Regional Inequality of Higher Education in China and the Role of Unequal Economic Development. by Frank Bickenbach, Wan-Hsin Liu

Transcription

1 Regonal Inequalty of Hgher Educaton n Chna and the Role of Unequal Economc Development by Frank Bckenbach, Wan-Hsn Lu No Aprl 2011

2 Kel Insttute for the World Economy, Hndenburgufer 66, Kel, Germany Kel Workng Paper No Aprl 2011 Regonal Inequalty of Hgher Educaton n Chna and the Role of Unequal Economc Development Frank Bckenbach and Wan-Hsn Lu Abstract: Over the past decade the scale of hgher educaton n Chna has expanded substantally. Regonal development polces attempted to make use of the scale expanson as a tool to reduce the nequalty of hgher educaton among dfferent regons wth dfferent development levels through provdng the poor regons preferental treatment and support n ths regard. Ths paper analyses a provncal dataset ( ), amng to provde comprehensve quanttatve evdence for the development of nequalty of hgher educaton opportuntes across provnces n Chna over the perod of the scale expanson, takng dfferent szes and economc development levels of provnces nto account. Results show that the regonal nequalty of hgher educaton relatve to provnces dfferent populaton szes clearly decreased over the research perod. Accompanyng the reducton n overall nequalty across provnces, the nequalty between the poor and the rch regons actually ncreased over the same perod. However, the ncrease was realsed n favour of the poor regon. The emprcal results are consstent wth the polcy orentaton of reformng the hgher educaton system and of promotng regonal development n Chna over the past decade. Keywords: hgher educaton, regonal nequalty, Chna, Thel ndex JEL classfcaton: C43. I23, I28, R53, R58 Frank Bckenbach Kel Insttute for the World Economy Hndenburgufer Kel, Germany Emal: frank.bckenbach@fw-kel.de Wan-Hsn Lu (Correspondng author) Kel Insttute for the World Economy Hndenburgufer Kel, Germany Emal: wan-hsn.lu@fw-kel.de Acknowledgement: We would lke to thank Mchaela Rank for her excellent research and techncal assstance. Lu would also lke to thank the German Research Foundaton (DFG) for ts fnancal support of the cooperatve project Regonal Aglty and Upgradng n Hong Kong and the PRD (Prorty Program 1233: Megactes Megachallenge: Informal Dynamcs of Global Change). The responsblty for the contents of the workng papers rests wth the author, not the Insttute. Snce workng papers are of a prelmnary nature, t may be useful to contact the author of a partcular workng paper about results or caveats before referrng to, or quotng, a paper. Any comments on workng papers should be sent drectly to the author. Coverphoto: un_com on photocase.com

3 1 Introducton Chna s strong economc growth over the past decades has attracted a great deal of attenton worldwde. The open door polcy snce the late 1970s has nduced the nflow of large foregn drect nvestments (FDI) especally n labour-ntensve ndustres, amng to proft from Chna s large resources of low-cost labour-force. As a result, the labour-ntensve ndustres have developed rapdly over the reform perod. The rapdly developng labour-ntensve ndustres are well ntegrated nto the global producton networks and are responsble for the low value-added part of the producton actvtes of the global value chan n partcular. Aganst the background of ntensfed market competton and a rsng challenge from ncreasng labour costs, the Chnese government has gradually revsed ts economcs polcy towards more nnovaton and upgradng promoton. A hghly qualfed labour-force s crucal for progressng n nnovaton and upgradng. To mprove the average educaton level of the Chnese labour-force and to ncrease the provson of hghly-qualfed workers for companes n Chna, the Chnese government has already made some progress n reformng ts hgher educaton system. Above all, over the past decade the scale of hgher educaton n Chna has expanded substantally. Local governments and unverstes have been granted more autonomy n managng unverstyrelated affars. Local governments have been encouraged to take over the operaton of some unverstes managed centrally before and to found new unverstes, takng nto account the local needs for promotng socetal and economc development. The central government has restrcted tself to focus on a much smaller number of role-model unverstes n Chna than before. 1 Though students from provnces other than the provnces where the role-model unverstes are located also have opportuntes through the natonal unversty entrance exam to study at these unverstes, the home-based pre-determned new student quotas of such unverstes restrct the cross-provncal moblty of students. 2 Over the reform process the local and regonal features of the hgher educaton system n Chna have become more pronounced (CCCPC (1985), (1993), (1999); NEC (1995); PRC (1999); SCPRC (1994)). 1 In 1998 about one thrd of all unverstes were centrally governed (MOE(a) (1999)). By contrast, n 2008 only about 5% of all unverstes were governed by central mnstres and agences (NBSC(a) (2009)). 2 Though the entrance exam s called the natonal unversty entrance exam, dfferent provnces may use dfferent exam sheets for the same subject. Moreover, students takng the exam compete drectly wth other students from the same provnce but not wth those from other provnces. The dstrbuton of students to unverstes s based on the students score-based postons among all exam partcpants from the same provnce and ther preference for majors and unverstes. However, the possblty of students to get access to unverstes, especally to non-local ones, strongly depends, n addton, on the pre-determned provnce-specfc new student quotas of the ndvdual unverstes (e.g. Chen (2004); He (2007)). 1

4 The scale expanson n hgher educaton has transformed the hgher educaton system n Chna from the one emphassng elte educaton to the one promotng mass hgher educaton, amng to enhance the average educatonal level and qualfcaton of the Chnese labour-force. The polcy decson for the scale expanson n hgher educaton (CCCPC (1999)) concerned all 31 provnces n manland Chna 3. However, due to regonal economc consderatons, there has been a partcular emphass by the central government on expandng hgher educaton n economcally backward provnces. Over the last decade, regonal economc polcy determned by the central government has ncreasngly ganed mportance. Its focus s on promotng economc development and/or ndustral structural change n the provnces n Western, Central and North-Eastern part of Chna 4. Compared to the provnces of the Coastal or Eastern regon 5, the economc development n these three regons s clearly laggng behnd. Takng nto account the dosyncratc characterstcs of the three regons, dfferent regonal polcy decsons have been announced by the central government. Reformng and/or expandng the local and regonal hgher educaton systems has been emphassed wth dfferent weghts n these regon-specfc polcy decsons. Accordng to the correspondng polcy decsons, the expanson and further mprovement of the hgher educaton system n partcularly poor provnces (Western regon) should obtan preferental treatment and support from the central government. The am s to support the convergence of the shares of the populaton wth hgher educaton n these provnces to the correspondng natonal average share (GOSCPRC (2001), (2002); OWDSCPRC (2002); SCPRC (2000), (2004)). For the North-Eastern and Central regons, whch are more developed than the Western regon, the regonal economc polcy plans also mentoned the mportance of hgher educaton. The promoton of hgher educaton and n partcular the ncrease of quanttatve hgher educaton opportuntes was less strongly emphassed for these regons, however, and no preferental treatment, comparable to that for the Western regon, was explctly provded (CCCPC (2003), (2006); GOCCCPC (2004); NDRC et al. (2007)). Have these polcy objectves and related decsons actually lead to a more equal dstrbuton of hgher educaton opportuntes among regons and provnces n Chna and have poor provnces actually benefted the most from the expanson of the hgher educaton system? 3 More specfcally, the provncal level dvson conssts of 22 provnces, 5 autonomous regons and 4 muncpaltes. For smplcty we wll refer to all of these unts as provnces. 4 The Western regon comprses 12 provnces (Chongqng, Gansu, Guangx, Guzhou, Inner Mongola, Nngxa, Qngha, Shaanx, Schuan, Tbet, Xnjang, Yunnan), the Central regon 6 provnces (Anhu, Henan, Hube, Hunan, Jangx, Shanx) and the North-Eastern regon comprses 3 provnces (Helongjang, Jln, Laonng). 5 The Coastal (Eastern) regon comprses 10 provnces: Bejng, Fujan, Guangdong, Hanan, Hebe, Jangsu, Shandong, Shangha, Tanjn, Zhejang. 2

5 Ths paper ams to answer these questons by provdng statstcal evdence on the development of the nequalty of hgher educaton opportuntes across provnces and regonally and economcally defned groups of provnces. In order to do so, t employs a provncal panel dataset for the perod from 1997 to 2008 to calculate a seres of generalsed Thel nequalty ndces wth dfferent regonal weghts and references, whch allow us to hghlght varous aspects of the ssue of the evoluton of the nequalty of hgher educaton opportuntes across (groups of) Chnese provnces. 6 The remander of the paper s organsed as follows. Secton 2 brefly descrbes some key quanttatve features of the (development of) the hgher educaton system n Chna and of nter-regonal dfferences n economc development. Secton 3 ntroduces the generalsed Thel ndex of nequalty (Secton 3.1), whch s then used to measure the nequalty n the dstrbuton of hgher educaton opportuntes across provnces (and groups of provnces) takng provncal heterogenety n sze and structural development nto account and to descrbe the evoluton of ths nequalty over tme (Secton 3.2). Secton 4 summarses the emprcal results and dscusses them n lght of the Chnese government s regonal economc polcy prortes. 2 The hgher educaton system and regonal economc development: an overvew Pror to the polces decson for a large scale expanson of hgher educaton opportuntes n 1999, the hgher educaton system n Chna was desgned to support hgher educaton of a rather small elte. There had already been some ncrease over tme n the number of unverstes and the number of students from 1978 (the year of the revval of the natonal unversty entrance exam) to That ncrease was rather modest, however, compared to the ncrease durng the perod snce 1999 n whch the hgher educaton system has been gradually transformed to promote mass hgher educaton. Whle t took about twenty years for the number of unverstes (or regular hgher educaton nsttutons) to ncrease from 598 unverstes n 1978 to 1,022 unverstes n 1998 (L and Xng (2010)), t just took another decade to more than double the number of unverstes to 2,263 n As regards the ncrease n the number of students, the dfference n growth rate was even larger (Fgure 1). Whle the number of newly enrolled students ncreased by about 5% annually from 0.4 mllon students n 1978 to 1.08 mllon n 1998 (L and Xng (2010)), t ncreased by more than 43% (to reach 1.55 mllon) n just one year from 1998 to Over the perod from 6 The provncal dataset was collected from the Educatonal Statstcs Yearbook of Chna (MOE(b) (varous years)) and from the Chna Statstcal Yearbook (NBSC(a) (varous years)). 3

6 1999 to 2008, the number of newly enrolled students contnued to ncrease at very hgh annual growth rate of an average 19%. As a result of the large scale expanson of the unversty system, there were about mllon unversty students n total n 2008 compared to just 3.41 mllon n In 2008, more than fve mllon students graduated from unversty, more than sx tmes the correspondng number n In total, about 6.7% of the Chnese populaton older than 6 (ncl.) years of age were hghly educated (wth unversty degree or hgher) n 2008, compared to 2.8% n Fgure 1: Number of unverstes and new students (n thousands) n Chna, unverstes new students n thousands un graduates n thousands Year Sources: MOE(b) (varous years). Classfyng the 31 provnces of Chna nto four regons (accordng to the regonal economc polcy classfcaton), the Coastal (Eastern) 8 regon s the regon wth the hghest level of development of both the economy and the hgher educaton system. The Coastal regon, whch accounts for roughly 10% of the geographc area of manland Chna, conssts of 10 provnces that together account for about 37% of the total Chnese populaton (n 2008). It was the poneer regon of the Chnese economc reform and s Chna s economcally the most advanced regon, wth ts Gross Domestc Product (GDP) per capta roughly 1.5 tmes the 7 These shares were calculated from offcal statstcs that were obtaned through the annual Natonal Sample Survey on Populaton Changes whch s based on a sample of roughly 1 of total populaton. 8 See Footnote 5 n Secton 1. 4

7 natonal average (25,012 RMB 9 ) n The regon hosted the largest share of unverstes (40% of unverstes n 2008 up from 38% n 1998) and the greatest share of new students was enrolled at the unverstes n the Coastal regon (40% n 2008 down from 43% n 1998). By contrast, the Western regon, whch conssts of 12 provnces 10 coverng more than seven tmes the area of the Coastal regon, s the economcally least developed Chnese regon, Its GDP per capta accounted for about 64 % of the natonal level (or about 43% of the level of the Coastal regon) n Whle the Western regon s populaton sze was about three fourth of that of the Coastal regon n 2008 (about 85% n 1998), the number of newly enrolled students n the West was only about half of that of the Coastal regon. More precsely, only about 24% of all unverstes were located n the West n 2008 (also 1998) and ts share of newly enrolled students was only 23% n 2008 (up from 21% n 1998). In sum, over the past decade of hgher educatonal reform there has been a massve expanson of the scale of the unversty system. Whle the number of unverstes more than doubled n just one decade the number of newly enrolled students grew even sxfold and the number of unversty graduates followed wth the correspondng three to four years delay. A frst bref comparson of some aggregate fgures for the most and the least advanced regons suggests that these massve changes may have come along wth surprsngly small changes n the regonal dstrbuton of unverstes and of unversty students among the Chnese regons. Whether ths frst mpresson s supported or refuted by a more systematc data analyss and whether t can possbly be generalsed to the dstrbuton of hgher educaton opportuntes across ndvdual provnces and (across and wthn) alternatve groupng of provnces wll be the subject of the next secton. 3 Inequalty of hgher educaton opportuntes across provnces 3.1 Measurng nequalty In ths paper we am to provde quanttatve evdence on the nequalty, or the concentraton, of hgher educaton opportuntes across the 31 Chnese provnces and ts development over tme. In addton, we am to trace the overall (changes n) nequalty to (changes n) the correspondng nequalty wthn and between meanngfully defned subgroups of provnces. To be able to do so we have to apply approprate measures of nequalty. In the economcs 9 Usng the average exchange rate (PBC (2008)), ths was roughly equal to US$3, See Footnote 4 n Secton 1. 5

8 lterature a large number of alternatve nequalty measures have been defned and appled to varous economc questons. Among the most frequently appled nequalty measures are the coeffcent of varaton, the Gn coeffcent, the Atknson class of measures, and the generalsed entropy (GE) class of measures. As the GE class of nequalty measures satsfes a number of crtera useful for our analyss, and as all nequalty measures that satsfy these propertes are ordnally-equvalent transformatons of GE measures (Cowell (2011): Secton 3.4), we choose GE measures for our analyss. One of the crtera that wll prove partcularly mportant for our analyss s the decomposablty of the measure. 11 Decomposablty mples that, for any mutually exclusve set of subgroups of provnces (whch may be defned on the bass of geographc (regonal) or economc crtera), the total nequalty across provnces can be meanngfully decomposed nto the nequalty wthn these subgroups and the nequalty between these subgroups. The GE class of nequalty measures s generally defned as: α I 1 1 X GE( α ) = α 2 α I = 1 1 1, (1) X I where I s the number of observatons (n our case provnces), X (=1,, I) s the realsaton of the varable of nterest (here the supply of hgher educaton opportuntes) for observaton, and X s the arthmetc mean of the varable of nterest over all observatons, X =(1/I)( X ). The parameter α, whch can take any real value, represents the GE-weght gven to dfferent observaton-specfc devatons from the mean. For lower values of α, the GE measure s more senstve to changes at the lower tal of the dstrbuton (of X), and for hgher values of α, the measure s more senstve to changes at the upper tal of the dstrbuton (Cowell and Flachare (2007)). For any α, the value of GE(α) s equal to zero n the case of complete equalty, X = X for all, otherwse t s strctly postve, wth hgher values of GE(α) representng hgher levels of nequalty (or concentraton) n the dstrbuton of X across observatons. The most frequently used values of α are 0, 1, and 2, where the formulae for GE(0) and GE(1) can be derved from formula (1) usng l'hoptal's rule. For α=0, we thereby get 11 The other crtera satsfed by the GE class of nequalty measures are the weak prncple of transfer, scale ndependence, and the prncple of populaton (for a dscusson see Cowell (2011): Secton 3.4). 6

9 GE(0) = 1 I I = 1 ln X X, a measure also known as the mean log devaton. (2) For α=1, we get GE(1) = 1 I X X ln X I = 1 X, a measure also known as the Thel ndex. 12 (3) For our emprcal analyss of the nequalty of hgher educaton opportuntes across Chnese provnces n the next secton, we wll manly use the Thel ndex (GE(1)). To test the robustness of our results and conclusons we also calculate the correspondng GE(0) and GE(2) measures. However, rather than confnng our analyss to the use of GE measures as defned by (1), and n partcular the Thel ndex as defned by (3), we wll make use of a slght generalsaton of the class of GE measures that has been ntroduced by Bckenbach and Bode (2008). Ths more generalsed form of GE measures wll allow us to better take nto account the heterogenety of provnces n terms of ther populaton szes or other relevant structural varables, whle nvestgatng the nequalty n hgher educaton n Chna. More concretely, followng Bckenbach and Bode (2008) we defne our generalsed, weghted relatve Thel ndex as follows: 13 GE wr X X I wr Π Π ( 1) = T = w ln, (4) X = X 1 w w Π Π where I and X are defned as before, and where s referred to as the reference for observaton and w (wth w = 1) s referred to as the weght of observaton. For the specal case of = 1 and w = 1/I for all we obtan our orgnal Thel ndex (3), whch we now call the unweghted absolute Thel ndex. Indces wth equal weghts (w = 1/I) and general references may accordngly be called unweghted relatve Thel ndces. As mentoned above, for our study the ndex denotes the 31 Chnese provnces (thus I=31). The varable X denotes the supply of hgher educaton opportuntes n provnce, whch we 12 The GE(2), whch can be drectly obtaned from formula (1), s a smple monotonc transformaton of the (frequently used) coeffcent of varaton (CV). More specfcally we have GE(2)=0.5CV The whole class of general entropy measures GE(α), ncludng the GE(0) and GE(2) measures used for our robustness tests, have been generalsed n the same way (for detals see Bckenbach and Bode (2008)). 7

10 proxy by the number of unverstes or the number of newly enrolled students n provnce, respectvely. Ths stll leaves us wth the choce of a set of weghts w and references. The role of weghts and references can best be llustrated by comparng formula (4) to formula (3) of the unweghted absolute Thel ndex. The provnce-specfc weght w, whch replaces the parameter 1/I of (3), enables us to redefne the basc unts of analyss and thus grant dfferent weghts to dfferent provnces. For example, nstead of gvng each provnce the same weght as n the unweghted Thel ndex (3), we may now decde to gve every nhabtant of Chna the same weght, whch mples that the provnces are assgned dfferent weghts correspondng to ther shares n total populaton when calculatng the nequalty measure. In other words, a gven level of undersupply (or oversupply) of hgher educaton opportuntes (relatve to the mean) n a specfc provnce s gven a greater weght n calculatng the nequalty measures, and s thus consdered to contrbute more to overall nequalty, f that provnce s larger n terms of populaton relatve to the other provnces. The second dfference to formula (3) s that a provnce-specfc reference s explctly consdered n the more generalsed formula (4). The ntroducton of provnce-specfc references enables us to nvestgate the nequalty across provnces of the dstrbuton of unverstes or unversty places relatve to relevant reference varables, whch we consder mportant n assessng the (nequalty of) hgher educaton opportuntes n the dfferent provnces. The absolute Thel ndex wthout provnce-specfc references (formula (3)) takes ts mnmum value of zero, whch s to be nterpreted as perfect equalty, f our varable of nterest X, e.g., the number of unverstes or unversty places, takes the same value for all provnces. The ntroducton of provnce-specfc references allows us to redefne ths perfect equalty benchmark (and the correspondng concept of equalty). For example, by takng the sze of the (young) populaton n the dfferent provnces as the reference mples that an equal dstrbuton of unversty places s no longer defned as the case of an equal absolute number of unverstes/unversty places n each provnce but as an equal rato of unverstes/unversty places to (young) nhabtants n each provnce. In other words, amng for equalty now mples that the number of unversty places should be proportonal to the sze of the young populaton. More generally, the relatve Thel ndex of nequalty s zero f 8

11 the dstrbuton of the varable of nterest X across provnces s proportonal to that of the reference varable, and t s strctly postve otherwse. 14 In our emprcal analyss of Secton 3.2 we wll also make use of dfferent specfcatons of weghts and references thereby focusng on dfferent facets of the ssue of the nequalty of hgher educaton opportuntes across Chnese provnces. As to the choce of weghts we wll calculate both unweghted Thel ndces (w =1/31) as well as populaton-weghted Thel ndces, where the weghts of the ndvdual provnces are gven by ther shares of total Chnese populaton. As to the choce of references, we wll make use of several alternatve varables total populaton, populaton under age 15, non-agrcultural employees and GDP (per capta) that correspond to dfferent supply sde, demand sde and development polcy consderatons on hgher educaton opportuntes. A dscusson of the motvaton for our dfferent choces of weghts and references and the mplcatons for the nterpretaton of the resultng measures are provded when we present our emprcal results n Secton 3.2. As mentoned above, we wll make explct use of the decomposablty property of the GE class measures of nequalty, whch allows us to trace (changes n) the nequalty of hgher educaton opportuntes across all provnces to (changes n) the nequalty wthn and (changes n) the nequalty between subgroups of provnces, that we defne on the bass of geographc (4 subgroups) and economc development (2 subgroups) crtera. Techncally, the decomposton of the weghted relatve Thel ndex, T wr, nto a wthn-group component and a between-group component s gven by: wr wr T = T wthn + T wr between = R r = 1 w r r w w w r X Π X Π r w w r r X Π w w r X Π ln r X Π w w r + X Π R r = 1 w r r w w w r X Π X Π ln r w w w r X Π X Π, (5) where r=1,, R denotes the mutually exclusve (sub)groups of provnces and I = r r w r w. 14 The upper bound of the weghted relatve Thel ndex s ln(1/w mn ), where w mn s the smallest weght. For an unweghted Thel ndex the upper bound s, thus, gven by ln(i). 9

12 The wthn-group component, wr T wthn, corresponds to the weghted average of the nequalty between provnces wthn each group r (r = 1,, R), whch s calculated based on the devatons of the relatve hgher educaton opportuntes (X/ ) of each provnce of a group from the (weghted) group mean. The between-group component, wr T between, corresponds to the nequalty n the hgher educaton opportuntes between the group and s calculated based on weghted devaton of the group mean of a certan group from the overall mean Regonal nequalty: emprcal results Absolute nequalty Aggregate statstcs presented n Secton 2 show that both the number of unverstes and the number of newly enrolled students (unversty places) have ncreased substantally snce Aganst the background of such a strong scale expanson, the dstrbuton of unverstes and unversty places between the most advances (Coastal) and the least advanced (Western) regons n Chna seemed to be only slghtly dfferent n 2008 from the correspondng dstrbuton n However, when turnng to a geographcally more dsaggregated level, even a bref look at the data shows that the number of unverstes, the number of students, and ther changes over tme dffer qute substantally across Chnese provnces. In 1997, the frst year of our observaton perod 16, there were only four unverstes n Tbet and fve n both Nngxa and Hanan, but there were 65 unverstes both n Bejng and n Jangsu. Whle there was a substantal ncrease n the numbers of unverstes n all provnces over the observaton perod, the dfferences across provnces remaned substantal. In 2008, there were sx unverstes n Tbet and nne n Qngha but as many as 125 n Guangdong and n Shandong and even 146 n Jangsu. The ncrease n the number of unverstes vared between 30.7% n Bejng (from 65 unverstes n 1997 to 85 n 2008) and 37.5% n Jln (from 40 to 55) to about 206% n Anhu (from 34 to 104) and even 220% n Hanan (from 5 to 16). In a large majorty of almost two-thrds of provnces the number of unverstes ncreased between 100% and 200%. 15 Smlar decomposton formulas hold for the GE(0) and the GE(2) measures, whch are used for robustness analyses n Secton There are two reasons for us to choose 1997 as the frst year of the panel dataset for analyss. Frst, ths was the year when Chongqng was upgraded to be the fourth muncpalty of Chna, next to Bejng, Tanjn and Shangha. Startng wth 1997 enables us to work on the largest balanced panel dataset, whch was avalable at the tme when data were collected. Second, startng wth 1997 but not later enables us to have at least two years of data for the perod before the large-sze scale expanson n hgher educaton n Chna. 10

13 As a measure of the overall nequalty of the number of unverstes across provnces, or the concentraton of unverstes across provnces, we calculated, for each year between 1997 and 2008, the unweghted absolute Thel ndex of the number of unverstes across provnces. The development over tme of ths Thel ndex s dsplayed n Fgure 2 (lne wth hollow squares). The value of the Thel ndex slghtly decreased from n 1997 to n 2008, wth a maxmum (mnmum) value of (0.131) n the year 1999 (2003). There s thus only lttle change and no clear tme trend n the measure between 1997 and Fgure 2: Inequalty n dstrbuton of unverstes and unversty places across provnces (unweghted absolute Thel ndex).3 unverstes unversty places Year Sources: MOE(b) (varous years). Own calculatons. As unverstes dffer n sze (the number of students), the number of unverstes n a provnce s obvously only a very rough measure of the opportuntes of potental students to obtan hgher educaton n the dfferent provnces. We may therefore prefer to look at the dstrbuton of the number of students, or even better the number of newly enrolled students, across provnces. Fgure 2, therefore, also dsplays, for each year, the unweghted absolute Thel ndex (of the concentraton) of the number of newly enrolled students across provnces (lne wth flled squares). A comparson of the two measures shows that the concentraton, or nequalty, of newly enrolled students across provnces s substantally larger than the 11

14 concentraton of the number of unverstes both n 1997 and n Actually the total number of newly enrolled students vares from 717 n Tbet and 2,619 n Qngha to 58,168 n Hube and 78,424 n Jangsu (more than 100 tmes the number of Tbet) n 1997 and from 8,520 n Tbet and 13,767 n Qngha to 410,705 n Jangsu and as many as 465,593 n Shandong n Smlar to the concentraton of unverstes, the concentraton of newly enrolled students across provnces s farly constant over tme. The unweghted absolute Thel ndex of newly enrolled students ncreased only very slghtly from about n 1997 to n 2008, wth a maxmum (mnmum) value of (0.207) n the year 2005 (2001). Ths reflects the fact that despte very hgh absolute growth rates of the number of newly enrolled students the dstrbuton of the 31 provnces shares of all newly enrolled students remans overall farly constant over tme. 18 As for the number of unverstes, ths does not mply that there was no change n the number of newly enrolled students or n the shares of ndvdual provnces n the total number of students newly enrolled, t rather suggests that the ncrease n the number of students was overall qute proportonal,.e. smlar n relatve terms, across provnces. Ths s true although the ncrease n the number of newly enrolled students vares from about 173% n Bejng (from 57,124 to 156,092) and 216% n Shangha (from 45,371 to 143,328) to 884% n Hanan (from 4,038 to 39,735) and 1,088% n Tbet (from 717 to 8,520). Stll for a large majorty of almost two-thrds of provnces the ncrease from 1997 to 2008 n the number of newly enrolled students les between 400% and 600%. Relatve nequalty The analyss so far does not consder dfferences n the sze of the populaton of the dfferent provnces. As the populaton szes of Chnese provnces vary very substantally n 2008 the populaton sze vared between less than three (2.9) mllon n Tbet and about 94 mllon n Shandong we can hardly consder t a reasonable poltcal objectve to have an equal or smlar absolute number of unverstes or unversty places (or students) n all provnces, dsregardng the dfferences n ther szes. In dscussng the nequalty n the supply of hgher educaton opportuntes across provnces we therefore have to take nto account the 17 Ths ndcates that unverstes are on average larger (n terms of newly enrolled students) n provnces wth a hgh numbers of unverstes as compared to provnces wth low numbers of unverstes. 18 Qualtatvely qute smlar results as for the Thel ndex,.e. the GE(1) the concentraton of newly enrolled students s larger than the concentraton of unverstes; both concentraton measures are farly constant over tme are obtaned for two alternatve general entropy (GE) measures, namely for the GE(0) and GE(2) (see Fgure A n Appendx). As ths s largely true for all of our analyss, we wll only focus on the results for the Thel ndex n the followng. 12

15 dfferences n the szes of the dfferent provnces; and n evaluatng changes n nequalty over tme we also have to take nto account that the populaton of provnces are growng at grossly dfferent rates, so that the relatve szes of the provnces as measured by ther shares n overall populaton are also changng over tme. For example, whle populaton shrank by about 6.7% n Chongqng and by about 3.5% n Schuan t ncreased by about 35%n Guangdong and by about 37% n Bejng over the tme perod consdered. Fundamentally, there are two mutually non-exclusve ways n whch our analyss can account for provnces dfferent populaton szes (cf. Secton 3.1). The frst way s to consder the dstrbuton across provnces of unverstes per capta (or newly enrolled students per capta) rather than the dstrbuton of the absolute number of unverstes (or newly enrolled students). In the terms of Secton 3.1, ths amounts to takng provnces populaton szes as the reference n calculatng relatve Thel ndces. The other way s to weght provnces by the sze of ther populaton when calculatng the ndex. Instead of provnces as n the unweghted Thel ndex t s the ndvdual nhabtants that are gven equal weghts n calculatng the populatonweghted Thel ndex. Ths mples that the undersupply of unverstes or unversty places n a gven provnce s taken to be a larger devaton from an equal supply, and thus a potentally larger problem for equalty, f ths provnce s larger n terms of populaton. Fgure 3 dsplays for each year between 1997 and 2008 the unweghted relatve and the populaton-weghted relatve Thel ndces both for the number of unverstes and for the number of newly enrolled students. Bearng the fndngs from Fgure 2 n mnd, there s a number of mportant observatons from Fgure 3: Frstly, at the begnnng of the observaton perod, the unweghted relatve Thel ndces for both the number of unverstes (hollow crcles) and the number of newly enrolled students (flled crcles) were even hgher than the correspondng unweghted absolute Thel ndces dsplayed n Fgure 2. Ths mples that n 1997 unverstes and unversty places were even more unequally dstrbuted across provnces once we consder ther supply relatve to the provnces dfferent populaton szes. Secondly, the weghted relatve Thel ndces for both the number of unverstes (hollow damond) and the number of newly enrolled students (flled damond) are lower n the early 13

16 years even substantally lower than the correspondng unweghted relatve measures. 19 They are also lower than the correspondng absolute measures. The lower values for the populatonweghted as compared to the unweghted relatve ndces ndcate that devatons from the average number of unverstes or students per nhabtant are on average more pronounced n smaller provnces (for more on ths see below). Thrdly and ths s the most mportant observaton from Fgure 3, and the most strkng dfference to Fgure 2 there s a strong and farly monotonc declne over tme of all relatve Thel ndces dsplayed n Fgure As a consequence these relatve nequalty measures are substantally lower than the correspondng absolute measures (c.f. Fgure 2) at the end of the observaton perod. Generally, the declne n nequalty s stronger for the number of newly enrolled students than for the number of unverstes. It s also slghtly stronger for the unweghted measures (crcles) than for the weghted measures (damonds). 21 The strong declne n the relatve measures suggests that the nequalty across provnces n the supply of hgher educaton opportuntes per capta has decreased substantally between 1997 and Whle there are some quanttatve dfferences between the development n the concentraton of unverstes per capta and the concentraton of newly enrolled students per capta, the general development s the same for both varables both have decreased substantally over tme. In the followng dscusson we therefore focus on only one of the two varables. Gven that unverstes may be of very dfferent szes the number of unversty places or newly enrolled students s generally a better proxy for the supply of hgher 19 Only n the early years of the observaton perod the unweghted relatve Thel ndex was hgher for the number of newly enrolled students (flled crcles) than for the number of unverstes (hollow crcles), as was the case for the absolute Thel ndces dsplayed n Fgure 2. (There was no such dfference for the weghted relatve Thel ndex.) In later years, by contrast, nequalty was hgher for the number of unverstes than for the number of newly enrolled students for both the weghted and the unweghted measures. Ths suggests that unverstes were on average larger (n terms of newly enrolled students) n larger provnces (n terms of populaton). 20 For most of the nequalty measures dsplayed n Fgure 3, there s a partcularly strong decrease of nequalty between 1999 and 2000 (we wll see smlar effects for several of the measures dscussed n the further course of our analyss). As 1999 and 2000 were the frst years of the new polcy of an enforced expanson of hgher educaton opportuntes and as the ncrease n the number of new students (but not the number of unverstes that saw ts greatest expanson between 2001 and 2004) was partcularly large n 1999 and 2000 (wth an ncrease of more than 40% n each of the two years), one could conjecture that the exceptonally strong declne n nequalty n these years may be a consequence of ths rapd expanson. We can show, however, that the strong decrease n 2000 s manly due to some extreme varaton n the populaton data for some of the provnces for the year Ths s also the reason why a correspondng declne for 2000 has not been observed for the absolute ndces dsplayed n Fgure 2 that do not make use of these populaton data.. 21 The dfference n the rate of declne between weghted and unweghted measures s relatvely small, however. For the number of unverstes (the number of new students enrolled) the unweghted relatve Thel ndex declnes by about 63.3% (77.4%) whereas the weghted relatve Thel ndex declnes (only) slghtly less by about 57.2% (74.6%). 14

17 educaton opportuntes than the number of unverstes. In the followng, we wll therefore focus on the concentraton of newly enrolled students. Fgure 3: Inequalty n dstrbuton of unverstes and unversty places across provnces relatve to populaton sze (unweghted relatve (UR) and weghted relatve (WR) Thel ndces).3.25 unverstes UR unverstes WR unvertty places UR unvertty places WR Year Sources: MOE(b) (varous years) and NBSC(a) (varous years). Own calculatons. The weghted relatve Thel ndex for the number of newly enrolled students (flled damonds) decreased substantally between 1997 and 2008 (from to 0.035). The correspondng unweghted relatve Thel ndex (flled crcles) s larger and decreased slghtly more strongly (from about n 1997 to about n 2008). The decrease of the two ndces ndcates a declne n the nequalty across provnces of the number of newly enrolled students per capta. For example, n 1997 there were as few as 29 newly enrolled students per 100,000 nhabtants n Tbet and just 36 per 100,000 nhabtants n Guzhou whle the correspondng numbers were 311 n Shangha and 461 n Bejng. In 2008 ths rato was hgher n all provnces; t was the lowest n Guzhou wth 226 and Yunnan wth 243 and the hghest n Bejng wth 921 and Tanjn wth a rato of 944. Whle the dfferences n these ratos across provnces were stll qute substantal n 2008 they were much smaller than n 1997 (n relatve terms). The hgher values of the unweghted as compared to the weghted Thel ndex reflects the fact the 15

18 provnces wth the lowest and those wth the hghest number of newly enrolled students per capta are generally of comparatvely small (Tbet, Tanjn, Bejng, Shangha), or medum (Guzhou, Yunnan) sze n terms of ther populaton whle the student to populaton ratos n the large provnces are closer to the average values across provnces. Gven the dfference dentfed between the results for the absolute nequalty ndces (Fgure 2) and the results for the relatve nequalty ndces (Fgure 3), we may wonder whether the results for the relatve ndces n partcular the strong ncrease of nequalty over tme hold true for alternatve reference varables that may be equally well or even better suted for our research purpose than the provnces total populaton szes. Two alternatve references are () the sze of the populaton under age 15 and () the number of non-agrcultural employees. Gven dfferences n the age structure of the provnces, the sze of the populaton under 15 as a proxy of the sze of the young populaton may be consdered a better ndcator of the regonal supply of potental students and thus the demand for unversty places than the overall populaton sze. Smlarly, gven dfferences n the (structural) economc development of the provnces the number of non-agrcultural employees may be consdered a better ndcator of the demand for hgher educaton graduates n the provnces labour markets than the overall populaton. 22 Fgure 4 dsplays the relatve weghted Thel ndces for the number of newly enrolled students for these two alternatve references (flled crcles for populaton under age 15 and flled squares for non-agrcultural employment). For ease of comparson Fgure 4 also dsplays once agan the relatve weghted Thel ndex wth the populaton sze as the reference (flled damonds) from Fgure 3. For all three Thel ndces we take provnces total populatons as weghts as before. Our choce not to use the populaton under age 15 or the non-agrcultural employment as weghts s based on the presumpton that an adequate supply of hgher educaton opportuntes s n the nterest of the whole populaton (and not just the young populaton or non-agrcultural employees) Our choce of non-agrcultural employment as a proxy for the demand for hgh-sklled workers s justfed by the fact that manufacturng and servces are more hgh-skll ntensve than (tradtonal) agrculture (NBSC(b) (2009)). Ths tendency s lkely to be renforced by the government-promoted shft towards the producton of more technologcally advanced products and an ncreasng mportance of more complex, knowledge-ntensve servces. 23 We have also calculated the unweghted relatve Thel ndces for the three references as well as the correspondng weghted and unweghted Thel ndces for the number of unverstes. The results of these calculatons, whch are not presented here, are largely n lne wth the results presented here. 16

19 Fgure 4: Inequalty n dstrbuton of unversty places across provnces relatve to alternatve sze measures (weghted relatve Thel ndces).3.25 unversty places - total populaton unversty places - young populaton unversty places - non-agrcultural employees Year Sources: MOE(b) (varous years) and NBSC(a) (varous years). Own calculatons. From Fgure 4 we observe that the weghted relatve Thel ndces for all three references are fallng between 1997 and Whle the declne n the ndex wth total populaton as reference was partcularly strong (75% n relatve terms), the absolute declne n the ndex wth non-agrcultural employment as reference was comparatvely small. Nevertheless, ths ndex was stll more than 18% smaller n 2008 than n As compared to ts maxmum value n 1998 (0.059) the relatve declne was even more than 33%. 24 Taken the populaton under age 15 as reference, the Thel ndex declned by about 51%. The results for the alternatve reference varables thus confrm the observaton that the regonal nequalty n the supply of unversty places has strongly decreased between 1997 and 2008 once we take nto account the provnces vastly dfferent populaton szes (relatve measures). Whle the decrease was somewhat weaker for the alternatve references as compared to the populaton reference, t was stll qute consderable n relatve terms. 24 The unweghted relatve measure wth the number of non-agrcultural employees as reference (not presented here) declnes even more strongly namely by more than 40% from n 1997 to n

20 As to the levels of nequalty, Fgure 4 shows that the nequalty across provnces was substantally larger for the number of newly enrolled students per nhabtant under age 15 than for the number of newly enrolled students per nhabtant. In contrast, relatve to the number of non-agrcultural employees the nequalty across provnces of the number of newly enrolled students was much lower than the nequalty relatve to the sze of total populaton n the early years of the observaton perod. Both measures were of smlar sze at the end of the observaton perod, however. The dfferences n the levels of the Thel ndces for the three references are reflected by correspondng dfferences n the span of values of the relatve numbers of newly enrolled students. For the number of newly enrolled students per nhabtant under age 15 the span between the maxmum and the mnmum rato across provnces s much hgher n relatve terms than the correspondng span for the number of newly enrolled students per nhabtant whch agan s much hgher than the span for the number of newly enrolled students per non-agrcultural employee. 25 These dfferences result from the structural dfferences between those provnces that have hgh ratos of students per capta and those that have low ratos. Provnces wth comparatvely hgh students per capta tend to have an older populaton and a larger share of manufacturng and servce sector employment. Compare, for example, the provnces that have the lowest numbers of students per capta, namely Tbet and Guzhou n 1997 and Guzhou and Yunnan n 2008, wth those provnces that have the hghest number of students per capta, namely Shangha and Bejng n 1997 and Tanjn n Tbet, Guzhou and Yunnan stll have a much hgher share of young people n total populaton and they also stll have a much hgher share of agrculture n total employment compared to Shangha, Bejng and Tanjn. Both wth respect to the demographc change towards an older populaton and wth respect to the structural transton of the economy from agrculture to manufacturng and servces the frst group of provnces s clearly lackng behnd the average Chnese provnce, whereas the second group s leadng the Chnese average n these transton processes. Across all provnces, there s a strong (but decreasng) negatve correlaton between the share of populaton under age 15 n total populaton and the share of non-agrcultural employment n 25 The number of newly enrolled students per 100,000 nhabtants under age 15 ranges from 88 n Tbet and 124 n Guzhou to 1,984 n Shangha and 2,738 n Bejng n 1997 and from 855 n Guzhou and 1,080 n Yunnan to as many as 9,544 n Shangha and 9,628 n Bejng n 2008, correspondng to a rato between mnmum and maxmum of about 1:30 n 1997 and 1:11 n By comparson, the number of newly enrolled students per 100,000 non-agrcultural employees ranges from 209 n Zhejang and 222 n Henan to 655 n Shangha and 968 n Bejng n 1997 and from 782 n Guzhou and 812 n Zhejang to 2,557 n Shaanx and 2,612 n Tanjn n 2008 correspondng to a rato of less than 1:5 n 1997 and close to 1:3 n For the total populaton reference the correspondng ratos were about 1:12 n 1997 and more than 1:4 n 2008 (for detaled fgures see above). 18

21 total employment. 26 The share of non-agrcultural employment n total employment tends to be largest n the Eastern/Coastal provnces where the openng of the Chnese economy and the accompanyng economc reform process started and where structural change and ncome growth have been strongest. At the same tme, these provnces tend to have a larger share of adults n total populaton due to a strcter enforcement of the One-Chld Polcy and an mmgraton of workers (and students) from less advanced, poorer provnces. The results dsplayed n Fgure 4 have clearly shown that takng nto account the provnces (hghly) dfferent (populaton) szes, the nequalty across the 31 Chnese provnces n the number of newly enrolled students, as measured by the (populaton-) weghted relatve Thel ndex, has clearly decreased snce In prncple, such a declne n the overall nequalty across provnces may go along wth an ncrease n the nequalty between larger regons (or groups of provnces), such as between the comparatvely well developed coastal provnces on the one hand and the less developed central or Western provnces on the other hand. To see whether ths s actually the case we have adopted the regonal classfcaton used by the Chnese government to desgn ts regonal development polcy, whch assgns the 31 provnces, accordng to ther geographcal locaton and developng status, nto four regonal groups: the Eastern or Coastal regon, the North-Eastern regon, the Central regon, and the Western regon. Makng use of the well-known decomposton propertes of the Thel ndex (see Secton 3.1), we have then decomposed each of the three Thel ndces from Fgure 4 nto a wthn-group component and a between-group component. For each of the three references the wthn-group component s equal to the weghted average of the four (respectve) relatve Thel ndces across the provnces wthn each of the four regons. It s a measure of the average nequalty n the number of newly enrolled students across the provnces wthn the ndvdual regons. The between-group component s the relatve Thel ndex across the four regons. It s a measure of the nequalty n the number of newly enrolled students across the four regons. Fgure 5 dsplays the resultng wthn-regon components (left panel) and between-regon components (rght panel) for each of the three weghted relatve Thel ndces from above The correlaton coeffcent between the two ratos has been n 1997 and n Thus the weghted relatve Thel ndex wth populaton sze as reference (flled damonds n Fgure 4) s, for nstance, equal to the sum of the correspondng wthn-group (flled damonds n Fgure 5) and between-group components (hollow damonds n Fgure 5). 19

22 Fgure 5: Inequalty n dstrbuton of unversty places across provnces decomposed by regons (weghted relatve Thel ndces from Fgure 4) (a) wthn-group nequalty (b) between-group nequalty.2 unversty places - total populaton unversty places - young populaton unversty places - non-agrcultural employees.2 unversty places - total populaton unversty places - young populaton unversty places - non-agrcultural employees Year Year Sources: MOE(b) (varous years) and NBSC(a) (varous years). Own calculatons. For all three references the wthn-regon component, whch s n all cases substantally larger than the between-regon component, declned strongly between 1997 and For the between-regon component there s a smlar declne for the total-populaton reference (hollow damonds) and young-populaton reference (hollow crcles). 28 In contrast, relatve to the number of non-agrcultural employment the between-group nequalty of newly enrolled students ncreased between 1997 and Actually, for ths reference the between-regon measure was about twce as hgh n 2008 as n 1997, whereas the wthn-regon measure has decreased by about a thrd. Whle the between-regon nequalty for non-agrcultural employment was comparatvely low n all years and took ts hghest value n 1999, where t was slghtly hgher than n 2008, the development of ths between-group nequalty mght nevertheless rase concerns that the regonal nequalty n hgher-educaton opportuntes may actually ncrease n the (unequal) process of economc development n Chna. Inequalty n hgher educaton and regonal economc development In further nvestgatng ths ssue t s useful to take dfferences n provnces economc development, measured by GDP per capta of Chnese provnces, more drectly nto account both n defnng the reference of the nequalty measure and n defnng groups of regons for our decomposton analyss. We thus calculated the weghted 29 relatve Thel ndces for the 28 For both these references the relatve declne of the between-regon component s of very smlar sze as the declne of the wthn-regon component. Both components declne by about 75% for the populaton reference and by close to 50% for the populaton-under-15 reference. 29 As n the prevous secton, the populaton share by provnce s used as the weght here. 20