Does The Keynesian Absolute Income Hypothesis Exist in Pakistan?

Transcription

1 MPRA Munch Personal RePEc Archve Does The Keynesan Absolue Income Hypohess Exs n Paksan? Muhammad Shahbaz and Kshwer Nawaz and Mohamed El Hed AROURI and Frédérc Teulon COMSATS Insue of Informaon Technology, Lahore, COMSATS Insue of Informaon Technology, Lahore, EDHEC Busness School, France, IPAG Busness School, IPAG Lab, France, Lnköpng Unversy, Sweden 1. July 2013 Onlne a hp://mpra.ub.un-muenchen.de/47923/ MPRA Paper No , posed 1. July :59 UTC

2 Does The Keynesan Absolue Income Hypohess Exs n Paksan? Muhammad Shahbaz COMSATS Insue of Informaon Technology, Lahore Emal: shahbazmohd@lve.com Kshwar Nawaz COMSATS Insue of Informaon Technology, Lahore Emal: kmn78642@yahoo.com Mohamed Arour EDHEC Busness School, France E-Mal: Mohamed.arour@u-clermon1.fr Frédérc Teulon IPAG Busness School, IPAG Lab, France E-Mal: f.eulon@pag.fr Gaz Salah Uddn Lnköpng Unversy, Sweden Emal: gaz.salah.uddn@lu.se Absrac: The presen paper conrbues n exsng economc leraure by nvesgang he valdaon of he Keynesan Absolue Income hypohess n Paksan by applyng he ARDL approach o conegraon. The fndngs of hs paper ndcae he valdaon of he Keynesan absolue ncome hypohess n Paksan, where publc savngs and fnancal developmen add n prvae savngs. Ths sudy opens up new nsghs for governmen o mprove he level of prvae savngs. JEL Classfcaon: E2, C4 Keywords: Prvae savngs, Co-negraon, Paksan 1

3 1. Inroducon The Permanen Income Hypohess (PIH) of Fredman (1957) explaned he households consumpon decsons depends on her permanen ncome raher han presen ncome. The major conrbuon of he Fredman sudy s ha he elascy of consumpon whch s nfluenced by he change n ncome. There s also some emprcal evdence on he permanen ncome hypohess n he exsng leraure such as Kelley and Wllamson, (1968) and Gupa, (1970, 1971). Second, Ando and Modglan s lfe cycle hypohess posulaes ha ndvduals smooh her consumpon over lfeme and lve by accumulang savngs durng peak earnng years whch help hem o manan her consumpon levels durng reremen. Oher emprcal sudes hghlgh he mpac of demographc facors on prvae savngs behavor, such as age groups (Kelley and Wllamson, 1968); brh raes (Leff, 1969 and 1971); dependency raos (Gupa, 1971); fnancal varables such as neres rae (Oulars, 1981) and nflaon rae (Koskela and Vren, 1982) ec. Paksan s an emergng growng economy n Souh Asa. I s suable o es he mpac of economc growh on prvae savngs.e. Keynesan Absolue Income Hypohess. Foregn savngs could be an mporan deermnan of naonal savngs o smoohen exernal lqudy consrans. Muradoglu and Taskn, (1996) found ha foregn savngs are no sgnfcan n nfluencng household savngs. Emprcal evdence raher suppors a negave relaonshp beween foregn savngs and naonal savngs [see Fry, (1995); Govannn, (1983)]. Expors have a posve mpac on prvae savngs as ncreased demand for expored goods leads o ncreased oupu leadng o an ncrease n savngs. For example, Lee (1971) corroboraed he hypohess ha he level of domesc savngs vary wh level of expors. The erms of rade are anoher possble deermnan for boh publc and prvae savngs. The declne n erms of rade 2

4 provdes he ndcaon o he domesc resdens o ncrease her savngs a curren perod n order o susan her sandard of lvng n fuure perods 1. An mprovemen n erms of rade leads o an ncrease n savngs along-wh rade balance mprovemens (Fry, 1995; Masson e al. 1995). In conras, Pava and Sarwa (2003) exposed ha exernal erms of rade, no only has a posve mpac on prvae savngs bu also on publc savngs. In Turkey, Ozcan e al. (2003) clam ha he negave nfluence of lfe expecancy gves he provson of he lfe-cycle hypohess and nflaon conrol he degree of macroeconomc synchronzaon and has a posve mpac on prvae savngs 2. Accordng o he Indan experence, Ahukorala and Sen, (2004) fnd ha demographc rends, real neres rae, and he nflaon rae affecng savngs and ha publc savngs seem o crowd ou prvae savngs, bu less han proporonaely 3. Pava and Sarwa, (2003) used Brazlan daa and concluded ha prvae savngs are nversely correlaed wh publc savngs and a 1 percen ncrease n publc savngs pushed gross naonal savngs upward by 0.2 percen f all else s same. Alan and Pelgrn, (2003) repored ha prvae savngs n OECD counres have been sgnfcanly nfluenced by publc savngs, he demographc srucure of he populaon, he growh rae of labor producvy, changes n he erms of rade, and he real neres raes. Bulíř and Swson, (2006) explaned ha n Mexco, prvae savngs are adversely affeced due o relance on 1 In heory, erms of rade changes are already a par of real GDP. However, n pracce he prce deflaors used n naonal accounng generally allow only for changes n he general level of prces and fal o capure prce srucural effecs on he level and growh of real ncome such as hose due o changes n he erms of rade. Thus changes n TOT can be expeced o have an addonal effec o ha of changes n GDP on savngs (Ahukorala and Sen, 2004) 2 Income level has a posve mpac on he prvae savng rae and growh rae of ncome s no sascally sgnfcan. Prvae cred and real neres raes ry o capure he severy of he borrowng consrans and he degree of fnancal represson for Turkey. 3 Oher facors ha effec prvae savngs lke he spread of bankng facles n he economy and he rae of nflaon seem o have a posve mpac and changes n he exernal erms of rade and mgran remances a negave mpac on prvae savng. 3

5 exernal savngs, relavely hgh populaon dependency rao, and less developed fnancal sysem. The leraure on economc growh and prvae savngs s nconclusve [see Carrol and Wel, (1994); Edwards, (1996); Hussan, (1996); Loayza e al. (2000); Ozcan e al. (2003); Ahukorala and Sen, (2004) and, Pava and Sarwa (2003)]. Modglan s (1970) lfe-cycle model reveals producvy growh makes workng young rcher han old and he young save more whle old dssave 4. Moreover, Carroll and Wel (1994) argued ha people adjus consumpon habs slowly, whch makes savngs posvely correlaed wh curren growh of ncome. The relaonshp beween neres raes and prvae savngs s unceran (see Schmd-Hebbel and Serven, 1999) whle Ogak e al. (1995) explored he sensvy of prvae savngs wh rae of neres. They argued ha prvae savngs are more reacve o raes of reurn a hgher ncome levels. A lower ncome levels, people canno smooh ou consumpon over me. Ths mples ha neremporal elascy of subsuon beween presen and curren consumpon vares wh he level of wealh. Leraure also reveals ha an mporan reason o save s he precauonary move, as people save more a mes of ndecson apprehendng he possbly of dffcul mes ahead. Such a source of uncerany s of macroeconomc naure. Ths can be refleced n hgh and nconssen nflaon, exchange rae volaly, cycles of boom and conracon, nsably n he fnancal sysem. One reply o hese unceranes s capal flgh as people leave domesc asses due o hs uncerany (see for deals; Edwards, 1996 and Taylor, 1996). 4 Aggregae ncome growh would follow from ncreasng he lfeme profles for succeedng generaons. In urn, hab formaon n consumpon s a facor ha helps o dmnsh he posve correlaon beween savngs and growh. Economc agens wll have negave savngs when hey are young and have very low ncome, posve savngs durng her producve years and negave savngs when hey are old and rered (Modglan, 1970) 4

6 Low fscal defcs or surpluses buld up naonal savngs, as complee Rcardan equvalence has been refued emprcally (.e. an ncrease n publc savngs s no fully offse by a declne n prvae savngs). Ths ype of effec s sronger n developng counres subjec o survval consumpon and lqudy consrans (see Corbo and Schmd-Hebbel, 1991). The evdence confrms ha governmen savngs parally offse prvae savngs wh he offse coeffcen n he range of percen 5. Dayal-Gula and Thmann, (1997) emprcally observed ha fscal polcy parcularly socal safey arrangemens nfluence prvae savngs. Macroeconomc sably and fnancal deepenng appear o be mporan deermnng facors of savngs behavor n he Souheas Asa and Lan Amercan economes. Furhermore, Edwards (1996) ncorporaed some polcy relaed, demographc, srucural, and polcal varables o deermne boh publc and prvae savngs. The domesc savngs (boh prvae and publc) are srongly nfluenced by he oupu growh per capa. However, polcal uncerany has negave nfluence on publc savngs. Ozcan and Ozcan, (2000) explaned ha he real per capa ncome s posvely assocaed wh prvae savngs, whch corroboraes he lfe-cycle hypohess. However, hey are sascally nsgnfcan and smlarly he Rcardan equvalence does no hold callously 6. In general, prvae savngs s nfluenced by he age dsrbuon of he populaon. Ths fundamenal hough s applcable o he rered people [Lahr, (1989); Edwards, (1996); Dayal- 5 Ths means ha 1 percenage of addonal governmen savngs (n erms of GDP) adds abou 0.5 percenage of GDP o naonal savngs. 6 Loayza, e al. (2000) he mean savng rae for MENA counres drops o 0.24 when he rae for Arab Emraes s excluded. Excludng hree hgh savng counres (Kuwa, Bahran and Arab Emraes), he mean savng rae becomes The medan value s If only seven MENA counres (Egyp, Iran, Jordan, Morocco, Syra, Tunsa and Turkey) are consdered he mean savng rae drops o a low of 0.17 whch s almos equal o he world savng rae. Ths paper also examnes emprcal deermnans of prvae savng for a sample of economes n he Mddle Eas and Norh Afrca (MENA) over he perod I argues ha he mean savng rao for he seleced MENA counres s over he mean world savng rao (0.26 vs. 0.16, respecvely) 5

7 Gula and Thmann, (1997), Loayza e al. (2000)]. Paul e al. (1993) suggesed ha publc savngs and foregn savngs parally offse prvae savngs n developng counres. In addon, hey found ha demographcs and growh raes are mporan deermnans of prvae savngs. There are many sudes abou he mpac of fnancal lberalzaon on prvae savngs n a number of boh developed [(Bayoum, (1993a, b); Caporale and Wllams, (2001); Chapple, (1991)] and developng counres [Melo and Tybou, (1986); (Bandera e al. (2000); Dayal- Gula and Thmann (1997) and; Fry, (1995); Hussan, (1996); Jbl e al. (1997); Loayza and Shankar, (2000)]. However, her resuls are ambguous. Recenly, Jongwanch (2010) nvesgaed he mpac of economc growh, nflaon and erms of rade on prvae savngs usng daa of Thaland. The emprcal resuls exposed ha economc growh and erms of rade ncrease prvae savngs bu nflaon declnes. Keho, (2011) nvesgaed he deermnans of prvae savng n case of Wes Afrcan Economc and Moneary Unon counres by applyng he ARDL bounds esng approach. The resuls showed ha fnancal lberalzaon plays mporan role o explan prvae savng behavor for sampled counres. Sackey, (2011) repored ha nernaonal remances play a sgnfcan role o boos prvae savng n case of Ghana. Johnson, (2011) nvesgaed he presence of Lfe Cycle Hypohess n case of Ngera. The emprcal resuls ndcaed ha economc growh, growh n dsposable ncome and real neres rae have posve effec on prvae savngs. Addonally, publc savngs crowd ou prvae savngs. Maur e al. (2012) examned he deermnans of prvae savngs usng Turksh daa. They appled Johansen conegraon and found he presence of conegraon among he seres. Ther emprcal evdence repored ha publc savngs, ncome per capa, fnancal developmen, old and young dependency raos, exchange rae, urbanzaon 6

8 crowd ou prvae savngs. Dsposable ncome per capa, nflaon, real neres rae, money supply, female labor force parcpaon, drec and ndc axes have posve effec on prvae savngs. Larb, (2013) examned he mpac of fnancal lberalzaon, ncome per capa, nflaon and fscal defc on prvae savngs n case of Ghana. The auhor found ha fnancal lberalzaon, ncome per capa and nflaon have posve mpac on prvae savngs. The Rcardan Equvalence hypohess s valdaed n Ghana because fscal defc ncreases prvae savngs. In case of Paksan, Burney and Khan (1992) found ha he burden of hgh dependency rao reduces household savngs. Husan (1996) analyzed he long-run behavor of savngs and exposed ha fnancal developmen adds n prvae savngs. Hgh raes of populaon growh have kep he age srucure of Paksan s populaon vrually unchanged. Laer on, Farhan and Akram (2011) denfed he deermnans of prvae savngs by applyng he ARDL bounds esng and error correcon mehod (ECM). They found ha ncome growh s posvely lnked wh household savngs bu nflaon and age dependency rao lower. Ismal and Rashd, (2013) also re-examned he deermnans of household savngs. Ther emprcal evdence repored ha nflaon reduces household savngs bu ncome per capa ncreases. Publc savngs and old dependency rao add n household savngs. In hs paper, our conrbuon s no n erms of mehodology. I s raher n erms of fndngs. In hs presen sudy, he ARDL bounds esng procedure o conegraon s appled o examne he long run relaons beween he real prvae savngs, real GDP, real domesc cred o prvae secor, real publc savngs, polcal nsably dummy and real agrculure value-added per 7

9 capa. Ths approach s suable for small sample daa. The ARDL bounds esng s flexble regardng order of negraon. We can apply f our varables are negraed a I(0) or I(1) or havng mxed order of negraon. Ths approach provdes conssen and effcen emprcal evdence compared o convenonal conegraon approaches. For shor run dynamcs, we use error correcon mehod (ECM) and long run margnal mpac of ndependen varables s nvesgaed by applyng OLS regresson. Our fndngs show he presence of long run relaonshp among he varables. We fnd ha he Keynesan absolue ncome hypohess holds n case of Paksan for long and shor runs. Furhermore, publc savngs and fnancal developmen add n prvae savngs and agrculure growh enhances he propensy of people o save n Paksan. 2. Modelng, and Mehodologcal Framework, and Daa We begn wh he followng basc equaon of prvae savngs for Paksan. In he presen sudy, prvae savngs and oher varables are n log-form excep ha of polcal nsably. A loglnear model provdes more comprehensve esmaes han he smple-lnear form [Bowers and Perce, (1975); Ehrlch, (1975); Ehrlch, (1977); Layson, (1983); Cameron, (1994) and Ehrlch, (1996)] 7. ln PS lny ln INF ln FD ln PUB PL ln AGR (1) Y INF FD PUB PL AGR where, PS s real prvae savngs per capa, Y s real GDP per capa, INF s nflaon proxes by consumer prce ndex, FD s real domesc cred o prvae secor per capa proxy for 7 For more deals see Shahbaz e al. (2010) 8

10 fnancal developmen, PUB s real publc savngs per capa, PL s for polcal nsably (D = 0 where here s democrac perod oherwse zero), AGR s real agrculure value-added per capa and s error erm assumed o be normally dsrbued. The daa perod of sudy consss on Daa on real GDP, consumer prce ndex, real domesc cred o prvae secor and real agrculure value added o GDP has been obaned from economc survey (varous ssues) of Paksan. The monhly sascal bullen of Sae Bank of Paksan s combed o collec daa on real prvae savngs and real publc savngs. We have use populaon seres o conver all seres no per capa excep consumer prce ndex. Sascal leraure provdes many economercal echnques o nvesgae he conegraon among he macroeconomc acors n developmen economcs. In hs sudy, he auoregressve dsrbued lag (ARDL) approach for conegraon by Pesaran e al. (2001) has been employed. Recen sudes have ndcaed ha he ARDL bounds esng approach o conegraon s preferable o oher convenonal conegraon approaches such as Engle-Granger (1987), and Hansen (1996). One of he reasons for preferrng he ARDL echnque s ha s applcable rrespecve of wheher he underlyng regressors are purely I(0), purely I(1) or muually conegraed. Ths procedure s he famlar wh Wald es or F-sascs n a generalzed Dckey-Fuller ype regresson, whch s used o es he sgnfcance of lagged levels of he varables under consderaon n a unresrced equlbrum error correcon model (UECM) 9

11 10 (Pesaran e al. 2001) 8. We use he augmened ARDL (p, q 1, q 2, q k ) developed by Pesaran and Pesaran, (1997); Pesaran e al. (2001) as followng: k w x q L y p L ' 1 ), ( ), ( (2) n, 1,... where k L L L q L L L L p L q q p p..., 1,2,... ), (... 1 ), ( Where ndependen varable s y, consan erm s α, lag operaor s denoed by L such ha Ly = y 1, w s s 1 vecor of deermnsc varables for example consan erm, me rend, or ndependen varables havng approprae lags. The x n equaon-2 s he ndependen varable where = 1, 2,, k. we may wre long run emprcal model wh consan erm as followng: k w x y ' 1 ) 1, ( p (3) The long run mpac of x varables on y varable can be examned by: p q p q ) (1, ) (1, 2 1 1,...,k 1,2 (4) 8 Anoher reason for usng he ARDL approach s ha s more robus and performs beer for small sample szes (such as n hs sudy) han oher co-negraon echnques.

12 Where pˆ and qˆ, = 1, 2,, k show he esmaes of pˆ and qˆ, = 1, 2,, k. we may esmae long run coeffcens usng followng equaon: q k ( p, q1, q2,..., ) (5) p The OLS esmaes of he ARDL model are ˆ( pˆ, qˆ, qˆ,... qˆ ) for equaon (2). Followng 1 2 k Pesaran e al. (2001), we can oban long run as well as shor run esmaes usng general o specfc verson of he ARDL bounds esng as followng: y k p1 k q 1 ' ( 1, p) EC 1 x w jy 1 jx, j (6) 1 j1 1 j1 where ECM s he error correcon model and s defned as follows: ECM y x ' w (7) x s he k-dmensonal forcng varables, whch are no co-negraed among hem. ε s a vecor of sochasc error erms, wh zero means and consan varance-covarance 9. There are wo seps o apply he ARDL bounds approach o conegraon. In frs sep, we examne he presence of 9 The exsence of an error-correcon erm among a number of co-negraed varables mples ha changes n dependan varable are a funcon of boh he levels of dsequlbrum n he co-negraon relaonshp (represened by he ECM) and he changes n he oher explanaory varables. Ths ells us ha any devaon from he long run equlbrum wll feed back on he changes n he dependan varable n order o force he movemen owards he long run equlbrum (Mash and Mash, 2002). 11

13 long run relaonshp beween he varables followng approprae lag lengh selecon. We have followed Akake nformaon crera o selec lag lengh. Secondly, we esmae long run as well as shor run coeffcens usng equaon-6. The second sep appled once conegraon beween he varables s confrmed (Narayan and Smyh, 2005). The unresrced nercep and unresrced rend (Pesaran e al. 2001) verson of he ARDL model s as followng: PS p1 0 PL c PL FD 1 FD c 1 1 AGR p1 0 PS PS AGR NS 1 1 NS p1 1 1 GDPC PS PS p1 0 GDPC PL 1 1 PL p1 0 1 INF GDPC p1 0 INF 1 GDPC AGR FD 1 AGR FD 1 1 p1 0 INF NS NS 1 INF 1 (8) The Wald es (F-sascs) for he null hypohess H : PS GDPC INF FD NS PL AGR0 and alernave hypohess H : 0. The asympoc 1 PS GDPC INF FD NS PL AGR dsrbuons of he F-sascs are non-sandard under he null hypohess of no co-negraon relaonshp beween he examned varables, rrespecve of wheher he varables are purely I(0) or I(1), or muually co-negraed. Two ses of asympoc crcal values provded by Pesaran and Pesaran, (1997). The frs se assumes ha all varables are I(0) whle he second se assumes ha all varables are I(1). If he compued F-sasc s greaer han he upper bound crcal value, hen we rejec he null hypohess of no co-negraon and conclude ha here exss seady sae equlbrum beween he varables. If he compued F-sasc s less han he lower bound crcal value, hen we canno rejec he null of no co-negraon. If he compued F-sasc fall whn he lower and upper bound crcal values, hen he resul s nconclusve. In hs case, followng Kremers e al. (1992) and Bannerjee e al. (1998), he error correcon erm wll be a 12

14 useful way of esablshng co-negraon. The second sep s o esmae he long-run coeffcen of he same equaon and he assocaed ARDL error correcon models. 3. Emprcal Resuls and her Dscusson Fndng he order of negraon of prvae savngs wh he baery of oher varables s prerequse o nvesgae conegraon hrough employng he ARDL bounds esng for long run assocaon. I s necessary for ARDL bounds esng ha varables should be negraed a I(0) or I(1) or I(0)/I(1). In dong so, Ng-Perron, (2001) un roo es s appled o make sure ha no varable s havng 2 nd dfference order of negrang. ADF, P-P and DF-GLS ess are no relable for small sample daa ses due o her poor sze and power properes (Dejong e al. 1992; Harrs and Solls, 2003). These ess seem o over-rejec he null hypoheses when s rue and vce versa. Ng-Perron, (2001) un roo es seems o solve hs problem. The resuls of Ng- Perron un roo es are repored n Table-1. We fnd ha ha NP does no show problem of un roo and fond o be saonary a level.e. I(0). The res of he varables such as and ln Y, ln INF, ln FD, ln PUB, PL and ln AGR are negraed a 1 s dfference.e. I(1) wh nercep and rend. Ths shows ha varables do have mxed order of negraon. In such suaon, he ARDL bounds esng approach o conegraon s suable o examne long run relaonshp among he varables. 13

15 Table-1 Un Roo Decson Ng-Perron a Level Varables MZa MZ MSB MPT ln PS *** ln Y ln INF ln FD ln PUB PL ln AGR Ng-Perron a 1 s Dfference Varables MZa MZ MSB MPT ln PS *** ln Y *** ln INF ** ln FD ** ln PUB ** PL ** ln AGR * Sensvy Analyss 14

16 Seral Correlaon LM, F = (0.6617) ARCH Tes = (0.6630) Normaly J-B Value = (0.5971) Heeroscedscy Tes, F = (0.4833) Ramsey RESET Tes, F = (0.7303) Noe: ***, ** and * show sgnfcan a 1%, 5% and 10% respecvely. We have o selec approprae lag order of he varables usng unresrced VAR. The approprae selecon of leg lengh s helpful o compue ARDL F-sasc n examne wheher conegraon exss or no. We selec lag lengh whch s 2 n our sample daa and our resuls are based on Akake nformaon creron (AIC). The AIC performs beer n selecon of lag lengh due o s superor power properes. Our resuls of he ARDL bounds esng are repored n Table-2 ha our calculaed F-sasc s more han upper crcal bound a 10 percen level of sgnfcance followng Pesaran e al. (2001). One may conclude ha here prevals a conegraon beween prvae savngs and s macroeconomc deermnans. Table-2 ARDL Bounds Tesng Lag Order F-Sasc * Pesaran e al. (2001) a Narayan P (2005) b 15

17 Crcal Value Lower Upper Lower Upper Bound Bound Bound Bound Value Value Value Value 10 1 percenage percenage percenage Noe: * here s conegraon among runnng acors n concerned model. Dependan Varable = ln PS Table-3 Long Run Elasces Coeffce Prob- Coeffcen Prob- Coeffcen Prob-values Varable n values values Consan ln PS ln Y ln PUB ln R ln INF ln AGR a Crcal values are obaned from Pesaran e al. (2001), Table CV (IV): Unresrced Inercep and no Trend b Crcal values are obaned from Narayan (2005), Table CV (IV): Unresrced Inercep and no Trend, p

18 ln FD PL ln INV ln TOT ln PREM ln DPEN ln M ln M ln ER R-Squared = Adj-R-Squared = Durbn-Wason = F-Sasc = R-Squared = Adj-R-Squared = Durbn-Wason = F-Sasc = R-Squared = Adj-R-Squared = Durbn-Wason = F-Sasc = The margnal mpac of macroeconomc deermnans on prvae savng s repored n Table-3. We fnd ha economc growh (ncome per capa) s posvely lnked wh prvae savngs. I s sascally sgnfcan a 1 per cen level of sgnfcance. A 1 per cen ncrease n economc growh (ncome per capa) ncreases prvae savngs by per cen, all else s same 11. Ths suppors he Keynesan absolue ncome hypohess. These fndngs are conssen wh Modglan (1986) and, Hussan and Thrlwall, (1999) who clamed ha rse n ncome per capa ncrease savngs capacy whch s he mos mporan deermnan of economc growh. 11 see for more deals; Edwards, (1996) 17

19 The relaonshp beween publc and prvae savng s posve and sgnfcan a 1 per cen level. A 1 per cen ncrease n publc savngs pushes he prvae savngs upwards by only per cen by keepng oher hngs consan. The mpac of macroeconomc nsably declnes prvae savngs and s sascally sgnfcan a 1 per cen by keepng oher hngs consan. More han 65 per cen populaon of Paksan lves n rural areas and more han 90 per cen of rural populaon s assocaed wh agrculural secor (AGR) drecly and ndrecly. Boos n economc acves of agrculural secor defnely mproves he economc suaon of vllage populaon. Ths also rases her power o save. The relaon beween agrculural growh and prvae savngs s posve a 1 per cen level of sgnfcance. We fnd ha a 1 per cen ncrease n agrculural growh rases prvae savngs by 1.37 per cen, all else reman same. The mpac of fnancal developmen s posve and s sascally sgnfcan a 1 per cen. Keepng oher hngs consan, a 1 per cen ncrease n fnancal developmen may rase prvae savngs per cen 12 [enhancemen n cred o prvae secor as share of GDP] 13. The lnk beween polcal nsably ( PL ) and prvae savngs s negave and s sascally sgnfcan a 5 per cen. The polcal nsably lowers prvae savngs va lowerng nvesmen acves. The declne n nvesmen acvy s lnked wh an ncrease n unemploymen. The ncreased unemploymen lowers prvae savngs. The mpac of prvae savngs n curren perod s posvely lnked wh prvae savngs n fuure. A 1 percen ncrease n prvae savngs n curren perod wll rase prvae savngs n he fuure by per cen by keepng oher reman same (see for deals; Bulr and Swson, 2006). The relaonshp beween nvesmen and prvae savngs s posve and s sascally sgnfcan a 1 per cen. Invesmen generaes employmen opporunes for boh sklled and 12 Two ndcaors of fnancal developmen Cred o prvae secor as share of GDP and M2/GDP ulzed. 13 See Loayza, Schmd-Hebbel and Servn (1999). 18

20 unsklled labor. Ths ndcaes he mporance of nvesmen o mprove he prvae savngs n Paksan. A per cen ncrease n prvae savngs s lnked wh a 1 per cen ncrease n nvesmen, all else reman same. The erms of rade ( ln TOT ) s also a deermnan of prvae savngs. Ths mples ha a permanen deeroraon n erms of rade s lnked wh hgh savngs n prevous perod jus o manan her sandard of lvng n curren perod confrmed by ADY, (1976); Pava and Sarwa, (2003). In case of Paksan, hgher prvae savngs are assocaed wh deeroraon of erms of rade. A 1 per cen deeroraon n erms of rade reduces prvae savngs by per cen by assumng oher hngs consan. Dependency rao s aken as demographc varable (DPEN) 14 o examne s mpac on prvae savngs accordng o precauonary savngs model. We fnd ha low savngs are alled wh hgh dependency rao. Ths confrms he Deaon s (1991) buffer sock approach o prvae savngs n Paksan. All oher facors reman same, a 1 per cen ncrease n dependency rao reards prvae savngs. Accumulaed wealh ( ln M 3 ) lessens margnal propensy o save of ndvduals n Paksan bu s nsgnfcan. The relaonshp beween quas money supply ( ln M 2 ) and prvae savngs s posve and s sgnfcan a he 5% sgnfcance level. A 1 per cen rase n money supply s boosed prvae savngs by per cen, all else s same. We fnd ha he relaonshp beween neres rae and prvae savngs s posve a 1% level of sgnfcance. Ths shows ha a rse n neres rae s lnked wh presen consumpon raher han fuure consumpon (subsuon effec) and hs nclnes he people o ncrease savngs 15. Thus, savngs respond posvely due o rse n he neres raes only f he subsuon 14 Ths shows he dependency rao 15 However, f he household s a ne lender, he neres rae rse also rases lfeme ncome, and hus ends o ncrease consumpon and decrease savng (he ncome effec). 19

21 effec s sronger han he ncome effec. I s argued ha real neres rae affecs prvae savngs posvely n Paksan. In Paksan, savngs process s money concenraed because of lmed porfolo choces. Due o change n neres rae, subsuon effec domnaes ncome as savngs come from small ndvduals compared o ele class populaon. Prvae savngs and foregn remances move n oppose drecons. The man reason s ha he recpens of remances fnd no ncenves o save her money n banks. Inally, hey spend her money on consumpon needs. Wh he passage of me, recpens of remances nves her money n consrucon of houses, real esaes, and wase huge amouns of money on weddng ceremones. Fnally, fnancal reforms seem o have negave correlaon wh prvae savngs bu s no nsgnfcan. We have appled error correcon model (ECM) o oban shor run dynamc relaonshp and resuls are repored n Table-4. We fnd ha esmae of lagged error erms s negave and s sascally sgnfcan. The sascal sgnfcance of lagged error erm wh negave sgn shows he speed of adjusmen from shor run owards long run equlbrum pah. Table-4: Shor Run Behavor Dependen Varable = ln PS prob. Varable Coeffcen T-sasc value Consan lny ln FD

22 ln AGR ln PUB ln INF PL ECM R-squared = Adj-R-squared = Durbn-Wason = F-sasc = AIC = SBC = The coeffcen s for shor-run model mples ha devaon from he long-erm n prvae savngs s correced by per cen every year. Sgnfcance of error erm s anoher proof of esablshed long run relaonshp beween prvae savngs and s deermnans n case for Paksan. Table-4 shows ha ncrease n GDP per capa mproves he prvae savngs bu s nsgnfcan. Fnancal developmen ncreases prvae savngs a 10 per cen level. Enhanced publc savngs and ncreased conrbuon of agrculure secor o GDP push ndvduals o save more. Macroeconomc nsably declnes he prvae savngs and s sascally sgnfcan a 10 per cen level. Polcal nsably reards prvae savngs bu s nsgnfcan. The model passes all sensvy analyss ess and ndcaes ha here s no seral correlaon and same nference s drawn for auoregressve condonal heeroscedscy. The error erm s normally dsrbued and here s no whe heeroscedscy. Ramsey Res es ndcaes ha he model s well specfed. 21

23 4. Concluson and Polcy Implcaons Ths sudy explaned he possble assocaon beween prvae savngs and macroeconomc deermnans n Paksan durng he perod of The ARDL approach has been employed o nvesgae he long run relaons beween hese varables. Our emprcal evdence hghlghs ha prevous savngs hke probably o save more n forhcomng perods o oban profs and o smooh ou her consumpon. Ths mples ha he rse n ncome mproves prvae savngs and confrms he exsence of Keynesan absolue ncome hypohess n he counry. Resuls also clam ha publc savngs and fnancal developmen are posvely correlaed wh prvae savngs. Ineresngly, agrculure secor s conrbuon o GDP and also mproves ndvduals capacy o save n he counry. The emprcal fndngs of he analyss provde he aenon of moneary and polcal nsables reduce prvae savngs, whereas mprovemens n nvesmen acves rase prvae savngs hrough absorpon of boh sklled and unsklled labor ha mproves no only her ncome bu also her capacy o save o smooh ou her consumpon. A deeroraon of erms of rade ha s perceved o be permanen may nduce he domesc resdens o ncrease her savngs o susan her sandard of lvng. Demographc varable has negave mpac on savngs and accumulaon of wealh. Ineresngly we observe ha ncremen n nernaonal remances pushes he prvae savngs downward n Paksan. In he conex of polcy recommendaons, here s need o conrol nflaon n he counry o boos prvae savngs. Subdued nflaon no only enhances ndvduals ably o save bu also reduces he cos of borrowng o nves n employmen generang acves. Agrculure secor s he backbone of he counry bu here are lmed fnancal openngs avalable o local peasans hus, encourage hem use her savngs o buy gold, houses, land ec. nsead of channelzng her 22

24 savngs no banks commercal or ndusral nvesmen (Shahbaz e al. 2013). There needs o mplemen such polces whch no only promoe agrculure secor bu also nroduce schemes o promoe savngs habs of peasans n he formal fnancal nsuons. Polcal sably s a prerequse n he counry o arac no only local nvesors bu also foregners. In a frendly polcal envronmen, nvesors are expeced o launch her long-erm projecs n he counry, whch, n urn, generae employmen opporunes leadng o ncreased prvae ncome and prvae savngs n case of Paksan. References 1. Alan de, S and F. Pelgrn., (2003). The Declne n prvae savng raes n he 1990 n OECD counres: how much can be explaned by Non Wealh deermnans?. OECD Economc Sudes No. 36, 2003/1 2. Ahukorala, P., Sen, K., (2004). The deermnans of prvae savngs n Inda. Workng Paper, No.2001/13, ASARC, Ausralan Naonal Unversy. 3. Bandera, O., G. Capro, P. Honohan, and F. Schanarell. (2000). Does fnancal reform rase or reduce prvae savngs? Revew of Economcs and Sascs, 82 (2), pp: Bannerjee, A., J. Dolado and R. Mesre., (1998), Error-correcon mechansm ess for conegraon n sngle equaon framework, Journal of Tme Seres Analyss, 19, pp: Bayoum, T. (1993a), Fnancal deregulaon and consumpon n he Uned Kngdom, Revew of Economcs and Sascs, 75, pp: Bayoum, T. (1993b). Fnancal savng and household savng. Economc Journal, 103, pp: Bowers, W. and Perce, G., (1975), The lluson of deerrence n Isaac Ehrlch s work on he deerren effec of capal punshmen, Yale Law Journal, 85, pp: Bulíø, A., and Swson, A., (2006), Wha explans prvae savngs n Mexco? IMF WP/06/

25 9. Burney, N. A., and A. H. Khan., (1992), Soco-economc characerscs and household savngs: an analyss of he households savngs behavor n Paksan, The Paksan Developmen Revew, 31(1), pp: Cameron, S., (1994), A revew of he economerc evdence on he effecs of capal punshmen, Journal of Soco-economcs, 23, pp: Caporale, G. and G. Wllams., (2001), Moneary polcy and fnancal lberalzaon: he case of Uned Kngdom consumpon Journal of Macroeconomcs, 23(1), pp: Carroll C., and Wel, D., (1994), Savng and growh: a renerpreaon, Carnege- Rocheser Conference Seres on Publc Polcy, 40, pp: Chapple, S., (1991), Fnancal lberalzaon n New Zealand, Uned Naons Conference on Trade and Developmen Dscusson Papers No. 35, Geneva, UNCTAD, March. 14. Corbo, V. and K. Schmd-Hebbel., (1991), Publc polces and savngs n developng counres, Journal of Developmen Economcs, 36, pp: Dayal-Gula, A. and C. Thmann., (1997), Savng n Souheas Asa and Lan Amerca compared: searchng for polcy lessons Workng Paper WP/97/ Deaon, A. (1991), Savngs and lqudy consrans, Economerca, 59(1), pp: Dejong, D.N., Nankervs, J.C., Savn, N.E., (1992), Inegraon versus rend saonary n me seres, Economerca, 60, pp: Edwards, S., (1996). Why are Lan Amerca s savngs raes so low? an nernaonal comparave analyss Journal of Developmen Economcs, 51, pp: Ehrlch, I., (1975), The deerren effec of capal punshmen A queson of lfe and deah, Amercan Economc Revew, 65, pp: Ehrlch, I., (1977), The deerren effec of capal punshmen reply, Amercan Economc Revew, 67, pp: Ehrlch, I., (1996), Crme, punshmen and he marke for offences, Journal of Economc Perspecves, 10, pp: Engle, R. F and Granger C. W. J., (1987), Conegraon and error correcon represenaon: esmaon and esng, Economerca, 55, pp:

26 23. Farhan, M., Akram, M., (2011). Does ncome level affec savng behavor n Paksan? An ARDL approach o conegraon for emprcal assessmen. Far-Eas Journal of Psychology and Busness. 3(3), Fredman, M., (1957), A heory of he consumpon funcon. Prnceon Unversy Press for he Naonal Bureau of he Economc Research. 25. Fry, M., (1995), Money, neres and bankng n economc developmen. 2nd Ed., London: Johns Hopkns Unversy Press. 26. Govannn, A., (1983), The neres elascy of savngs n developng counres: he exsng evdence, World Bank, 11 (July). 27. Gupa, K. L., (1970), Foregn capal and domesc savngs: a es of Haavelmo's hypohess wh cross-counry daa: a commen, Revew of Economcs and Sascs, 52, pp: Gupa, K. L., (1971), On some deermnans of rural and urban household savngs behavor, Economc Record, 46, pp: Hansen, B. E., (1996), Erraum: he lkelhood rao es under nonsandard condons: esng he Markov Swchng Model of GNP, Journal of Appled Economercs, 11(2), pp: Harrs, R., Solls, R., (2003), Appled Tme Seres Modelng and Forecasng. Wley, Wes Sussex. 31. Hussan, A., (1996), Prvae savng and s deermnans: he case of Paksan, The Paksan Developmen Revew, 35(1), pp: Hussan, N., (1996), Fnancal lberalzaon, currency subsuon and nvesmen: he case of Egyp, Afrcan Developmen Bank Economc Research Papers, No. 24, Adds Ababa, Afrcan Developmen Bank. 33. Ismal, A., Rashd, K., (2013). Deermnans of household savng: conegraed evdence from Paksan ( ). Economc Modelng, 32, Jbl, A., Enders, K. and Trechel, V. (1997), Fnancal Secor Reforms n Morocco and Tunsa. Inernaonal Moneary Fund Workng Paper 81, Washngon, DC: IMF, July. 35. Johnson, A. O., (2011). The nexus of prvae savngs and economc growh n emergng economy: a case of Ngera. Journal of Economcs and Susanable Developmen, 2(6),

27 36. Jongwanch, J., (2010). The deermnans of household and prvae savngs n Thaland. Appled Economcs, 42, Keho, Y., (2011). Long-run deermnans of savngs rae n WAEMU counres: an emprcal assessmen from ARDL bounds esng approach. Souh Afrcan Journal of Economcs, 73(3), Kelley, A. and Wllamson, J. (1968), Household savngs behavor n he developng economes: he Indonesan case, Economc Developmen and Culural Change, 16(3), pp: Koskela, E. and Vren, M., (1982), Savng and nflaon: some nernaonal evdence, Economcs Leers, 9(4), pp: Kremers, J. J. M., N. E. Ercsson and J. J. Dolado., (1992), The power of conegraon ess, Oxford Bullen of Economcs and Sascs, 54, pp: Lahr, A., (1989), Dynamcs of Asan savngs, Saff Papers, Inernaonal Moneary Fund, 36, pp: Larb, D. A., (2013). The long run deermnans of prvae domesc savngs n Ghana: a conegraon approach. Journal of Economcs and Susanable Developmen, 4(4), Layson, S., (1983), Homcde and deerrence: anoher vew of he Canadan me seres evdence, Canadan Journal of Economcs, 16, pp: Lee, J-K, (1971). Expors and he Propensy o Save n LDCs. The Economc Journal, 81, pp Leff, N. H., (1969), Dependency raes and savngs raes, Amercan Economc Revew, 59, pp: Leff, N. H., (1971). Dependency raes and savng raes: reply Amercan Economc Revew, 61, pp: Loayza N., Schmd-Hebbel K and Serven, L., (2000), Wha drves prvae savngs across he world?, Revew of Economcs and Sascs, 82 (2), pp: Loayza, N. and R. Shankar., (2000), Prvae savngs n Inda. The World Bank Economc Revew, 14(3), pp: Mash and Mash, (2002), Energy consumpon, real ncome and emporal causaly, Energy Economcs, 18, pp:

28 50. Masson, P., Bayoum, T. and Same, H., (1995), Inernaonal evdence on he deermnans of prvae savngs, IMF Workng Paper 95/51, Washngon: Inernaonal Moneary Fund. 51. Maur, E, P., Sabuncu, A., Bahcec, S., (2012). Deermnans of prvae savng and neracon beween publc & prvae savngs n Turkey. Topcs n Mddle Easern and Afrcan Economes, 14, pp Melo, J. and J. Tybou., (1986), The effecs of fnancal lberalzaon on savngs and nvesmen n Uruguay, Economc Developmen and Culural Change, pp: Men Ozcan, K. and Ozcan, Y. Z. (2000). Deermnans of prvae savng n MENA regon, Iran and Turkey. Deermnans of prvae Savngs n MENA regon, Iran and Turkey. Presened a World Bank MDF 3 Conference held n Caro and avalable n he Proceedngs of he World Bank MDF3 Conference webse a hp:// (K. Men- Özcan). 54. Modglan, F., (1970), The lfe cycle hypohess of savngs and ner counry dfferences n he savng rao n nducon, growh and rade, Eded By W. Els, M Sco and J. Wolfe, London: Oxford Unversy Press. 55. Modglan, F., (1986), Lfe cycle, ndvdual hrf and he wealh of naons, Amercan Economc Revew, 76, pp: Muradoglu, G., and Taskn, F. (1996). Dfferences n household savngs behavor: Evdence from Indusral and Developng counres. The Developng Economes, 34, pp: Narayan, P. K., (2005), The savngs and nvesmen nexus for Chna: evdence from conegraon ess, Appled Economcs, 37(17), pp: Narayan, P.K., Smyh, R., (2005b), Temporal causaly and he dynamcs of democracy, emgraon and real ncome n Fj, Inernaonal Revew of Appled Economcs, 19 (2), pp: Ng, S., and Perron, P., (2001), Lag lengh selecon and he consrucon of un roo es wh good sze and power, Economerca, 69, pp: Ogak, O. and Renhar., (1995), Savng behavor n low and mddle-ncome developng counres: a comparson, IMF Workng Paper No. 95/ Oulars, S. (1981), Household savngs and he rae of neres, Economc Record, 57, pp:

29 62. Ozcan, K. M., A. Gunay and S. Erac., (2003), Deermnans of prvae savngs behavor n Turkey, Appled Economcs, 35, pp: Pava, C and Sarwa, J., (2003), An emprcal sudy of prvae savngs n Brazl, Brazlan Journal of Polcal Economy, 23(1), pp: Paul R. Masson, T. Bayoum, and H. Same., (1993), Inernaonal evdence on he deermnans of prvae savngs, The World Bank Economcs Revew, 12 (3), pp: Pesaran, M. H., Pesaran, B. (1997). Workng wh Mcrof 4.0: Ineracve Economerc Analyss, Oxford: Oxford Unversy Press. 66. Pesaran, M. H., Shn, Y. and Smh, R. J., (2001), Bounds esng approaches o he analyss of level relaonshps, Journal of Appled Economercs, 16, pp: Sackey, F. G., (2011). Mcroeconomc deermnans of prvaes of prvae nward remances o households n Ghana: a case sudy. Journal of Economcs and Susanable Developmen, 2(2), Schmd-Hebbel, K. and Serven, L., (1999), The economcs of savngs and growh: heory, evdence and mplcaons for polcy. Cambrdge: Cambrdge Unversy Press. 69. Shahbaz, M., Shabbr, M. S., Bu, M.S., (2013). Effec of Fnancal Developmen on Agrculural Growh n Paksan: New Exensons from Bounds Tes o Level Relaonshps and Granger Causaly Tess. Inernaonal Journal of Socal Economcs, 8, 70. Shahbaz. M, Ahmad, N. and Wahd. N. M. A, (2010), Savngs Invesmen Correlaon and Capal Ouflow: The Case of Paksan Transon Sudes Revew, 17, pp: Taylor, A. M. (1996), Inernaonal capal mobly n hsory: he savng-nvesmen relaonshp. NBER Workng Paper Seres, No