Budget Deficit, Economic Growth and FDI in the Baltics: A Cross-Section Dependence Panel Approach.

Transcription

1 Ques Journals Journal of Research n Humanes and Socal Scence Volume 4 ~ Issue 12 (2016) pp: ISS(Onlne) : Research Paper Budge Defc, Economc Growh and FDI n he Balcs: A Cross-Secon Dependence Panel Approach. 1 kolaos Drsaks, 2 Pavlos Samaou Deparmen of Appled Informacs, Unversy of Macedona, Economcs and Socal Scences, 156 Egnaa Sree, Thessalonk, Greece e-mal: drs@uom.gr Deparmen of Appled Informacs, Unversy of Macedona, Economcs and Socal Scences, 156 Egnaa Sree, Thessalonk, Greece e-mal: samaou@uom.edu.gr Receved 14 Dec, 2016; Acceped 26 Dec, 2016 The auhor(s) Publshed wh open access a ABSTRACT: Ths sudy nvesgaes he mporance of budge defc and foregn drec nvesmens on economc growh of he Balc counres. In our analyss, we use a panel daa se for all hree counres over he perod Ths paper, herefore, apples panel un roo ess (frs and second generaon), panel conegraon mehods and panel causaly es hrough he Vecor Error Correcon Model. The emprcal fndngs reveal a posve and sgnfcan long-run relaonshp beween foregn drec nvesmen and economc growh n Balc counres. In conras, he relaonshp beween budge defc and economc growh s found o be negave. Also, he causaly resuls show, boh n shor and long-run, undreconal causal relaonshp from foregn drec nvesmens o economc growh as well as from budge defc o economc growh. Resuls on hs paper ndcae ha he ranson counres, whch mplemen he prvazaon programs successfully, arac foregn drec nvesmen faser whch n urn promoes economc growh. Keywords: Balc Counres, Budge Defc, Economc Growh, FDI, Panel Daa I. ITRODUCTIO The global fnancal crss of 2008 s consdered by many economss as he wors fnancal crss snce he Grea Depresson of he 1930s. The recen crss sared n he US (Uned Saes) morgage marke and wen global when morgage-backed secures raded by fnancal nsuons. Alhough no all subprme morgages were bad loans, he crss was spread quckly because nvesors los her fah n hs caegory of nvesmens. Τhe nerdependence of he fnancal markes had as a resul he ransmsson of crss n Europe. Today, he fnancal condons have been mproved n mos European counres. These mprovemens are due o he massve suppor measures aken by he governmens and he European Cenral Bank. On he oher, even oday many years afer oubreak of he crss, here are sll counres n Europe ha are facng serous srucural and fnancal problems. The Balc Saes were arguably some of he European counres whch affeced sgnfcan by he global fnancal crss. The collapse n demand and he shocks n he fnancal markes led o oupu reducon n hese economes. In 2009, he declne of GDP n Esona, Lava and Lhuana was 14.1%, 17.9% and 14.7% respecvely. Moreover, he reducon n domesc demand and expors had as a resul a sgnfcan ncrease on unemploymen. In all hree counres, he unemploymen rae was almos hree mes hgher a he end of 2009 han was n 2007 (WDI, 2016). In response o crss, he Balcs chose he nernal devaluaon of her domesc currency, whch means downward adjusmen n nomnal wages. The governmens of hese counres were srongly opposed o he exernal devaluaon, belevng ha hs sraegy would be an obsacle on her quck accesson o he Eurozone. As far as he fscal polcy, he response o he crss was dfferen across he Balcs. Unl 2007 he budge n Esona was balanced, whle n he oher wo counres was n defc. Afer he oubreak of he crss, Lava s and Lhuana s budge defcs reached almos 10% of her GDP. On he oher, Esona managed o keep s defc almos a 2% of GDP, n Ths drec fscal adjusmen, compared wh he gradual approaches of Lava s and Lhuana s, was due o Esona s effor o jon Eurozone as soon as possble. In Deparmen of Appled Informacs, Unversy of Macedona, Economcs and Socal Scences, 156 Egnaa Sree, Thessalonk, Greece 65 Page

2 Budge Defc, Economc Growh and FDI n he Balcs: A Cross-Secon Dependence Panel early 2010, he Balc counres seem o have lef behnd he dffcules of he crss. Ther economes have reurned o growh, wh reduced unemploymen raes and ncreased quany of expors. The relaonshp beween budge defc and economc growh has been he focus of a consderable number of academc sudes. In he leraure here are hree perspecves as far as he relaonshp beween hese varables: Rcardan heory suppors ha here s neural relaon beween budge defc and economc growh. Keynesan, whch says ha here s posve relaon beween budge defc and economc growh. Fnally, eoclasscal heory suppors ha exss an nverse relaonshp beween he wo varables. Foregn drec nvesmens (FDI) had a val role n he world economy he las weny years. Foregn drec nvesmens are an mporan source of prvae capal snce hey are dreced owards producve acves, and herefore conrbue sgnfcanly o economc developmen (Gorbunova e al., 2012). As far as he Balc counres, FDI have played a key facor n her ranson process, from planned o marke economes. Durng hs perod, hese counres have receved a lo of FDI. In Lava and Lhuana he average level has been over han 3% of GDP. However, n Esona hs number has been much larger (8% of GDP) (UCTAD, 2016). The polcal deermnaon of Esona o mplemen fas economc lberaon was he man reason for hs large share of FDI. The am of hs paper s o examne he relaonshp beween budge defc, FDI nflows and economc growh n he hree Balc counres, vewed as group, usng annual daa for he perod Alhough here s a large number of academc sudes ha examne he lnk beween hese varables, lle academc research has been done on he Balcs. Perhaps due o he small sze of Balc economes, whch also may reflecs he weakness of he domesc economc research (Suela, 2001). Moreover, he Balc economes were closed and negraed pars of Sove sysem unl 1990, so no FDI flows were before. The srucure of he paper s as follows: Secon 2 brefly revews he heorecal framework. Secon 3 presens he emprcal leraure. Secon 4 descrbes daa and mehodology. Emprcal resuls are dscussed n secon 5. Concludng remarks are gven n he fnal secon. II. THEORETICAL FRAMEWORK The recen economc crss had undoubedly a negave mpac on he economes of all EU counres. However, he magnude of he mpac was no he same for all members. Ths depended on he nfrasrucure and he level of he domesc economy of each counry. For hs reason, he governmens of hese counres have mplemened dfferen ypes of fscal measures n order o manan economc sably. These measures had a sgnfcan effec n macroeconomc varables such as GDP, governmen deb, budge defc and FDI. Budge defc s he amoun by whch governmen expendures exceed s revenues over a parcular perod of me, usually a year. Accordng o European Sysem of negraed economc Accouns (ESA) here s a dsncon beween he cenral governmen defc and he general governmen defc, dependng on he governmen s secors ha are ncluded. In addon, here s a dsncon beween prmary defc and oal defc. The prmary defc does no nclude he neress on deb servce. All counres can have defcs n her budges. However, s necessary o keep her defcs below ceran quanave lms. Theorecally, hese lms are defned from he long erm deb susanably. In pracce, he lms are defned from he possbly of he defc fnancng by he bankng sysem. The governmen defc s funded by loans provded by he banks and oher nsuonal nvesors and also by he bonds ssued by he naonal governmens. In general, here are hree schools of economc hough regardng he mpac of budge defcs: eoclasscal: Accordng o hs heory, he fscal defcs ncrease he overall consumpon and shf he axes o fuure generaons. In full employmen of he economy s resources, an ncrease n consumpon wll reduce savngs. Subsequenly, he neres raes should be ncreased n order o brng balance n he capal markes. In hs way, fscal defcs crowd ou he prvae capal accumulaon. Buer (1997) suppored ha, when he economy s a full capacy and employmen of resources, any ncrease n publc expendures should necessarly lead o a reducon of he same amoun of prvae spendng (crowdng-ou) and creae negave effecs on economc growh. Keynesan: Accordng o Keynesan heory, fscal defc s used as a means for mprovng he economc suaon and maxmzng he socal welfare. Ths approach suppors ha here s a negave relaon beween budge defc and unemploymen as well as a posve relaon beween budge defc and real growh. In 1989, Berhem argued ha appropraely med defcs are benefcal for he economy. Rcardan Equvalence: Rcardo suppors ha he budge defc ncreases due o an ncrease n governmen spendngs. Ths defc can be pad now or a a laer sage. Therefore ax cus, arsng from he budge defc polcy, have no effec on consumpon and savngs. So, employmen does no change as well as oher macroeconomc varables ncludng economc growh. Saleh (2003), suppored ha budge defcs have no sgnfcan mpac on nvesmens, savngs and overall on economc performance. 66 Page

3 Budge Defc, Economc Growh and FDI n he Balcs: A Cross-Secon Dependence Panel A foregn drec nvesmen s an nvesmen made by a company or eny based n one counry, no a company or eny based n anoher counry. The effec of FDI on a domesc economy has been exensvely suded durng he las years. Mos of he sudes show posve effecs of FDI n ransonal and long erm economc growh hrough capal accumulaon and ransfer of knowledge (Basu e al., 2003). Densa (2010) beleved ha FDI promoes economc growh boh drecly and ndrecly hrough expors. However, he posve effec of FDI on a hos counry appears o be greaer n developed economes, snce hey can absorb more easly new echnologes (Blomsrom e al., 1992). On he oher, some sudes have shown ha hese posve resuls may be nsgnfcan or even negave due o he concenraon of domesc capal (Carkovc and Levne, 2005). Reuber e al. (1973) dspued he posve effecs of FDI because of he shor-erm unemploymen ofen causes and he hgh adjusmen cos of domesc companes n order cope he new compeve condons. III. REVIEW OF EMPIRICAL LITERATURE In prevous years, sudes on he relaonshp beween budge defc, FDI and economc growh, usng eher me seres or panel daa, revealed dfferen resuls whch seem o depend on he counres (developed or developng), he economercs mehods and he perod he analyss was carred ou. Ball and Mankw (1995) proved ha here s an nverse relaonshp beween budge defc and economc growh for he case of he US, usng daa over he perod I was found ha economes wh budge defcs have lower growh raes han economes wh budge surplus. The connuous ncrease n budge defcs can lead o a bankrupcy problem. Asogwa and Okeke (2013) suppored ha crowdng ou effec of budge defc on prvae nvesmen n he gera s economy has a sgnfcan mpac on he economy s oupu, he level of employmen and he sandard of lvng. The geran governmen should reduce recurren expendure and ncrease s capal expendure n order o encourage prvae nvesmens and promoe economc growh. Hussan and Ahmad (2015) nvesgaed causal lnk beween FDI, budge defc and economc growh for Paksan usng daa over he perod of The analyss ndcaed ha here are bdreconal causaly relaons beween FDI and economc growh and beween economc growh and budge defc, as well as a undreconal causaly runnng from FDI o budge defc. Fndngs suppor ha FDI s an mporan facor for economc growh n Paksan. Also, fndngs ndcae ha here s a long run assocaon beween economc growh and budge defc. Ths means ha governmen s nvesng more for he welfare of he naon. However, hs spendng mus be a he rgh rack. As far as he sudes whch conduced under he panel framework, Bose e al. (2007) found ha budge defc can help economy o grow faser. For her analyss hey used a panel daa se of 30 developng counres from 1970 o They suppored ha he defcs were due o expendures n felds such as educaon and healh. Afonso and Jalles (2011) nvesgaed he mpac of budge defc on producvy and economc growh usng a large panel of 155 counres over he perod Ther resuls revealed ha he defcs have sgnfcan negave effecs on economc growh, whle he overall effcency of he producon facors has a posve and sgnfcan mpac. Meza (2012) examned he mpac of budge defcs and nvesmen spendng on curren accoun defcs usng panel daa for 20 OECD counres over he perod of Boh budge defcs and domesc nvesmen are found o Granger-cause he curren accoun. Larger budge defcs lead o hgher curren accoun defcs especally n he shor erm. Increases n nvesmen spendng have a smlar effec over me. Meza, concluded ha he resuls of hs sudy do no suppor he Rcardan Equvalence hypohess. Yucel (2014) nvesgaed he relaonshp beween FDI and economc growh for he Balc counres usng panel daa for he perod The resuls showed ha FDI has a posve effec on economc growh for he group of he counres. The aracon of foregn capal nflows wll ncrease FDI, wll boos economc developmen and wll help n reducng unemploymen. Bökemeer (2015) found ha here s a sgnfcan negave relaonshp beween budge defc and economc growh for he group of egh EU counres, ha joned he unon n The sample of he sudy consss of annual panel daa durng he perod The deb or defc fnancng nvolves fuure paymens on neres and redempons, whch n urn reduce he scope of budge or he fscal polcy decsons. The lnk beween hese varables seems o be sronger afer Akbaş and Lebe (2016) nvesgaed he valdy of he rple defc hypohess n he G7 counres, usng panel daa over perod Ther resuls revealed ha budge defc and savngs gap have mporan role on curren accoun defc. In addon, b-dreconal causaly beween he curren accoun defc and he savngs gap and beween he budge defc and he savngs gap were deermned. They conclude ha rple defc hypohess s vald n G7. 67 Page

4 Budge Defc, Economc Growh and FDI n he Balcs: A Cross-Secon Dependence Panel IV. DATA AD METHODOLOGY 1. Daa The varables ha are used n hs sudy are gross domesc produc (GDP), foregn drec nvesmens nflows (FDI) and budge defc (BD) expressed n mllon US dollars. The daa are annual coverng he perod All Daa colleced from World Developmen Indcaor (WDI, 2016), publshed by he World Bank and from Uned aons Conference on Trade and Developmen (UCTAD, 2016). The curren prces of GDP, FDI nflows and BD were deflaed by he GDP deflaor of each counry (2005=1) n order o be convered n consan prces. 2. Mehodology The purpose of he paper s o examne he causal relaonshp beween foregn drec nvesmens, budge defc and economc growh n he Balc counres, vewed as group, over he perod Followng Hussan and Ahmad (2015) we specfy he funconal form as follows: GDP a FDI BD u (1) where: Defc, 1 2 GDP, Gross Domesc Produc, FDI, Foregn Drec Invesmens Inflows, BD, a = Inercep, 1 = Esmaed coeffcen of FDI, 2 = Esmaed coeffcen of BD, u, Budge Error erm, he number of ndvdual members and he number of observaon over me. Afer he specfcaon of he model, our paper nvolves four objecves:. The frs s o examne he saonary of he varables. Panel un roo ess are dvded no frs and second generaon. The frs generaon ess assume ha he uns of he panel are ndependen each oher. On he oher, he second generaon ake under consderaon he cross seconal dependency. Accordng o Breusch and Pagan (1980) and Pesaran (2004) he exsence of cross seconal dependence beween he seres can affec he whole panel sgnfcanly. In hs sudy, we begn applyng he frs generaon un roo ess. Consderng ha panel daa models are lkely o exhb subsanal cross seconal dependence n he errors, necessary o es for cross seconal dependence n order o avod based esmaes and spurous resuls. Snce cross seconal dependence exss, we connue applyng he.. v. second generaon ess. Havng defned he order of negraon, we connue applyng panel conegraon mehodology. The classcal mehod of Pedron (1999) under he assumpon of cross-seconal ndependence and he Weserlund (2007) es robus o cross seconal dependence are compued. The hrd s o esmae he long run relaonshp wh he panel Dynamc Ordnary Leas Square (DOLS) mehod. The fourh am s o esmae a dynamc panel vecor error correcon model (VECM) n order o provde us wh he Granger causal relaonshps. 2.1 Frs Generaon Un Roo Tess The frs sep n panel causaly analyss s o defne he order of negraon of he varables ncluded n he sudy. The leraure proposes several approaches for esng un roo n panels. Takng under consderaon ha hese mehods may gve dfferen resuls we selec ess suggesed by Breung (2000), Levn e al. (2002) (LLC), Im e al. (2003) W-es (IPS), ADF-Fsher Ch-square es, PP Fsher Ch-Square es, Maddala and Wu (1999), and Hadr (2000). In all cases excep Hadr, he null hypohess s ha he varable conans a un roo. All hese ess assume ha panel daa do no exhb a cross-seconal dependence n he errors. The frs generaon un roo ess can be dvded no homogeneous and heerogeneous models. Levn e al. (2002) and Breung (2000) ess, assume homogeney n he dynamcs of he auoregressve coeffcens for all ndvduals of he panel. Phllps and Sul (2003) proved ha hs hypohess n many cases can lead o he rejecon of he null hypohess ncorrecly. Also, Hadr (2010) suggesed a resdual based Lagrange mulpler es for he null ha he me seres are saonary, for each, around a deermnsc rend agans he alernave of a un roo n panel daa. However, he res ess are no so resrcve. Im e al. (2003) proposed an alernave snce allows for heerogeneous coeffcens n he panel. Moreover, Maddala and Wu (1999) proposed wo dfferen ess, based on he Augmened Dckey-Fuller and he Phllps-Perron ess, n order o check he saonary n panel daa. In hese ess, he null and alernave hypoheses are he same as n he IPS es. 2.2 Cross Seconal Dependence Tess As we menon before, he frs generaon un roo ess have he drawback o suppose ha he cross secons n he panel are ndependen. However, hs assumpon s no vald n he emprcal nvesgaon. Panel daa models are more lkely o exhb a cross-seconal dependence n he errors, whch may arse due o 68 Page

5 Budge Defc, Economc Growh and FDI n he Balcs: A Cross-Secon Dependence Panel unobserved common facor, he presence of common shocks, regonal and macroeconomc lnkages and general resdual nerdependence (De Hoyos and Sarafds, 2006) Promnen cross-seconal dependence ess are hese proposed by Breusch-Pagan (1980) and Pesaran (2004). Followng Pesaran (2004), consder he panel daa model: y a x u wh = 1,, and = 1, T (2) where a represens he nercep, x, s a k dmensonal column vecor of regressors, s a k x 1 vecor of parameers o be esmaed, u, s assumed o be ndependen and dencally dsrbued over me perods and across cross-seconal uns. Pesaran s cross dependence (CD) es s based on he correlaons beween he dsurbances n dfferen cross secon uns: CD 1 2T ( j ) ( 1) (3) 1 j1 whrere uu j 1 j j 2 2 ( u ) ( u j ) 1 1 1/2 1/2 The mos well-known cross secon dependence dagnosc s he Breusch-Pagan (1980) es. Breusch and Pagan (1980) proposed a Lagrange Mulpler (LM) sasc gven by: 1 2 LM T( ) (4) 1 j1 j where j are he correlaon coeffcens obaned from he resduals of he model as descrbed above. For boh ess, he null hypohess s ha here sn cross-seconal dependence n he panel. Under he null hypohess, he sasc has a ch-square asympoc dsrbuon wh (-1)/2 degrees of freedom. In he case ha me dmenson s larger han cross secon dmenson Breusch-Pagan (1980) es s proposed. On he oher hand, f me dmenson s smaller han cross secon dmenson Pesaran (2004) es s preferred. 2.3 Second Generaon Un Roo Tes Snce cross-seconal dependence among counres has been deermned, we connue applyng he second generaon un roo ess. The mos common of hese ess s he Cross Seconally Augmened IPS (CIPS) proposed by Pesaran (2007). The procedure of CIPS es begns wh he OLS esmaon for he h cross secon n he panel consderng he followng Cross-Seconal Augmened Dckey Fuller (CADF) regresson: y y y y y, 1, 1 j, j, j j0 j0 k k (5) 1 where: y, 1 ( ) y, 1, 1 y ( ) y, 1 and (, ) s he -sasc of ha s used for he 1 compuaon of he ndvdual ADF sasc. CIPS sasc s based on ndvdual CADF average sascs. Equaon of CIPS s as follows: CIPS CADF 1 ( ) Pesaran (2007) has abulaed he crcal values for CIPS for varous deermnsc erms. If calculaed CIPS value s smaller han he able he crcal value, he null hypohess of non saonary s rejeced. 2.4 Panel Conegraon Tess Snce he order of saonary has been defned, our nex sep s o use panel conegraon mehodology n order o examne he long run relaonshp beween he varables. The classcal mehod of Pedron (1999) under he assumpon of cross-seconal ndependence and he Weserlund (2007) es robus o cross-seconal dependence are appled. (6) 69 Page

6 Budge Defc, Economc Growh and FDI n he Balcs: A Cross-Secon Dependence Panel Pedron (1999) nroduces seven conegraon sascs ha allow for heerogeneous slopes coeffcens across cross secons. Of hese seven sascs, four are based on he whn-dmenson of he panel and hree on he beween-dmenson (see Appendx). The null hypohess of no conegraon s he same for each sasc H0 p 1 for all. For he beween-dmenson sascs he alernave s H1 p 1 for a leas one, whle for he whn-dmenson he alernave s H1 p p 1 for all. The second es proposed by Weserlund (2007). Ths es examnes he presence of conegraon by deermnng f he error correcon erm n a condonal error correcon model exss or no. Consder he followng error correcon model: Y a ( Y X ) Y X u p,, 1, 1, k, k, k, k, k1 k1 p (7) where Y, sands for he GDP,, gves a se of he exogenous varables, s he speed of adjusmen erm and u j,, s he dsurbance erm assumed o be uncorrelaed wh zero means. The null hypohess of no error correcon erm s acceped n he case ha 0. If 0, he model s error correcng whch means ha he varables are conegraed. Weserlund (2007) proposed wo ypes of ess. The frs ncludes he whole panel sascs, namely and P a, whle he second ncludes he group mean sascs namely G and G a. The null hypohess s he same for all sascs H : 0 0 for all. On he oher, for he whole panel sascs he alernave s H : 0 1 for all and for he group mean es sascs he alernave s H : 0 1, for a leas one. One propery of Weserlund (2007) s ha provdes p-values que robus o cross-seconal dependence hrough o he boosrap approach used by Chang (2004). 2.5 Panel DOLS Esmaes Snce our varables are conegraed he nex sep s he esmaon of he long-run equlbrum relaonshp. Accordng o Kao and Chang (2000) he ordnary leas square (OLS) esmaor s based and nconssen when appled o conegraed panels. Pedron (2000) argued ha only n he case ha he regressors are srcly exogenous he OLS esmaors are unbased and could be generally used for vald nferences. So, we use he dynamc OLS (DOLS) esmaor proposed by Kao and Chang (2000) n order o esmae he long-run conegraon vecor. The DOLS esmaor allow for greaer flexbly n he exsence of heerogeney of he conegrang vecors (Pedron 1999; 2000). Consder he followng fxed effecs panel regresson: y x u for = 1 members and = 1 T (8) where y s a marx (1,1), s a vecor of slopes ( k,1) dmenson, s he ndvdual fxed effec, u are he saonary dsurbance erms. x ( k,1) vecor assumed o be an negraed process of order one for all, where x x 1 e. The DOLS s an exenson of Sock and Wason (1993) esmaor and s consruced makng correcons for endogeney and seral correlaon o he OLS esmaor (Phllps, 1995). Followng equaon s used o oban he DOLS esmaor: jq y x c x v (9) j, j jq Where c j represens he lead or lag coeffcen of explanaory varables a frs dfferences. The DOLS esmaor s obaned as followng: 1 * zz z y (10) z [ x x, x,..., x ] s 2( q 1) 1 vecor of regressors. Kao and Chang (2000) DOLS where, q, q suppored ha he DOLS esmaor s less based and has superor small sample properes compared wh FMOLS esmaor. P 70 Page

7 Budge Defc, Economc Growh and FDI n he Balcs: A Cross-Secon Dependence Panel 2.6 Panel Causaly Analyss The exsence of conegraon beween he examned varables mples ha here s causaly relaon among hem n a leas one drecon (Engle and Granger, 1987). We connue wh he esmaon of he VECM n order o fnd he shor and long run dynamcs beween budge defc, FDI and economc growh. The equaons ha are used o es Granger causaly are he followng: p p p (11) GDP a GDP BD FDI ECT u, 1, 1,1,, k, k 1,2,, k, k 1,3,, k, k 1,, 1 1,, k1 k1 k1 p p p (12) BD a GDP BD FDI ECT u, 2, 2,1,, k, k 2,2,, k, k 2,3,, k, k 2,, 1 2,, k1 k1 k1 p p p (13) FDI a GDP BD FDI ECT u, 3, 3,1,, k, k 3,2,, k, k 3,3,, k, k 3,, 1 3,, k1 k1 k1 Where Δ s he frs dfference operaor, k = 1,,p s he opmal lag seleced by he Schwarz, ECT, 1 sands for he lagged error correcon erm from he long-run conegraon equaon, j, s he adjusmen coeffcen (j = 1, 2, 3, 4) and u j,, s he dsurbance erm assumed o be uncorrelaed wh zero means. V. EMPIRICAL RESULTS 1. Frs Generaon Un Roo Resuls We begn applyng he un roo ess of LLC, Breung, IPS, ADF-Fsher, PP-Fsher and Hadr. The resuls of level and frs dfference un roo ess for he hree varables are provded n Table 1. LLC Breung IPS ADF PP Hadr LLC Breung IPS Table 1: Panel Un Roo Resuls Level GDP FDI BD Inercep Inercep and Trend Inercep (0.154) (1.000) (0.076) (0.001)*** (0.002)*** (0.034)** Inercep and Trend (0.408) (0.028)** (0.289) Inercep (0.714) (0.320) (0.005)*** Inercep and Trend (0.176) (0.058) (0.007)*** Inercep (0.890) (0.374) (0.010)** Inercep and Trend (0.260) (0.064) (0.013)** Inercep (0.906) (0.173) (0.092) Inercep and Trend (0.945) (0.338) (0.301) Inercep (0.011)** (0.008)*** Inercep and Trend (0.052) (0.113) (0.345) Frs Dfferences GDP FDI BD Inercep Inercep and Trend (0.181) Inercep Inercep and Trend (0.650) Inercep Inercep and Trend (0.019)** (0.007)*** Inercep Page

8 Budge Defc, Economc Growh and FDI n he Balcs: A Cross-Secon Dependence Panel (0.002)*** ADF Inercep and Trend (0.030)** (0.014)** Inercep (0.027)** PP Inercep and Trend (0.401) Inercep (0.681) (0.502) (0.724) Hadr Inercep and Trend (0.237) (0.001)*** (0.113) oes: Panel daa nclude all counres. The numbers n parenheses denoe p-values. ***, ** denoes rejecon of null hypohess a he 1% and 5% level of sgnfcance, respecvely. The null hypohess s ha he panel seres has a un roo excep wh he Hadr es whch has no un roo n panel seres. Lag lengh selecon auomac based on Schwarz creron. As can be seen from Table 1, he resuls showed ha GDP conan a un roo (non-saonary) n levels, whle he resuls for he oher wo varables are nconclusve. In he cases of FDI and BD wo ou of sx ess and hree ou of sx ess respecvely fal o rejec he saonary hypohess, eher wh nercep or ncludng nercep and rend, n 5% level o sgnfcance Evdenly, he resuls ndcaed ha all varables are saonary n her frs dfferences (.e. I(1)). 2. Resuls of Cross-Seconal Dependence Tes All he above un roo ess have he drawback o suppose ha he cross secons are ndependen each oher. The fac ha he un roo ess resuls for he varables of FDI and BD are nconclusve could be arbued o he cross-seconal ndependence hypohess (Breusch and Pagan 1980; Pesaran 2004). In hs sudy, he cross-seconal dependence ess proposed by Breusch and Pagan (1980) s appled, snce here are 3 counres (=3) and 18 years (T=18). Table 2: Cross-Seconal Dependence Resuls Tes Sasc Probably d.f Breusch-Pagan (LM) 9.139** oe: ** ndcaes rejecon of null hypohess a 5% level of sgnfcance. As can be seen from Table 2, he Breusch-Pagan es srongly rejecs he null hypohess of no crossseconal dependence a 5% level of sgnfcance. Ths means ha a shock whch has come from one of he counres, affecs he ohers. 3. Second Generaon Un Roo Resuls Snce cross-seconal dependence has been deermned among counres, we proceed applyng he CIPS es proposed by Pesaran (2007). We run he es for each varable up o 4 lags wh and whou rend. Table 3: CIPS Resuls Level GDP FDI BD CIPS Sasc Inercep Inercep and rend Frs Dfferences GDP FDI BD CIPS Sasc Inercep *** *** *** Inercep and rend ** *** *** oes: *** and ** denoes rejecon of null hypohess a a 1% and %5 levels of sgnfcance. The null hypohess of non saonary s rejeced n he case ha CIPS sascs s smaller han crcal value. The crcal values n he case of nercep are and a 1% and 5% levels of sgnfcance respecvely. In he case of nercep and rend are and a 1% and 5% level respecvely. The resuls of Table 3 suppor ha all varables are no saonary n levels bu n her frs dfferences. Consderng he frs generaon un roo ess resuls, he cross-seconal dependence resuls and he resuls of second generaon un roo es we conclude ha GDP, FDI and BD are negraed of order one (.e. I(1)). 4. Panel Conegraon Resuls Snce he order of negraon has been confrmed, we proceed applyng panel conegraon mehodologes o es wheher here s long-run relaonshp beween he examned varables. The resuls of 72 Page

9 Budge Defc, Economc Growh and FDI n he Balcs: A Cross-Secon Dependence Panel Pedron (1999) and Weserlund (2007) ess are repored n Table 4. Table 4: Panel Conegraon Resuls Pedron (GDP as dependen varable) Tes sasc Probably Whn-Dmenson Panel-v ** Panel-rho * Panel-PP ** Panel-ADF * Beween-Dmenson Group-rho Group-PP Group-ADF ** Weserlund (GDP as dependen varable) Sasc Value Z-value P-Value G *** Ga P *** Pa *** oes: Under he null ess, all varables are dsrbued normal, (0, 1). ***, **, * sgnfcan a 1%, 5% and 10% levels. In Weserlund es Akake Informaon Creron s used for opmal lag/lead lengh selecon. The resuls of Table 4 show ha here s a conegrang vecor beween he examned varables, for he group of he Balc counres. The null hypohess of no conegraon s rejeced a panel level wh boh mehods. In oher words, he resuls show ha GDP, FDI and BD are movng ogeher n he long run. 5. Panel Conegraon Esmaon Resuls We proceed by esmang he parameers of he long-run equlbrum relaonshp. The resuls of Dynamc OLS esmaons are provded n Table 5. Table 5: DOLS Esmaons Resuls (GDP as dependen varable) DOLS Independen Varables FDI Coeffcen (4.016***) BD (-4.057***) R Adj. R oes: The numbers n parenheses denoes -sasc. Asympoc dsrbuon of -sasc s sandard normal as T and go o nfny. *** sgnfcan a 1% level. Lag and lead mehod seleced by Akake. The esmaon resuls show ha here s a posve relaonshp beween FDI and economc growh n 1% level of sgnfcance. On he oher, budge defc has a negave mpac on economc growh n 1% level of sgnfcance. 6. Panel Conegraon Esmaon Resuls The panel daa resuls on he shor and long-run dynamcs beween economc growh, foregn drec nvesmens nflows and budge defc for he group of he Balc counres are provded n he nex able. Table 6: Panel Causaly Resuls Dependen Varable Source of Causaon (ndependen varables) -es F-sasc Shor-run Long-run ΔGDP ΔFDI ΔBD ECT ΔGDP *** ** ** ΔFDI ΔBD oes: Δ denoes frs dfference operaor. *** and ** sgnfcan a 1% and 5% levels. Shor-run causaly s deermned by he sascal sgnfcance of he paral F-sascs assocaed wh he rgh hand sde varables. Long-run causaly s revealed by he sascal sgnfcance of he respecve error correcon erms usng a - es. 73 Page

10 Budge Defc, Economc Growh and FDI n he Balcs: A Cross-Secon Dependence Panel From he resuls of Table 6 we see ha here s a shor-run undreconal causal relaonshp beween budge defc and economc growh wh drecon from budge defc o economc growh, as well as a shorrun causaly runnng from foregn drec nvesmens o economc growh. The esmaed coeffcen 1, of ECT, 1 n equaon (11) s sascally sgnfcan a 5% level of sgnfcan mplyng ha GDP could play an mporan adjusmen role n he long-run equlbrum. Consequenly fndngs sugges ha, n he long run, here s a undreconal causaly runnng from FDI o economc growh and a undreconal causaly relaon runnng from budge defc o economc growh. The knowledge abou he drecon of causaly wll help polcy makers o race ou polces for faser economc growh n he Balc counres. VI. COCLUSIO AD POLICY IMPLICATIOS On hs paper we examne he relaonshp among budge defc, foregn drec nvesmen and economc growh n hree ranson economes such as Balc counres. Daa of hs economerc sudy showed ha here s a sgnfcan relaonshp beween foregn drec nvesmen and economc growh as well as beween budge defc and economc growh. The error correcon model, whch was used n order o capure he shor and long run dynamc relaonshps beween seres, revealed boh n shor and long run undreconal causales runnng from foregn drec nvesmen o economc growh and from budge defc o economc growh. However, he resuls of causaly ddn show a relaonshp beween budge defc and foregn drec nvesmen. The fndngs of he sudy can be explaned as follows: The fnancal secor flourshed n he Balc counres because of he fnancal negraon. The resul was he rapd growh of cred expanson, whch accordng o he World Bank, was almos 47% of GDP durng he perod of , whle he perod was 85%. However, he resul s no he same for all hree counres. Durng he perod of , he cred expanson has been rased by hree, sx and nne mes n Esona, Lhuana and Lava, reachng n 2008 o 130%, 85% and 120% of her GDP respecvely. The savngs raes were low n he early years. However hs rend was reserved afer 2009, as he propensy o savng ncreased sgnfcanly. Durng he perod of , savngs raes of he Balc counres were beween 12% and 14% of her GDP (AMECO, 2016). The Balc economes are ofen presened as an example of successful ausery program, mplemenng he nernal devaluaon as a sraegy for economc recovery. Accordng o professor Kael (2012), n he md 2000s he Balc counres were he counres wh he hghes GDP growh raes n Europe (Esona 7.6%, Lhuana 5.9%, Lava 5.5%). Ths happened no because he mposed ausery program was successful, bu because he Balcs ousourced her recovery. There are wo reasons for he Balc recovery. The frs s he massve use of he EU budgeary funds. The second refers o he exreme negraon of he expor secor wh Scandnavan producers. Bu hese sources of growh canno be susanable n he long-run. The fundng from he EU s runnng ou and he expors manly o he Scandnavan counres are unceran. Snce reganng her ndependence, he Balc counres are dsngushed from all oher Cenral and Easern European counres, whch also were n ranson. Ther enhusasm as far as he mplemenaon of neolberal reforms s concerned. In he early 90s, hese counres adoped a mxure of polces o ensure sablzaon, fscal dscplne, prce and rade lberalzaon and large-scale prvazaon. Indeed, he Balc counres developed as sandards of neolberal ranson, wh very open economes. Afer her accesson o he EU, all hree economes had an unprecedened perod of economc boom. These economes recorded very hgh annual growh raes. However, hs remarkable growh was accompaned by double dg nflaon, rse n real esae, ncrease n real exchange raes, wage growh acceleraon beyond he ncrease of producvy, rapd accumulaon of ne foregn lables and ncreasng defcs on curren accoun balances whch, as has been proved n our sudy, have negave mpac on economc growh. REFERECES [1]. Afonso, A. and Jalles, J. T. (2011). Growh and Producvy: he Role of Governmen, Unversy of Lsbon Workng Paper, o. 13. [2]. Akbaş, Y. E. and Lebe, F. (2016). Curren Accoun Defc, Budge Defc and Savng Gap: Is he Twn or Trple Defc Hypohess Vald n G7 Counres? Prague Economc Papers, 25(3), DOI: /j.pep.565. [3]. AMECO (2016). Economc and Fnancal Affars. European Commsson: Annual Macro-Economc Daabase. [4]. Asogwa, O. and Okeke, I. C. (2013). The Crowdng Ou Effec of Budge Defcs on Prvae Invesmen n gera. European Journal of Busness and Managemen, 5(20), 161. [5]. Ball, L., Mankw,. G. (1995). Wha do budge defcs do? Harvard Insue of Economc Research Workng Paper, 1740, [6]. Basu, P., Chakrabory, C. and Reagle, D. (2003). Lberalzaon, FDI, and Growh n Developng Counres: A Panel Conegraon Approach. Economc Inqury, 41(3), DOI: /e/cbg Page

11 Budge Defc, Economc Growh and FDI n he Balcs: A Cross-Secon Dependence Panel [7]. Berhem, B. D. (1989). A eoclasscal Perspecve on Budge Defc. Journal of Economc Perspecves, 3(2), [8]. Blomsrom, M., Lpsey, R. E. and Zejan, M. (1992). Wha Explans Developng Counry Growh? aonal Bureau of Economc Research (BER) Workng Paper, o [9]. Bökemeer, B. (2015). Economc Growh and he Publc Defc n EU Member Saes n Cenral and Easern Europe. Romanan Journal of Fscal Polcy, 6(1), [10]. Bose,., Haque, M. E. and Osborn, D. R. (2007). Publc Expendure and Economc Growh: A Dsaggregaed Analyss for Developng Counres. The Mancheser School, 75(5), [11]. Breung, J. (2000). The Local Power of Some Un Roo Tess for Panel Daa, n Balag, B. H., ed., Advances n Economercs, on-saonary panels, panel conegraon, and dynamc panels. Amserdam: JAI Press, pp [12]. Breusch, T. S. and Pagan, A. R. (1980). The Lagrange Mulpler Tes and s Applcaon o Model Specfcaons n Economercs. Revew of Economc Sudes, 47, [13]. Buer, H. W. (1977). Crowdng Ou and he Effecveness of Fscal Polcy. Journal of Publc Economcs, 7(3), DOI: / (77) [14]. Carkovc, M. and Levne, R. (2005). Does Foregn Drec Invesmen Accelerae Economc Growh?, n Moran, T. H., Graham, E. M., Blomsrom, M., ed., Does Foregn Drec Invesmen Promoe Developmen? Washngon, DC: Insue of Inernaonal Economcs, pp [15]. Chang, Y. (2004). Boosrap Un Roo Tess n Panels wh Cross-Seconal Dependency. Journal of Economercs, 120(2), DOI: /S (03) [16]. De Hoyos R. E., Sarafds, V. (2006). On Tesng for Cross-Seconal Dependence n Panel Daa Models. The Saa Journal, 6(4), [17]. Densa, V. (2010). Foregn Drec Invesmen Theory: An Overvew of he Man FDI Theores. European Journal of Inerdscplnary Sudes, 2(2), [18]. Dckey, D. A. and Fuller, W. A. (1979). Dsrbuons of he Esmaors for Auoregressve Tme Seres wh a Un Roo. Journal of Amercan Sascal Assocaon, 74,(366), DOI: / [19]. Dckey, D. A. and Fuller, W. A. (1981). Lkelhood Rao Sascs for Auoregressve Tme Seres wh a Un Rroo. Economerca, 49(4), DOI: / [20]. Engle, R. F. and Granger C. W. J. (1987). Conegraon and Error Correcon: Represenaon, Esmaon, and Tesng. Economerca, 55(2), DOI: / [21]. Fsher, R. A. (1938). Sascal Mehods for Research Workers. Ollver and Boyd: Ednburgh, 7 h Edon. ISB [22]. Gorbunova, Y., Infane, D. and Smrnova, J. (2012). ew Evdence on FDI Deermnans: An Apprasal over he Transon Perod. Prague Economc Papers, 21(2), DOI: /j.pep.415. [23]. Hadr, K. (2000). Tesng for Saonary n Heerogeneous Panel Daa. The Economercs Journal, 3(2), DOI: / X [24]. Hussan, M. A. and Ahmad,. (2015). Economc Growh, FDI and Budge Defc n Paksan: : Evdence from Causaly and Conegraon Analyss. Ausralan Journal of Basc and Appled Scences, 9(5), [25]. Im, K. S., Pesaran, M. H. and Shn, Y. (2003). Tesng for Un Roos n Heerogeneous Panels. Journal of Economercs, 115(1), DOI: /S (03) [26]. Kao, C. and Chang, M.H. (2000). On he Esmaon and Inference of a Conegraed Regresson n Panel Daa. Advances n Economercs, 15, [27]. Kael, R. and Raudla, R. (2012). Ausery ha ever was? The Balc Saes and he Crss. Levy Economcs Insue of Bard College Polcy oe, 5. [28]. Levn, A., Ln, C. F. and Chu, J. (2002). Un Roo Tes n Panel Daa: Asympoc and Fne-Sample Properes. Journal of Economercs, 108, DOI: /S (01) [29]. Maddala, G. S. and Wu, S. (1999). A Comparave Sudy of Un Roo Tess wh Panel Daa and a ew Smple Tes. Oxford Bullen of Economcs and Sascs, 61(S1), DOI: / s1631. [30]. Meza, I. (2012). Fscal Defcs, Curren Defcs and Invesmen: A Panel Causaly Framework of 20 OECD Counres. Appled Economercs and Inernaonal Developmen, 12(1). [31]. Pedron, P. (1999). Crcal Values for Conegraon Tess n Heerogeneous Panels wh Mulple Regressors. Oxford Bullen of Economcs and Sascs, 61(S1), DOI: / s1653. [32]. Pedron, P. (2000). Fully Modfed OLS for Heerogeneous Conegraed Panels. Advances n Economercs, 15, DOI: /S (00) [33]. Pesaran, M. H. (2004). General Dagnosc Tess for Cross Secon Dependence n Panels. Cambrdge Workng Papers n Economcs, o [34]. Pesaran, M. H. (2007). A Smple Panel Un Roo Tes n Presence of Cross Secon Dependence. Journal of Appled Economercs, 22, DOI: /jae.951. [35]. Phllps, P. C. B. (1987). Tme Seres Regresson wh a Un Roo. Economerca, 55(2), DOI: / [36]. Phllps, P. C. B. (1995). Fully Modfed Leas Squares and Vecor Auoregresson. Economerca, 63(5), DOI: / [37]. Phllps, P. C. B. and Perron, P. (1988). Tesng for a Un Roo n Tme Seres Regresson. Bomerka, 75(2), DOI: /bome/ [38]. Phllps, P.C.B. and Sul, D. (2003). Dynamc Panel Esmaon and Homogeney Tesng Under Cross Secon Dependence. The Economercs Journal, 6(1), OI: / X [39]. Reuber, G. L., Crookell, H., Emerson, M. and Gallas-Hamonno, G. (1973). Prvae Foregn Invesmen n Developmen. Oxford: Clarendon Press. [40]. Saleh, S. A. (2003). The Budge Defc and Economc Performance: A Survey. Unversy of Wollongong Workng Paper, o. 12. [41]. Sock, J. H. and Wason, M. W. (1993). A Smple Esmaor of Conegrang Vecors n Hgher Order Inegraed Sysems. Economerca, 61, DOI: / [42]. Suela, P. (2001). Managng Capal Flows n Esona and Lava. Bank of Fnland-Insue for Economes n Transon (BOFIT) Dscusson Papers, o 17. [43]. UCTAD (2016). UCTAD Sascs. Geneva: Uned aons Conference on Trade and Developmen. [44]. WDI (2016). The World Bank. Washngon: World Developmen Indcaors. [45]. Weserlund, J. (2007). Tesng for Error Correcon n Panel Daa. Oxford Bullen of Economcs and Sascs, 69, DOI: /j x. 75 Page

12 Budge Defc, Economc Growh and FDI n he Balcs: A Cross-Secon Dependence Panel [46]. Yucel, G.E, (2014). FDI and Economc Growh: The Case of Balc Counres. Research n World Economy, 5(2). DOI: /rwe.v5n2p115. Appendx: Pedron Tes Sascs 1. Panel v-sasc Z v T 1 1 L 2. Panel ρ-sasc 2 2 ˆ11 ˆ e, ˆ2 2 ˆ2 ˆ 11, 1 11 ˆ ˆ ˆ, 1, Z L e L e e 3. Panel -Sasc (non-paramerc) 1/ Z pp 2 Lˆ eˆ Lˆ eˆ eˆ ˆ 11, 1 11, Panel -Sasc (paramerc) 1/2 * ˆ*2 ˆ2 *2 ˆ2 * * ˆ ˆ ˆ 11, 1 11, 1, Z S L e L e e 5. Group ρ-sasc ~ Z 1 T 2 eˆ eˆ eˆ ˆ, 1, 1, Group -Sasc (non-paramerc) ~ Z pp 1/ 2 T 2 ˆ e e e 2 ˆ ˆ ˆ ˆ, 1, 1, Group -Sasc (paramerc) 1/2 T ˆ*2 *2 * * ˆ ˆ ˆ, 1, 1, Z S e e e 76 Page