Impact of Stock Markets on Economic Growth: A Cross Country Analysis

Transcription

1 Impac of Sock Markes on Economc Growh: A Cross Counry Analyss By Muhammad Jaml Imporance of sock markes for poolng fnancal resources ncreased snce he las wo decades. Presen sudy analyzed mpac of sock markes on he level of GDP usng daa of egheen counres from 980 o 997. Though n analyss fxed wo way fxed-effecs are preferred over one way fxed-effecs and common effec models bu resuls do no conrbue sgnfcanly o he hypohess ha sock markes play mporan role o enhance he level of GDP. I. INTRODUCTION Fnancal markes, especally sock markes, have grown consderably n developed and developng counres over he las wo decades. Several facors have aded n her growh, mporanly mproved macroeconomc fundamenals, such as more moneary sably and hgher economc growh. Fnancal globalzaon has also advanced n he las wo decades wh ncreased cross-border capal flows, gher lnks beween fnancal markes and greaer commercal presence of foregn fnancal frms n counres around he world. Levne (003) founds ha fnancal markes end o develop as ncome per capa grows, fnancal reform progresses and nsuonal envronmen mproves. Smlarly Shees (996) found ha resdens decde o nves her wealh abroad due o an adverse domesc nvesmen clmae, ncludng macroeconomc nsably and weak nsuons. Accordng o Prasad e al. (003) fnancal markes affec economc growh hrough drec and ndrec effecs. Drecly sock markes enhance economc growh by augmenaon of domesc savngs, lowerng cos of capal due o beer rsk allocaon, ransferrng of echnology and hrough developmen of fnancal secors. Indrecly sock markes promoe economc growh hrough he promoon of specalzaon, nducemen for beer polces and hrough enhancemen of capal nflows by sgnalng beer polces. The grown conrbuons of sock markes o he level of gross domesc produc (GDP) n all over he world ask for crcal aenon. Presen sudy ams o check he effec of sock markes on he level of GDP usng he panel daa. Followng he secon of nroducon secon II dscusses dfferen mehodologcal ssues along wh her benefs and selecon of bes. In secon III we dscuss daa and n The auhor s suden of Phd n Deparmen of Economcs, Unversy of Venna, Ausra. The paper s wren for he requremen of course Economercs of Panel Daa augh by Prof. Kuns n Deparmen of Economcs Unversy of Venna, Ausra. Please see he sudy map n Appendx A.

2 secon IV we wll presen resuls of analyss and her nerpreaon. A he end we wll presen concluson of he sudy. II. Mehodologcal Issues The sudy uses cross secon daa of egheen counres. 3 Techncal help regardng panel daa esmaon s aken from Aserou (006). Gross domesc produc (GDP) s used as he dependen varable whle Marke capalzaon rao (MCR), Tradng value rao (TVR), Turn Over Rao (TOR), Local Index (LI), Toal reurn ndex (TRI) and Inernaonal Fnancal Cooperaon calculaed marke capalzaon (IFC_MC) are used as explanaory varables represenng effcency of he sock markes. Panel daa esmaons are consdered o be he mos recen and effcen analycal mehod n handlng economerc daa. In hs secon frs of all we dscuss benefs of usng panel daa. Afer ha we wll dscuss dfferen models ha can be used for analyss. Then a he end we wll dscuss selecon of models based on dfferen ess. II. Benefs of Panel daa Panel daa analyss has become popular among socal scenss because allows he ncluson of daa for N cross-secons and T me perods. A common problem of meseres esmaons s ha whle esmang samples wh very few observaons, s dffcul for he analys o oban sgnfcan -raos or F-sascs from regressons. Poolng daa no a panel of me seres from dfferen cross-seconal uns solve hs problem. Three major benefs of usng panel daa can be saed as: Frs major advanage of usng panel daa s ha provdes more effcen esmaons of parameers by consderng broader sources of varaon. Second ousources more nformaon o he analys. Thrd hey allow he sudy of he dynamc behavor of he parameers. II. Analycal models based on panel daa: Poolng of daa generaes dfferences among he dfferen cross-seconal or me-seres observaons ha can acually be capured wh he ncluson of dummy varables. In general, smple lnear panel daa models can be esmaed usng hree dfferen mehods: II.. Common Consan /Pooled OLS Mehod: The common consan mehod (pooled OLS) of esmaon presens resuls under he prncple assumpon ha here are no dfferences among he daa marces of he crossseconal dmenson (N). Model can be saed as = α MRC TVR 3TOR 4LI 5TRI 6IFC _ MC +υ 3 Please see he ls of counres n Appendx B.

3 Model esmaes a common consan α for all cross-secons. The common consan mehod mples ha here are no dfferences beween he esmaed cross-secon and s useful under he hypohess ha he daa se s a pror homogeneous. However hs case s que resrcve and cases of more neres nvolve he ncluson of fxed and random effecs n he mehod of esmaon. II.. Fxed Effecs Mehod: Fxed effecs allow dfferen consans for each group. The fxed effec esmaor s also known as he leas-squares dummy varables (LSDV) esmaor because n order o allow for dfferen consans for each group, ncludes a dummy varable for each group. Fxed effecs can be reaed n wo ways. Frs, one-way-fxed-effecs: n hs dummy varables are nroduced wh respec o cross secons or wh respec o me seres. If dummy varables are nroduced wh respec o cross secon un hen can be saed as = α MRC TVR TOR LI TRI IFC _ MC + υ N = α D MRC TVR 3TOR 4LI 5TRI 6IFC _ MC + υ = Where α denoe consan for each cross secon un. Oher way o ncorporae one-wayfxed effecs s o nroduce dummy varables wh respec o me. In hs case model can be saed as: = α MRC TVR 3TOR 4LI 5TRI 6IFC _ MC +υ T = α D MRC TVR 3TOR 4LI 5TRI 6IFC _ MC + υ = Where α represen he me specfc consans. Second way o ncorporae fxed-effecs s wo-way-fxed-effecs model: n hese dummy varables for me effecs as well as for cross-secon effecs are nroduced a he same me. N = α + α D + α D T = = IFC _ MC 6 + υ MRC TVR TOR 3 LI 4 TRI 5 Where α and α represens cross-secon specfc and me specfc consans. Problems wh fxed effec models are ha gnores all explanaory varables ha don vary over me. I means ha do no allow us o use oher dummy varables n he model. Fxed effec models are very neffcen because esmaes a very large number of parameers. I uses up o /T of he degrees of freedom avalable, whch s a major cos. Fxed effecs also make very hard for any slowly changng explanaory varables o be ncluded n he model, because hey wll be hghly collnear wh he effecs.

4 Thus, we should no use fxed effecs whou hnkng very carefully abou her valdy usng approprae ess. Due o he sgnfcan dsadvanages of he fxed effecs menoned above some prefers o use random effecs mehod. II.. Random Effecs Mehod: The dfference beween fxed effecs and random effecs model models s ha he laer handles he consans for each secon no as fxed, bu as random parameers. Smlar o fxed-effecs model random effecs can also be nroduced wo ways. Frs, one-wayrandom-effecs model: n hs random effecs are nroduced eher n cross-secons or n me seres. Once way random effecs can be saed as follow: = α MRC TVR TOR LI TRI IFC _ MC + ε + υ = α MRC TVR TOR LI TRI IFC _ MC + ε + υ Where ε and ε are he random effecs nroduced n cross-secon and n me seres respecvely. Second way o nroduce random effecs s wo-way-random-effecs model ha s nroducng random effecs wh respec o me and cross secon a a me. Twoway random effecs model can be saed as: = α MRC TVR 3TOR 4LI 5TRI 6IFC _ MC + ε + ε +υ Random effecs models have advanage n erms of fewer parameers o esmae compared o fxed effecs models. I also allows for he use of addonal dummy varables as he explanaory varables. II. Selecon of Model: One mgh expec ha he use of he random effecs esmaor s superor compared o he fxed effecs esmaor, because he former s he GLS esmaor and he laer s acually a lmed case of he random effecs model, as corresponds o cases where he varaon n ndvdual effecs s relavely large. Sascal es has been developed o es he superory among fxed effecs and common consan mehod. Null hypohess s ha all he consans are he same (homogeney) and ha herefore he common consan s applcable. H = α = α =... = α N 0 or 0 T H = α = α =... = α The F-sascs under he null hypohess can be saed as F ( R R ) /( T ) FE CC = ~ F N, NT N k ( RFE ) /( NT N k)

5 Where R FE and R CC are R obaned hrough fxed-effec and common coeffcen models respecvely. Smlar we can es effecveness of me effecs agans crosssecon effecs. The F-Sascs under he null hypohess can be saes as: F ( R R ) /( T ) TFE FE = ~ F T,( N )( T ) k ( RTFE ) /(( N )( T ) k ) Followng a deal dscusson of mehodologcal ssues n he followng secon we wll dscuss daa. 4 III Daa: The sudy uses annual daa from 980 o 997 for 8 counres. 5 Varables ha are used for analyss are Marke capalzaon rao (MCR), Tradng value rao (TVR), Turn Over Rao (TOR), Local Index (LI), Toal reurn ndex (TRI), Inernaonal Fnancal Cooperaon calculaed marke capalzaon (IFC_MC) and Gross Domesc Produc (). The sudy used he rch daabase, emergng markes daabase (EMDB) by Inernaonal fnancal cooperaon (IFC). Some of he mssng values for he gross domesc produc are aken from Inernaonal Fnancal sascs daa CD 005. Some of he daa pons for dfferen varables of dfferen counres are mssng from 980 o 987 whle from 980 o 997 daa for each varable for every counry s complee. The sudy analyzed daa n wo seps: frs, from 980 o 997 wh unbalanced panel and second from 988 o 997 for balanced panel. 6 III Resuls and Resuls dscusson: The fxed effec models and random effec models are appled on he daa wh he help of Evews. Resuls wh he sgn of he coeffcens and her sgnfcance are presened n able. Resuls ndcae ha value of R and adj-r are very low as compared o fxed effecs and random effecs models. R and Adj-R reman hgher for wo-way fxed effec models ha are above 80% for unbalanced panel whle near 90% for balanced panel. Lowes R and adj-r are observed when random effecs are nroduced n me seres. R and adj-r for wo-way random effecs wh balanced daa are also very close o he lowes R and adj-r. Smple f one wans o decde abou he model on he bass of R and adj-r hen wo way fxed effecs wll be good choce. 4 All he analycal work s done by he help of Evews and Evews do no provde he facly of Hauseman es whle manually s very dffcul o calculae nverse of 8 by 8 varance-covarance marx. So n he presen sudy Hauseman es was no appled. 5 Daa on he varables beyond 997 s publshed by publshed by Sandard s & Poor s wh he name of Global Sock Markes Facbook. 6 If panel consss of full nes of daa boh across counres and across me hen daa se s called balanced whle when observaons are mssng for he me perods of some of he cross-seconal uns hen he panel daa s called unbalanced.

6 Accordng o heory all he varables represenng effcency of sock markes should have posve mpac on he level of GDP. Resuls n erms of sgn and sgnfcance of he varables are very poor. Marke capalzaon rao showed negave sgn when analyzed hrough fxed-effecs of me and wo way fxed effecs whle posve for oher cases. Marke capalzaon rao s only sgnfcan a 0% level of sgnfcance when s analyzed fxed effecs are nroduced n cross-secons. Table : Effec of Sock markes on he level of GDP Resul-Summary R Adj-R MCR TVR TOR L TRI IFC_MC Common Unbalanced (+) (-)* (+)* (+) (+) (-) Coeffcen Balanced (+) (-)* (+)* (+) (+) (-)** Cross Unbalanced (+)*** (-)* (+)* (-) (+)** (+)* Secon Balanced (+) (-)** (+) (+)* (+)* (+)* Fxed Tme Unbalanced (-) (-)* (+)* (-) (-) (-)* Effec Seres Balanced (-) (-)* (+)* (+) (-) (-)* Two- Unbalanced (-) (-) (+)* (-)*** (-) (+) Way Balanced (-) (-) (+)** (+) (-) (-) Cross Unbalanced (+) (-) (+) (-) (+) (+) Secon Balanced (+) (-)** (+) (+)** (+)** (+)*** Random Effec Tme Seres Unbalanced (+) (-)* (+)* (+) (+) (-) Two- Way Balanced (+) (-)** (+) (+)** (+)** (+)** * Sgnfcan a %, ** Sgnfcan a 5%, *** Sgnfcan a 0% Tradng value rao (TVR) always appear wh negave sgn and mosly sgnfcan excep for wo way fxed effecs and one way random effecs n cross secon wh unbalanced daa. On he oher hand urn over rao (TOR) always appear wh posve sgn and sgnfcan mos of he me. Turn over rao s he only varable n our all ype of analyss ha has sgn accordng o he heory. Local ndex (L) showed mxed resuls ha are some me appears wh posve sgn and some mes appear wh negave sgns and found nsgnfcan mos of he me. Toal reurn ndex (TRI) appears o be showed smlar sgn as he marke capalzaon rao ha s negave when marke capalzaon rao s negave and posve when marke capalzaon rao s posve. Agan oal reurn ndex s found o be nsgnfcan mos of he me. IFC calculaed marke canalzaon (IFC_MC) also showed mxed resuls ha are some me posve and some me negave bu sgnfcan mos of he me. One way fxed-effecs, one way random effecs and wo way fxed effecs and wo way random effecs are gven Appendx C. One way fxed effecs wh cross secon uns ndcae Korea domnaes n case of unbalanced panel whle Brazl domnaes n case of balanced panel. Mos of he counres appear wh negave fxed effecs. Fxed effecs wh me seres ndcae negave coeffcens for frs half of he perods whle posve for he remanng weaher appled on balanced or unbalanced pool. Two way fxed effec models ndcae Brazl domnaes n cross secon followed by Mexco and Inda. Smlar o one way fxed-effec model mos of he counres has negave coeffcen of fxed effec and coeffcens of me effecs are negave n he begnnng perod whle posve a he endng perods.

7 When random effecs are nroduced n he cross secons Brazl domnaes n case of unbalanced analyss whle Chna domnaes n case of balanced panels. Whle very less number of coeffcens are negave as compared o one-way fxed effec models. On he oher hand when random effecs are appled wh respec o me hen all he coeffcens are observed o be zero. Resuls of wo-way-random effecs ndcae Brazl domnaes followed by Mexco and Inda. Smlar o one way random effec-models coeffcens for he me effecs are agan observed o be zero. Now o decde whch model s repressng a rue pcure of he analyss we used F- sascs menon n III.. Table represens he resuls of he F-Sascs. Calculaed values of F-sascs are greaer han crcal values n all cases ndcaed hose null hypoheses are rejeced n all cases. In he frs case rejecon of null hypohess ndcae all he counry effecs are no homogenous. Smlarly rejecon of null hypohess a he second case ndcaes me effecs are no homogenous and common coeffcen wll no able o represen rue pcure. Rejecon of null hypohess a he hrd sage ndcae ha n he presence of cross secon fxed effecs me effecs also play mporan role n he mprovemen of resuls. Table : Selecon of model based on F-Sascs F Crcal Calculaed Decson Unbalanced Rejec H 0 n Cross Secon Panel favor of H Balanced Rejec H 0 n Fxed Panel favor of H Effec Unbalanced Rejec H 0 n Tme Seres Panel favor of H Balanced Rejec H 0 n Panel favor of H Unbalanced Rejec H 0 n Effecveness of Tme Effecs Panel favor of H Balanced Panel Rejec H 0 n favor of H Though F-sascs ndcae ha wo way fxed effecs are beer han one way fxed effec bu regresson resuls do no conrbue sgnfcanly o he hypohess ha sock markes play mporan role o enhance he level of GDP. IV Concluson: The presen sudy analyzes he mpac of sock markes on GDP usng dfferen varables represenng effcency of he sock markes. The sudy uses panel daa for egheen counres from 980 o 997. Sudy uses common effec, fxed effecs and random effecs o verfy he hypohess. Turn over rao s he only varable ha shows sgn accordng o

8 heory n all cases. Tradng value rao always comes up wh conradcory sgn as compared o heory. Oher varables represen a mxed resuls means some me negave and some me posve sgn and smlarly some cases sgnfcan and n some cases nsgnfcan. F-sascs ndcae wo-way-fxed effecs do are preferred over one wayfxed effec and common effecs models. However resuls do no conrbue sgnfcanly o he hypohess ha sock markes play mporan role o enhance he level of GDP One possble reason mgh be along wh he effcency measures of sock markes here s need o nroduce some macroeconomc varables n he analyss as used by Mohad, Hamd and Agarwal Sum (---). References Aserou, Dmros (006) Appled Economercs: A Modern Approach usng Evews and Mcrof s edon Palgrave: Macmllan Levne, R. (003) Law, fnance and Growh: Theory, Evdence and Mechansm prepared for he Handbook of Economc Growh Mohad, Hamd and Agarwal Sum (---), Sock Marke Developmen and Economc Growh: Evdence from Developng Counres Inerne source Prasad Eswar, Kenneh Rogoff, Shang Jn We and M. Ayan Kose (003) Effecs of Fnancal Globalzaon on Developng Counres: Some Emprcal Evdence Inernaonal Moneary Fund Shees, N. (996) Capal flgh from he counres n Transon: Some Theory and Emprcal Evdence Federal Reserve Board, nernaonal Fnance dscusson Paper No. 54 Inernaonal Fnancal Cooperaon, Emergng Sock Markes Fac Book Issues 998, 979 Inernaonal Fnancal Sascs Daa CD 005

9 Appendx A: Framework of he paper Effec of Sock Markes on GDP Inroducon Mehodologcal Issues Resuls Benefs of Panel Daa Analycal Models based on Panel Daa Selecon of Model Unbalanced Panel (980-97) Common Effecs Balanced Panel (988-97) Fxed Effecs Unbalanced Panel (980-97) One Way Balanced Panel (988-97) Two Way Unbalanced Panel (980-97) Random Effecs Balanced Panel (988-97) Unbalanced Panel (980-97) Concluson References

10 Appendx B: Ls of Counres Argenna Brazl Chle Columba Greece Inda Jordan Korea Malaysa Mexco Ngera Paksan Phlppnes Porugal Tawan (Chna) Thaland Venezuela Zmbabwe Appendx C : One way fxed effec (Cross-secon) Counry Unbalanced Balanced Counry Unbalanced Balanced Argenna Mexco Brazl Ngera Chle Paksan Columba Phlppnes Greece Porugal Inda Tawan (Chna) Jordan Thaland Korea Venezuela Malaysa Zmbabwe Appendx C : One way fxed effec (Tme-Seres) ear Unbalanced Balanced ear Unbalanced Balanced

11 Appendx C 3 : Two-Way-Fxed-Effecs Counry Unbalanced Balanced Counry Unbalanced Balanced Argenna Brazl Chle Columba Greece Inda Jordan Korea Malaysa Mexco Ngera Paksan Phlppnes Porugal Tawan (Chna) Thaland Venezuela Zmbabwe Appendx C 4 : One way fxed effec (Cross-secon) Counry Unbalanced Balanced Counry Unbalanced Balanced Argenna Mexco Brazl Ngera Chle Paksan Columba Phlppnes Greece Porugal Inda Tawan (Chna) Jordan Thaland Korea Venezuela Malaysa Zmbabwe

12 Appendx C 5 : Two-Way-Random-Effecs Counry Unbalanced Balanced Counry Unbalanced Balanced Argenna Brazl Chle Columba Greece Inda Jordan Korea Malaysa Mexco Ngera Paksan Phlppnes Porugal Tawan (Chna) Thaland Venezuela Zmbabwe