranked by the educaton degree from the top to the bottom, the weghted annual ncome on average of UNIV level, NUT level, SSS level, SHS level, JHS level, and PS level were Yuan, Yuan, Yuan, Yuan, 9032 Yuan, and 8937 Yuan, respectvely. General speakng, salary or annual ncome concdes wth the labor productvty. The above data showed that educaton have had remarkable postve mpacts on labor productvty, and hgher educated people are ndeed earnng more money n the labor market n Chna. Takng the senor hgh school as benchmark of 100, then the averaged ndces of relatve annual ncome were 138 and 86 for hgher educaton and for educaton below senor hgh school, respectvely. Durng , for 20 countres of OCED, the averaged ndces of relatve annual ncome were 143 for hgher educaton and 78 for educaton below senor hgh school (OECD, 2002). The ncome gap among three knd of educaton of Chna was less than OCED countres on average, but bgger than some OCED countres such as Sweden and Denmark Comparsons of expected net revenue n lfetme The above statstc data showed that people had more annual ncome f they receved more educaton. But, the above comparson ddn t take educaton cost nto account. To analyze the demand for a certan educaton level, both the revenue and the cost should be consdered together. Fgure 1 showed that all the expected net revenue of a certan educaton level compared to next level s also postve. Frst, compared to non-unversty tertary, the ENR of unversty educaton was 56,187 Yuan, and ths gap was bggest and sgnfcant. Second, compared to specalzed secondary school, the ENR of non-unversty tertary was 19,353 Yuan. Thrd, compared to senor hgh school, the ENR of specalzed secondary school was 23,476 Yuan. Fourth, compared to junor hgh school, the ENR of senor hgh school was 25,454 Yuan. Fnally, compared to prmary school, the ENR of junor hgh school was only 3,822 Yuan. 16

17 The data shows that the two bggest ENRs belong to unversty educaton and senor hgh school educaton, whch reveal also that the potental demand for these two level educatons should be stronger. Thus, there are needs for rapd expansons of both unversty educaton and hgh school educaton. 4. Conclusons Most exstng studes n the lterature are based on data from the 1980s and cover only part provnces of Chna and fnd that the rates of return to educaton n Chna are remarkably low. Ths study uses the most recent Chnese Urban Household Survey data that cover all provnces except Tbet n 1991 and 2000 to estmate returns to educaton n urban Chna. The objectve of ths study s to nvestgate econometrcally whether the returns to educaton has a sgnfcant rsng trend n the 1990s based on Mncer-type equaton, and to estmate statstcally whether the expected net revenues of educatons at varous levels are postve sgnfcantly when educaton costs are consdered. Several conclusons appear to be warranted based on the regresson results and statstc estmates. Frst, the Mnceran rate of return to educaton s postve sgnfcantly and ndeed has a remarkable rsng trend n the 1990s. The rate of return to educaton rose from 3.03% n 1991 to 8.47% n 2000 based on smple Mncer equaton, and ncreased from 2.07% n 1991 to 6.22% n 2000 when the gender, ndustry, regon, and ownershp dummy varables are added nto the regressons. Second, the expected net revenues at all educatonal levels are postve. Most exstng studes only use the rate of return to educaton to compare whether values are bg and whether the value ncreases. However, even the return to educaton s postve, t s unnecessary to guarantee there exsts postve expected net revenue when educaton costs are consdered. Ths study fnds that, compared to next level of educaton, the unversty, non-unversty tertary, specalzed secondary school, senor hgh school, and junor hgh school have expected net revenues of Yuan, Yuan, Yuan, Yuan, and 3822 Yuan, respectvely. Thrd, the dfference among varous educaton levels s very bg. The expected net 17

18 revenues for unversty educaton s bggest so that we can explan the reason why the hgher educaton expanson s so rapd n recent years. Though each unverstes of Chna started to charge tutons and fees n the late 1990s, and the number of tutons and fees ncreased very quckly n recent years, even let the students pay for the total recurrent expendture per student, the expected net revenues stll wll be postve. So the strong demand for hgher educaton n Chna wll contnue n the future. 18

19 Fourth, the expected net revenues for senor hgh school s the second bggest. However, the percentage of graduates of junor hgh schools enterng senor hgh schools n 2000 s only 51.1%. The senor hgh school educaton needs expanded urgently. Fnally, the expected net revenues for junor hgh school s the smallest. When educaton opportunty cost s consdered, ths fgure could be negatve. If the parents thnk that there s no hope for ther chldren to enter senor hgh school or unversty n the future, they wll choose not to let ther chldren enter the junor hgh school. Ths can partly explan why the dropout rate of Junor hgh school students s very hgh. References: Byron, Raymond P. and Evelyn Q. Manaloto, 1990, Returns to Educaton n Chna, Economc Development and Culture Change, 38, L, Hazheng (2003), Economc Transton and Returns to Educaton n Chna, Economcs of Educaton Revew, 22 ( ). Maurer-Fazo, M., 1999, Earnngs and Educaton n Chna s Transton to a Market Economy Survey Evdence from 1989 and 1992, Chna Economc revew, 10 (1): Mncer, J. (1974), Schoolng, experence and earnngs. New York: Columba Unversty Press for the Natonal Bureau of Economc Research. OECD, 2002, Educaton at Glance, OECD Indcators 2001, Head of Publcaton Servce. Psacharopoulos, George, 1994, Returns to Investment n Educaton: A Global Update, World Development, 22, Yuan, Guren etc., 2003, Strde from a Country of Tremendous Populaton to a Country of Profound Human Resources (n Chnese), Hgher Educaton Press, Bejng. Zhang, Junsen and Yaohu Zhao (2002), Economc Returns to Schoolng n Urban Chna, , paper presented at the 2002 meetngs of the Alled Socal Scences Assocaton, Washngton DC. State Statstc Bureau, 2002, Chna statstcal Yearbook, the Chna Statstcs Press. 19

20 Table 1 Varable defnton and descrptve statstcs Varable Defnton Mean 1991 Mean 2000 Schoolng Years of school UNIV Unverstes (Daxue Benke) 5.1% 8.5% NUT Non-unversty Tertary (Dazhuan) 10.3% 19.5% SSS Specalzed Secondary School (Zhongzhuan) 14.5% 14.9% SHS Senor Hgh School 26.9% 28.3% JSH Junor Hgh School 34.7% 25.4% PS Prmary School 8.1% 3.3% Experence Years Gender 1 for male, o for female 52.4% 54% Industry1 Farmng, Forestry, Anmal Husbandry and Fshery 1.1% 1.1% Industry2 Manufacturng; Constructon 37.5% 36% Industry31 Health Care, Sports & Socal Welfare; Educaton, Culture 20.2% 28.2% and Arts; Scentfc Research and Polytechnc Servces; Government Agences, Socal Organzatons Industry32 Fnance and Insurance; Real Estate; Socal Servces 6.2% 10.5% Industry4 Other ndustres 35% 24.2% Regon1 Bejng, Tanjn, Shangha 4.6% 8.8% Regon2 Coastal provnces (7) 30.4% 29.7% Regon3 Inland provnces (19) 65% 61.5% Ownershp1 State Owned Untes 81.3% 79.6% Ownershp2 Collectvely Owned Enterprses 18.1% 10.8% Ownershp3 Other Enterprses 0.6% 9.6% Observatons Sze of sample Data source: the Urban Socoeconomc Survey Teams of the State Statstcs Bureau, Chnese Urban Households Survey data of 1991 and

21 Table 2. Mnceran Rate of Return to Educaton of Chna n 1991 and 2000 Independent varables (1) (2) (3) (4) Const *** (0.0150) *** (0.0158) *** (0.0268) *** (0.0280) Schoolng *** (0.0010) *** (0.0011) *** (0.0017) *** (0.0018) Experence *** (0.0010) *** (0.0010) *** (0.0018) *** (0.0017) Experence *** (0.0002) *** (0.0000) *** (0.0000) *** (0.0000) Dummy_Gender *** (0.0055) *** (0.0085) Dummy_Industry ** (0.0264) *** (0.0398) Dummy_Industry *** (0.0116) *** (0.0147) Dummy_Industry *** (0.0073) *** (0.0112) Dummy_Industry ** (0.0066) *** (0.0110) Dummy_Regon *** (0.0129) *** (0.0173) Dummy_Regon *** (0.0058) *** (0.0092) Dummy_Ownershp *** (0.0073) *** (0.0139) Dummy_Ownershp *** (0.0374) *** (0.0152) Adj. R Observatons Data source: the Urban Socoeconomc Survey Teams of the State Statstcs Bureau, Chnese Urban Households Survey data of 1991 and

22 Table 3. The relatonshp between ndustry compostons of urban employed persons by educatonal attanment and average wage of staff and workers by ndustry Employment composton (2001) PS JHS SHS COLL UNIV Average wage (2000, Yuan) Farmng, Forestry, Anmal Husbandry and Fshery Mnng and Quarryng Manufacturng Producton and Supply of Electrcty Gas and Water Constructon Geologcal Prospectng and Water Conservancy Transport, Storage, Post & Telecommuncaton Servces Wholesale and Retal Trade & Caterng Servces Fnance and Insurance Real Estate Socal Servces Health Care, Sports & Socal Welfare Educaton, Culture and Arts, Rado, Flm and Televson Scentfc Research and Polytechnc Servces Government Agences Party Agences and Socal Organzatons Others Whole ndustres Data source: Chna Statstcal Yearbook,

23 Table 4. Annual returns to educaton by levels of educaton (%) Unversty over Senor hgh school Non-unversty tertary over Senor hgh school 10.2 Senor hgh school over Junor hgh school Junor hgh school over Prmary School Note: The results are run based on smple Mncer equaton (1). Table 5. Returns to educaton by gender, ndustry, regon, and ownershp (%) By gender: Male Female By ndustry: Industry Industry Industry Industry Industry By regon: Regon Regon Regon By ownershp: Ownershp Ownershp Ownershp Note: The results are run based on smple Mncer equaton (1). Table 6. Comparsons of revenues among varous educaton levels Educaton levels UNIV NUT SSS SHS JHS PS Annual ncome on average (Yuan) Income ndex (SHS=100) Expected work years n whole lfe Annual drect cost on average (Yuan) Net revenue compared to SHS (Yuan) Net revenue compared to lower level (Yuan) a Observatons Real work years on average Note: a, The rule of the prohbton of employng chldren ssued by the State Councl of Chna prohbt any chldren under 16 year old from beng employed. Hence, we suppose there s no opportunty cost for junor hgh school students. Data source: the Urban Socoeconomc Survey Teams of the State Statstcs Bureau, Chnese Urban Households Survey data of

24 JHS/PS SHS/JHS SSS/SHS NUT/SSS UNIV/NUT Fgure 1 the dfferences of expected net revenue of each educaton level compared to the next educaton level (Yuan, 2000) 24