Measuring the impact of educational expenditures on economic growth: evidence from Pakistan

Transcription

1 Educaional Research (ISSN: 4-56) Vol. (3) pp , December Special Issue Available hp:// Copyrigh Inernaional Research Journals Full Lengh Research Paper Measuring he impac of educaional expendiures on economic growh: evidence from Pakisan Saima Riasa, Rao Muhammad Aif and Khalid Zaman 3* //3 Deparmen of Managemen Sciences, COMSATS Insiue of Informaion Technology, Abboabad, Pakisan Acceped December, Educaion plays a vial role in human capial formaion. I raises he produciviy and efficiency of individuals and hus produces skilled manpower ha is capable of leading he economy owards he pah of susainable economic developmen. The objecive of his sudy is o empirically invesigae he impac of educaion expendiures on economic growh of Pakisan over a period of 97 o, using bonds esing approach. The empirical resuls show ha educaion expendiures have a posiive and significan impac on economic growh in he long run. The elasiciies of he model sugges ha if here is a one percen increase in educaion expendiures, i increases up o.39 percen in oupu in he long run. Keywords: Educaion expendiures, economic growh, employmen, bonds esing approach, Pakisan. INTRODUCTION Educaion plays a vial role in human capial formaion. I raises he produciviy and efficiency of individuals and hus produces skilled manpower ha is capable of leading he economy owards he pah of susainable economic developmen (Zaman, 8). Moreover educaion is expeced o ac posiively owards world economic developmen and povery alleviaion, boh of which are he prioriies of he world communiy. An invesmen in human capial, especially in educaion allows each person o conribue o heir sociey in a producive way. I becomes an imporan facor of an economy s capabiliy o achieve high level of growh wih low unemploymen, high wages and srong social uniy. Therefore, along wih elemens such as low unemploymen and balance of paymens equilibrium, educaion is an imporan issue ha each counry s governmen deals wih and srives o improve. Conribuion of educaion o economic developmen works in wo ways: firsly, hrough he economy's organizaion ha is is division of asks and secondly hrough he economy's performance ha how much i produces. The organizaion of an economy is geing *Corresponding Auhor khalidzaman@cii.ne.pk; Phone #: specialized in is division of asks. These asks include schools ha rain sudens for efficien and effecive performance. Produciviy of he labor force deermines he performance of economy. Because he educaional level of he labor force is a deerminan of is produciviy, schools and oher educaional faciliies make an imporan conribuion o economic developmen. All his could be made possible by a reasonable financing o educaion secor boh by governmen and educaion secor iself. In Pakisan, public expendiure on educaion lies on he fringes of % of GDP. According o Pakisan living and social sandard survey (8-9) repor, public expendiure on educaion as a percenage of GDP was acually reduced in 6 years and mainained for 5 years beween and 8-9. Thus, ou of oal 37 years since 97, public expendiure on educaion as a percenage of GDP eiher decreased or remained sagnan for years. Governmens spending on educaion in Souh Asia currenly average abou 4. percen of GDP (GoP, ). The governmen spending increased from 3.% in 98 o 6.% in 99 while i decreased sharply o 9.9% in 9. Expendiures for he period of 98-9 are demonsraed in Figure. Pakisan s economy has undergone a variey of sages over he 3 year period (98-9) which comprised siuaions of decreased and elevaed growh

2 84 Educ. Res Toal Expendiure as % of GDP GEXP Years Figure. Governmen Expendiures of Pakisan (98-9) Source: ADB (9) Growh Rae (%) Years GDP Figure. GDP growh Rae of Pakisan (98-9) Source: GoP () which provides wih an ineresing case sudy. The daa illusraes ha economic growh descended seadily from 5.4% in 98 o.% in 9. GDP growh rae of Pakisan from 98 o 9 is given in Figure. The above discussion confirms a srong linkage beween educaion expendiures and economic growh. The objecive of his paper is o criically examine he impac of educaion expendiures on economic growh of Pakisan, using ime series daa from 97-. The more specific objecives are: i. How much GDP growh change wih he change of educaion expendiures in he long run and shor run? ii. To esimae he dynamic shor-run causaliy effecs of educaion expendiures, gross capial formaion and employmen owards economic growh in Pakisan. A coinegraion echnique is used for analysis. The Auo-Regressive Disribued Lag (ARDL) model is used o calculae shor-run and long-run esimaes. The sudy arrange in he following manners: afer inroducion, Secion describes lieraure review. Daa Source and Mehodological Framework are included in Secion 3 o share vision wih he readers. Resuls and discussion are carried ou in Secion 4. Conclusion of he sudy is presened in he las. Lieraure Review There are number of empirical sudies regarding he impac of educaional expendiures on economic growh, however, hose sudies gives conradicory resuls, for examples, Barro (99) and Benhabib and Speigel (994) esimaed a posiive relaionship beween educaion expendiure and economic growh. However, Devarajan e al. (996) find a negaive relaionship beween educaion expenses and economic growh. Bose e al. (3) sudy he growh effecs of governmen expendiure, paricularly focusing on secoral expendi-

3 Riasa e al. 84 ures for a panel of hiry developing counries by using daa from 97 o 99. They find ha he share of governmen capial expendiure on gross domesic produc is posiively and significanly correlaed. Moreover, a he secoral level, governmen invesmen and oal expendiures in educaion are significanly associaed wih growh. Todd and Kenneh (3) illusrae he linkage beween governmen educaion expendiures and economic growh. An imporan aspec of his linkage is ha i is exremely hard o demonsrae any effec of increased funding o schools. They find an insignifican relaionship beween hese wo variables. Teles and Andrade (8) sudied he relaion beween governmen spending on basic educaion and he human capial accumulaion process. They observed he impacs of spending on economic growh and individual invesmens in higher educaion and ha he significance of he relaion beween public spending on educaion and economic growh is alered by changes in he composiion of governmen spending wih regard o basic and higher educaion, and his relaion may be insignifican when higher educaion is no promoed. Blankenau e al (4) find ha here exiss a posiive relaionship beween public educaion expendiures and long run growh for developed counries. By using panel daa from 3 counries hey concluded ha his relaionship is sensiive o he imposiion of he governmen budge consrain and here are no significan growh effecs of public educaion expendiures. Angelopoulos e al. (7) finds ha public spending on educaion brings growh and promoes he welfare. They furher argue ha if he composiion of public spending can be changed so ha he share of educaion spending is more han oher componens of oal governmen spending hen he welfare gains will increase by four percen of consumpion. Akram and Khan (7) examine he incidence of governmen spending on educaion in Pakisan a boh urban and rural provincial level by employing he hreesep Benefi Incidence Approach mehodology on primary daa of he Pakisan Social Sandard Living Measures Survey 4-5. They find ha a regional and he provincial levels he overall expendiure on he educaion secor is progressive in Pakisan. While some dispariies exis in he shares of differen income groups benefi from he provision of educaional faciliies creaed by public expendiures. Nurudeen and Usman () find ha governmen oal capial expendiure, oal recurren expendiures and governmen expendiure on educaion have negaive effec on economic growh. While rising governmen expendiure on ranspor and communicaion, and healh resuls o an increase in economic growh. Afzal e al. () find he exisence of direc relaionship beween school educaion and economic growh in Pakisan. Macroeconomic insabiliy due o inflaion reards economic growh boh in he shorrun and he long-run while i reards school educaion only in he long-run. They observed a saisical significan and inverse relaionship beween school educaion and economic growh in he shor-run. Faridi e al () find a posiive relaionship beween higher educaion and earnings of he sudens per monh. They use he Mincerian human capial model. They sugges ha governmen should increase is expendiures on educaion by giving scholarships o he sudens and by making proper arrangemens for he needy sudens because i could have ulimae posiive effec on economic growh. Educaion expendiures and economic growh have been well debaed in developing counries, hence here is a pressing need o evaluae and analyze he educaiongrowh nexus and o find ou he iner relaionship. In he subsequen secions an effor has been made o empirically find ou he long-run relaionship beween educaion expendiures and economic growh in he conex of Pakisan. METHODOLOGICAL FRAMEWORK Theoreical Framework The model used in his paper is based on he following aggregae producion funcion i.e., () Where Y is real income ha is real GDP K is physical capial L is he number of workers H is he oal amoun of human capial A is he echnology parameer is he observaion subscrip and, α, β, and γ are parameers o be esimaed. Human capial can be defined as he produc of average level of educaion per worker (E ) and number of oal workers (L ). Mahemaically we can wrie i as;.. () I is assumed ha he average level of educaion per worker and average expendiure on educaion per worker has direc relaionship. Subsiuing equaions () ino () obains;.. (3) Where δ = β + γ. I is from equaion (3) ha we develop he economeric equaion and uses i o assess empirically he impac of governmen educaion expendiures on economic growh. Theoreically, a posiive correlaion is expeced beween growh in oupu

4 84 Educ. Res. on one hand and increases in capial sock, employmen and educaion of workers on he oher hand. The daa used in his sudy is aken from differen sources, i.e., GoP () and ADB (9). Empirical Model Bound Tesing Approach The use of he bounds echnique is based on hree validaions. Firs, Pesaran e al. () advocaed he use of he ARDL model for he esimaion of level relaionships because he model suggess ha once he order of he ARDL has been recognized, he relaionship can be esimaed by OLS. Second, he bounds es allows a mixure of I() and I() variables as regressors, ha is, he order of inegraion of appropriae variables may no necessarily be he same. Therefore, he ARDL echnique has he advanage of no requiring a specific idenificaion of he order of he underlying daa. Third, his echnique is suiable for small or finie sample size (Pesaran e al., ). Following Pesaran e al. (), we assemble he vecor auoregression (VAR) of order p, denoed VAR (p), for he following growh funcion: Z p = + β i z i i = µ + ε where z is he vecor of boh x and y, where y is he dependen variable defined as economic growh (GDP), x is he vecor marix which represens a se of explanaory variables i.e., gross fixed capial formaion (GFC), employmen (E) and educaion expendiures per worker (EEWE) and is a ime or rend variable. According o Pesaran e al. (), y mus be I() variable, bu he regressor (4) x can be eiher I() or I(). We furher developed a vecor error correcion model (VECM) as follows: p i i p z = µ + α + λz + γ y + γ x + ε where is he firs-difference operaor. The long-run muliplier marix λ as: λyy λyx λ = λxy λxx The diagonal elemens of he marix are unresriced, so he seleced series can be eiher I() or I(). If λ YY =, hen Y is I(). In conras, if λ YY <, hen Y is I(). The VECM procedures described above are imperaive in he esing of a mos one coinegraing vecor beween dependen variable y and a se of regressors x. To derive model, we followed he posulaions made by i (5) Pesaran e al. () in Case III, ha is, unresriced inerceps and no rends. Afer imposing he resricions λ YY =, µ and α =, he hypohesis funcion can be saed as he following unresriced error correcion model (UECM): ( GDP) = β + β ( GDP) + q β ( GFC) 6 i + r β ( E) 7 + β ( GFC) i + s + β ( E) β ( EEWE) 8 3 i + β ( EEWE) + u 4 β ( GDP) (6) Where is he firs-difference operaor and u is a whie-noise disurbance erm. GDP = Gross Domesic Produc (GDP) in US $ million; GFC = Gross Fixed Capial Formaion in percenage; E = Toal Employmen in percenage; EEWE = Governmen educaion expendiures per worker. Equaion (6) also can be viewed as an ARDL of order (p, q, r, s). Equaion (3) indicaes ha economic growh ends o be influenced and explained by is pas values. The srucural lags are esablished by using minimum Akaike s informaion crieria (AIC). From he esimaion of UECMs, he long-run elasiciies are he coefficien of one lagged explanaory variable (muliplied by a negaive sign) divided by he coefficien of one lagged dependen variable (Bardsen, 989). For example, in equaion (3), he long-run invesmen, employmen and educaion expendiures per worker elasiciies are ( β / β ), ( β / β 3 ) and β / β 4 respecively. The shor-run effecs are capured by he coefficiens of he firs-differenced variables in equaion (6). Afer regression of Equaion (6), he Wald es (Fsaisic) was compued o differeniae he long-run relaionship beween he concerned variables. The Wald es can be carry ou by imposing resricions on he esimaed long-run coefficiens of economic growh, invesmen, employmen and educaion expendiures per worker. The null and alernaive hypoheses are as follows: H β = β = β = β (no long-run relaionship) : 3 4 = Agains he alernaive hypohesis H β β β β (a long-run relaionship A : 3 4 exiss) The compued F-saisic value will be evaluaed wih he criical values abulaed in Table CI (iii) of Pesaran e al. (). According o hese auhors, he lower bound criical values assumed ha he explanaory variables x are inegraed of order zero, or I(), while he upper bound criical values assumed ha x are inegraed of order one, or I(). Therefore, if he compued F-saisic is + p 5 i

5 Riasa e al. 843 Table. Augmened Dickey-Fuller (ADF) Tes Level s Difference Variables Consan Consan Order of Inegraion GDP.53() () *** I() E -.779() () *** I() GCF.396() () *** I() EEWE -.64*() -4.65() I() Noe: The null hypohesis is ha he series is non-saionary, or conains a uni roo. The rejecion of he null hypohesis is based on MacKinnon (996) criical values. The lag lengh are seleced based on SIC crieria, his ranges from lag zero o lag four. *** represens significan a percen level. Table. Esimaed Models based on Equaion (6)Dependen Variable: log(gdp) Variable Coefficien -Saisic Prob. β log( GDP ) log( GCF ) log( E ) log( EEWE ) log(gcf) log( GCF ) log(e) log( E ) log(eewe).76795** *** * ** *** *** ** Model crieria / Goodness of Fi: R-square =.79; Adjused R-square =.699; Wald F-saisic = 8.44 [.] *** Noe: ***, ** and * indicae significance a.,.5 and. level respecively. smaller han he lower bound value, hen he null hypohesis is no rejeced and we conclude ha here is no long-run relaionship beween economic growh and is deerminans. Conversely, if he compued F-saisic is greaer han he upper bound value, hen economic growh and educaion expendiures have a long-run level relaionship. On he oher hand, if he compued F- saisic falls beween he lower and upper bound values, hen he resuls are inconclusive. RESULTS AND DISCUSSION The sandard Augmened Dickey-Fuller (ADF) uni roo es was exercised o check he order of inegraion of hese variables. The resuls obained are repored in Table. Based on he ADF es saisic, i was concluded ha he GDP, Gross fixed capial formaion (GCF) and employmen (E) are non-saionary series a level. Though, when we ake he firs difference of hese variables, hen hey becomes saionary i.e., I(). Governmen educaion expendiures per worker (EEWE) is saionary a level, herefore, we considered as I() variable. Noiceably, he mixure of boh I() and I() variables would no be possible under he Johansen procedure. This gives a good jusificaion for using he bounds es approach, or ARDL model, which was proposed by Pesaran e al. (). The uni roo es explains ha all variables are saionary a firs difference excep he educaion expendiures per workers which is saionary a level. In he nex sep, we esimae he model in order o analyze he impac of educaion expendiures on GDP growh boh in shor-run and long-run. The ime series propery of economerics may compel us o employ he auoregressive disribuive lag model (ARDL) due o he mixure of I() and I() variables. The esimaion of Equaion (3) using he ARDL model is repored in Table. We follow he general o specific rule in order o remove he highly insignifican or less

6 844 Educ. Res. Table 3. Diagnosic and Sabiliy Tes Tess F-Saisics Probabiliy NORM WHITE RAMSEY ARCH SerialCorr Noe: For normaliy es, we repor Jarque-Bera saisics. NORM, WHITE, RAMSEY, ARCH, SerialCorr are non-normal errors normaliy es, whie heeroskedasiciy es, Ramsey Regression Specificaion Error Tes, and Auo regressive Condiional Heeroskedasiciy (ARCH Tes), Serial correlaion Lagrange Muliplier Tes (LM-ype Breusch-Godfrey- Tes). These saisics are disribued as Chi-square values and capure degree of freedom in parenhesis. significan variables from our regression analysis. This process may help us o follow he imporan propery of auoregressive disribuive lag (ARDL) model and make our regression analysis mos appropriae or desirable. The above procedure in esimaing equaion may appear as resul of one or none lags wih he explanaory variables. Ineresingly, i capures all lags ha may have desirable significance power in erm of saisical inference. Table shows ha all he variables are significan in he long run. The value of R-squared and adjused R-squared is.79 and.699 respecively. The resuls of he bounds co-inegraion es demonsrae ha he null hypohesis of agains is alernaive is easily rejeced a he % significance level. The compued F-saisic of 8.44, ha is greaer han he upper bound value of 5.6, hus indicaing he exisence of a seady-sae long-run relaionship among GDP, GCF, E and EEWE. Table 3 show he saisics of differen diagnosic and sabiliy es o check overall sabiliy of he model. The robusness of he model has been definie by several diagnosic ess such as Breusch- Godfrey serial correlaion LM es, ARCH es, Jacque-Bera normaliy es and Ramsey RESET specificaion es. All he ess disclosed ha he model has he aspiraion economeric properies, i has a correc funcional form and he model s residuals are serially uncorrelaed, normally disribued and homoskedasic. Therefore, he oucomes repored are serially uncorrelaed, normally disribued and homoskedasic. Hence, he resuls repored are valid for reliable inerpreaion. The plos of CUSUM and CUSUM-square saisics are in criical bounds and hey do no diverge from ha criical region ha indicaes he sabiliy of coefficien in he esimaed model as shown in Figure 3. I also confirms ha he sabiliy of long and shor-run esimaes of he model from GDP o educaion expendiures. The absence of divergence and presence of convergence mainly suppor o he long-run as well as shor-run sabiliy of he model paricular o he long-run esimaes. The esimaed coefficiens of he long-run relaionship beween GDP growh rae, Gross fixed capial formaion, employmen rae and educaion expendiures per worker are shown in Table 4. The resuls reveal ha here is a posiive and significan relaionship wih GCF, E and EEWE on GDP boh in he long run and he shor-run. The coefficien of.4 and.34 suggess ha if here is one percen change in he capial formaion, i increases he GDP o.4 percen in he shor run and.34 percen in he long run. Similarly, if here is one percen increase in employmen rae, i increases he GDP by.5 percen in shor run and.74 percen in long run. Educaion expendiure is saisically insignifican in he shor run bu in he long run is impac is quie significan. The coefficien of.39 depics ha if here is one percen increase in educaion expendiures, i increases he GDP by.39 percen. CONCLUSION The objecive of he sudy is o measuring he impac of educaion expendiures on economic growh of Pakisan from 97-. The esimaed resuls confirm ha educaion expendiure have a significan impac on longrun economic growh. The resuls conclude ha if here is one percen increase in he capial formaion, i increases

7 Riasa e al CUSUM 5% Significance CUSUM of Squares 5% Significance Figure 3. CUSUM Plos Table 4. Long-run and Shor-run Elasiciies of GDP Growh Funcion Dependen Variable = log(gdp) Variables Shor Run Long Run Log(GCF).434*. 344* Log(E).5783*.743*** Log(EEWE) *** Noe: ***, **, * denoe he significance level a %, 5% and % level of significance, respecively. he GDP up o.4 percen in he shor run and.34 percen in he long run. Similarly, if here is one percen increase in employmen rae, i increases he GDP by.5 percen in he shor run and.74 percen in he long run. Educaion expendiure is saisically insignifican in he shor run bu in he long run is impac is quie significan as i increases economic growh up o.39 percen in he long run. The resuls have shown ha invesmen maer for economic growh boh in shor run and long run. Thus, macroeconomic policies aimed a increasing invesmen hrough foreign direc invesmen (FDI) and domesic sources. Securiy and poliical sabiliy also plays a par in aracing FID and reaining human capial. Educaion secor should be reaed as a special secor by immunizing budgeary allocaions for i from fiscal sresses and poliical and economic insabiliies. Allocaions for educaion should no be affeced by squeezed fiscal space or surge in miliary expendiure or debs. REFERENCES ADB (9). Asian Developmen Oulook - 9, Asian Developmen Bank, Manila.

8 846 Educ. Res. Afzal M, Farooq MS, Ahmad HK, Begam I, Quddus MA (). Relaionship beween School Educaion and Economic Growh in Pakisan: ARDL Bounds Tesing Approach o Coinegraion. Pakisan Economic and Social Review, 48 () (Summer ), Akram M, Khan FJ (7). Public Provision of Educaion and Governmen Spending in Pakisan. PIDE Working Papers 7:4. Online available a: hp:// Angelopoulos, Konsaninos, Jim Malley, Aposolis Philippopoulos (7). Public Educaion Expendiure, Growh and Welfare. Cesiifo Working Paper, No.37. hp:// DB_cesifowp/Q_CESIFO_WORKING_PAPERS_7/6_CESIFO_ WORKING_PAPERS_JUNE_7/cesifo_wp 37.pdf. Bardsen G (989). Esimaion of long-run coefficiens in error correcion models. Oxford Bullein of Economics and Saisics, 5, Barro R (99). Economic Growh in Cross-Secion of Counries. Quarerly J. Econ. 6 (), Benhabib J, Speigel M (994). The Role of Human Capial in Economic Developmen: Evidence from Aggregae Cross-Counry Daa. J. Moneary Econ. 34 (), Blankenau, William F, Nicole B, Simpson (4). Public Educaion Expendiures and Growh. J. Dev. Econ. 73(), Bose N, Haque EM, Osborn RD (3). Public Expendiure and Economic Growh: A Disaggregaed Analysis for Developing Counries. Online available a: hp:// Devarajan S, Swaroop V, Zou H (996). The Composiion of Public Expendiure and Economic Growh. J. Moneary Econ. 37: Faridi ZM, Hussain S, Bashir F (). Impac of Educaion on Sudens Earnings: A Case of Public Secor Universiies in Pakisan. In. Res. J. Finance and Econ. 4, GoP (). Governmen of Pakisan, Economic Survey 9-, Economic Advisor's Wing, Minisry of Finance, Islamabad (). GoP (). Governmen of Pakisan, Economic Survey -, Economic Advisor's Wing, Minisry of Finance, Islamabad (). Nurudeen A, Usman N (). Governmen Expendiure And Economic Growh In Nigeria, 97-8: A Disaggregaed Analysis. Bus. and Econ. J. Volume : BEJ-4, -. Pesaran MH, Y Shin, Smih R (). Bounds esing approaches o he analysis of level relaionships, J. Appl. Econome., 6, Teles KV, Andrade J (8). Public Invesmen in Basic Educaion and Economic Growh. J. Econ. Sud. 35 (4), 8, Todd Pera E, Kenneh I Wolpin (3). On he Specificaion and Esimaion of he Producion Funcion for Cogniive Achievemen. Econ. J. 3(485):F3-F33. Zaman K (8). An Invesigaion for Pro-Poorness of Governmen Educaional Policy in Pakisan (99-7). nd Inernaional Conference on Assessing Qualiy in Higher Educaion, s 3rd December, 8, Lahore Pakisan.