THE FELDSTEIN-HORIOKA PUZZLE AND TWIN DEFICITS IN PAKISTAN

Transcription

1 THE FELDSTEIN-HORIOKA PUZZLE AND TWIN DEFICITS IN PAKISTAN Sumaira Saeed Inernaional Islamic Universiy, Islamabad, PAKISTAN. M. Arshad Khan Paisan Insiue of Developmen Economics, Islamabad, PAKISTAN ABSTRACTS The previous lieraure poins o a high correlaion beween domesic raes of invesmen and saving among OECD counries. Some ae his as evidence of limied financial inegraion in he indusrialized world. This paper ries o find any evidence of Feldsein-Horioa (F-H) puzzle in he presence of Twin deficis for Paisan using annual ime series daa for he period 1972 o The coinegraion resuls found evidence in favor of a high degree of capial mobiliy by rejecing he F-H puzzle in case of Paisan. So, F-H puzzle no exiss alhough Paisan is no perfecly inegraed in o he world economy. Keywords: Domesic raes of invesmen, OECD counries, Twin deficis INTRODUCTION The Feldsein and Horioa (1980) finding ha changes in domesic invesmen are very sensiive o changes in domesic savings for OECD counries launched a debae regarding he degree of financial inegraion and financial openness wihin he indusrialized world. The issue of he sensiiviy of domesic invesmen o changes in domesic savings has been acled mainly in he case of he indusrialized naions. However i is also relevan for developing counries. Twin deficis are defined as posiive long run relaionship beween he curren accoun and budge deficis. The win defici issue can also be lined o he degree of capial mobiliy across borders and o he Feldsein and Horioa (FH) (1980) puzzle. In fac, if domesic savings and invesmens are no highly relaed (due o free mobiliy) he BDEF and he CAD can be expeced o move ogeher. If he Ricardian equivalence does no hold, hen an increase in public borrowing (due o budge defici) is liely o decrease naional saving (boh domesic and foreign). For a given amoun of invesmen, his decrease in naional saving leads o an increase in he curren accoun defici. Hence, he perfec capial mobiliy resuls ino win deficis in a non-ricardian world. In oher words, if he FH puzzle holds here is no possibiliy of win deficis o coexis. In his sudy an aemp has been made o empirically es he validiy of Feldsein Horioa puzzle for Paisan and find any evidence in favor of a high degree of capial mobiliyusing annual ime series daa for he period 1972 o The res par of he paper is organized as follows. Secion 2 reviews he exising lieraure on Feldsein-Horioa puzzle. Theoreical and analyical framewor is presened in secion 3. Secion 4 gives daa descripion and economeric mehodology. Secion 5 discusses he esimaion resuls while secion 6 is devoed o conclusion. LITERATURE REVIEW In his secion, we have reviewed he some exising heoreical as well as empirical lieraure relaed o Feldsein-Horioa puzzle. There would be lile correlaion beween he saving and he domesic invesmen in ha counry wih perfec capial mobiliy across border. In conras, if porfolio preferences and insiuional rigidiies hinder he flow of long-erm capial across counries, hen an increase in domesic saving would be refleced primarily in addiional domesic invesmen. Feldsein and Horioa (1979) saisically invesigaed he wo views of inernaional capial mobiliy. The evidence on he relaion beween domesic invesmen and saving implied ha he ruh lies closer o he second view han o he firs. Feldsein (1982) reesed and confirmed he earlier findings of 525

2 Feldsein and Horioa (1980) ha susained increase in domesic saving rae induce approximaely an equal increase in domesic rae of invesmen. New esimaes for he pos-opec period ( ) implied ha each exra dollar of domesic saving increases domesic invesmen by approximaely 85 cens in a sample of 17 OECD counries. Jerry e al. (1998) invesigaed as o how economiss responded o he Feldsein Horioa (FH) view ha a high saving-invesmen associaion across OECD counries concealed low capial mobiliy. However, he debae wheher saving-invesmen co movemens are indicaion of capial mobiliy is sill inconclusive. The FH resul of a high cross-secion associaion beween saving and invesmen raes in OECD counries has remained remarably srong. Sachsida and Caeano (1999) corroboraed he evidence ha in he real side of he economy, Feldsein- Horioa es does no reflec capial mobiliy. The F-H coefficien does no mean capial mobiliy, bu jus a subsiuabiliy relaion beween exernal and domesic savings. Founas and Tsouis (2000) examined empirically he ineracions among he curren accoun, budge balances and he real ineres rae as i can provide more informaion abou he effecive degree of financial openness han simple saving-invesmen correlaions. They found ha in shor run here is some evidence in favor of he win deficis and curren accoun argeing hypoheses. Fidrmuc (2003) examined he evidence of win deficis and he F H puzzle for such economies and found a posiive long-run relaionship beween he win deficis in several counries. Invesmen in some EU counries is financed o a relaively high degree via he inernaional financial mares involving ha he F H puzzle is less significan in he EU counries. The FH puzzle is revisied by Caporalea e al. (2003) employing a variey of efficien coinegraion esimaors They concluded ha despie evidence supporing he FH resul, here appears o be considerable heerogeneiy in erms of he savings-invesmen associaion. Levy (2004) found ha long-run invesmen-saving correlaion follows direcly and his does no depend on he degree of inernaional capial mobiliy. Khedhiri andhebiri (2005) found evidence of a significan correlaion beween savings and invesmen, and hus no soluion o he F-H puzzle, despie he implicaions of inernaional capial mobiliy in six counries such as Algeria, Egyp, Morocco, Saudi Arabia, Syria and Tunisia. Rao e al. (2008) found he F-H puzzle exiss in a weaer form wih a reduced saving reenion coefficien he F-H equaion for 12 OECD counries. The FH resul of a high cross-secion associaion beween saving and invesmen raes in OECD counries has remained remarably srong. The response of economiss o he FH view regarding a high associaion beween saving-invesmen and implicily meaning low capial mobiliy across OECD counries is much discussed. The long-run invesmen-saving correlaion follows direcly and his does no depend on he degree of inernaional capial mobiliy. Unless he budge consrain is removed, he ime series of invesmen and saving show high coinegraion. However, he debae over he poin wheher saving-invesmen co movemens are indicaion of he degree of capial mobiliy is sill unresolved. TWIN DEFICITS AND THE FELDSTEIN-HORIOKA PUZZLE Wih perfec capial mobiliy worldwide, domesic saving is no necessarily relaed o domesic invesmen. However, here is srong empirical evidence ha domesic invesmen and naional saving are correlaed. Much of he evidence is based on cross-secion regressions of 16 OECD counries for he periods. This empirical finding is nown as he Feldsein-Horioa (1980) puzzle. According o his puzzle, he relaionship beween naional savings and domesic invesmen can be used as a measure of inernaional capial mobiliy. Under perfec mobiliy, he invesmen is no conrolled by domesic saving bu only by he accessibiliy of funds in he inernaional fully inegraed capial mare. Saving in each counry should reac o he worldwide opporuniies for invesmen while invesmen in ha counry should be financed by he inernaional pool of funds. The imporan facor of concern for invesors should be he rae of reurn. In a closed economy, domesic saving mus finance invesmen bu in case of open economy some of he invesmen may be financed by foreign 526

3 saving. Therefore, saving and invesmen could move independenly of each oher. Alernaively, he high I-S correlaion suggess ha capial migh no be fully mobile across borders. (Levy 2004) 10 Analyical Framewor for Twin Deficis and F-H Puzzle Feldsein and Horioa (1980) proposed an assessmen of he degree of capial mobiliy by measuring he correlaion beween invesmens and saving. They esimaed cross secion regressions of he form: ( I ) i = a + b ( S ) i + u (3.1) Y Y i In he above relaion, I is he raio of gross domesic invesmen o gross naional produc (GNP) Y and S is he raio of naional saving o GNP. For small counries, he value of slope parameer b Y should be close o zero if he inernaional capial mobiliy is perfec. In conras, he parameer should be large if capial is immobile. Such relaionships beween naional invesmen and saving shares of GDP, or heir varians such as firs differences have been essenial for many counries. These are called FH regressions where b is he FH coefficien represening he saving-invesmen associaion. The resuls for 16 OECD counries for he periods indicaed a very high saving-invesmen associaion alhough he capial mobiliy is near o perfec. This coinegraion exiss under fixed exchange rae and his relaion is an indicor of economic solvency (Rocha and Zerbini, 1998). Imporance of F-H Puzzle I is very criical o now he degree of capial mobiliy for financial decisions. Is imporance can be seen in following perspecives: The effec of fiscal policy crucially depends on he exen of capial mobiliy. The cos of adjusmen o exernal shocs ges reduced via he access of an economy o capial mares. Capial mobiliy deermines he rae a which incomes converge o equilibria. Perfec mobiliy is ofen assumed o hold in macroeconomic models (Levy, 2004). The win defici hypohesis has a clear lin wih he so-called F-H puzzle since he laer is concerned wih he source of financing he exernal defici. The causaliy beween exernal and fiscal balance also sresses he role of privae invesmen. Fidrmuc (2003) presened a regression model ha encompasses boh he win defici hypohesis and he Feldsein-Horioa puzzle. The model maes no difference beween ne expors and curren accoun balance. Rearranging he macro ideniy, one obains he relaion: X M = Y ( C + G ) I = S I (3.2) This ideniy implies ha he rade balance mus equal he difference beween naional savings (defined as oupu less oal consumpion) and invesmen (gross capial formaion). I provides a lin beween he exernal balance and saving-invesmen decisions. Hence, an increase in invesmen has a negaive impac on he exernal rade balance. On he oher hand, policies ha reduce (public or privae) consumpion are expeced o have a posiive impac on he exernal balance via increased naional savings. Naional savings can be subdivided ino privae (Sp) and public savings (Sg). The laer corresponds o he budge balance and given by he difference beween ax revenues and governmen expendiures. Privae savings are defined as he disposable income less privae consumpion. Thus he macro ideniy may be rewrien as under: 10 See levy (2004) for furher deails. 527

4 X M = Y T C ) + ( T G ) I = Sp, + Sg, ( I (3.3) This ideniy moivaes he esing of a long run relaionship among he curren accoun, he budge defici and oal invesmen. Therefore, gross capial formaion denoed by Invesmen as share of GDP, is included ino he relaionship beween he curren accoun (X-M) and he fiscal balance (T-G) X M = β ( I / Y ) + u (3.4) 1 + β 2 ( T G ) β 3 Here, he expeced sign of he coefficien for fiscal balance is posiive and ha of invesmen is negaive. I implies hahe curren accoun balance is worsened by budge defici and high invesmen. If an economy is perfecly inegraed ino he world mare, hen he coefficiens of boh he variables should equal o uniy. In his case, he budgeary as well as invesmen expendiures are financed by he world financial mare. However, i is shown in he sudy by Feldsein and Horioa (1980) ha a large porion of domesic invesmen is sill financed from domesic sources (savings). The above resul, generally referred o as he Feldsein Horioa puzzle, has been widely confirmed by subsequen inensive research. Following Fidrmuc (2003), we may specify he following regression model; wih variables (in small caps) expressed as a share of GDP: x m = 1 + β 2 g ) β 3 β ( inv + ε (3.5) Keeping in view he ideniy, a posiive coefficien is expeced for privae savings and he budge surplus (β 2, β 3 >0), and a negaive coefficien invesmen (β 4 <0). Hence, boh he budge defici and high invesmen are expeced o be ranslaed ino curren accoun defici. Moreover, he coefficiens of boh he variables should approach uniy if a counry is perfecly inegraed ino he world economy, implying ha boh he budge defici and invesmen are financed hrough he world capial mare. However, if he Feldsein-Horioa puzzle is here, he β 4 coefficien is significanly lower han uniy. Liewise, a negaive β 3 coefficien leads o he rejecion of he win defici hypohesis. 11 ECONOMETRIC METHODOLOGY AND DATA DESCRIPTION In his secion, we discuss he mehodology of research and he daa used in he analysis wih reference o Paisan. In his sudy, we employed he uni roo ess, Johansen co- inegraion echnique and he Error Correcion Model o aain our objecives. The main purpose of co-inegraion analysis is o verify he naure of long run relaionship beween a se of ime series variables. However, i is essenial o chec each ime-series for saionariy before saring he co-inegraion ess. In case he ime-series a hand is non saionary, hen he regression analysis carried ou in he usual manner may produce spurious resuls. So he uni roo ess are conduced firs o examine his propery of he ime-series. Co Inegraion Analysis If he variables of ineres share a common sochasic rend, hey are said o co-inegraed in he long run (Chrisensen, Nielsen 200). The concep of co-inegraion was firsly inroduced by Granger (1981) and furher formalized by Engel and Granger (1987) by inroducing a very simple mehod o chec he exisence of long run relaionship beween he variables. Alhough he EG es is very simple and convenien o implemen, bu i does no inimae as o which of he variable should be used as regressor and why. I may lead o conradicory resuls, especially when here are more han wo I(1) variables under consideraion (Pesaran and Pesaran, 1997). Therefore Johansen (1988; 1991) and Johansen and Juselius (1992) ess are employed in mulivariae analysis. We discuss he Johansen coinegraion procedure briefly: 11 These specificaions are used for Paisani daa and resemble hose adoped by Marinheiro (2008) for analysis of Egyp daa. Only he proxy variable of wealh is excluded from model. 528

5 Les assume ha we have hree variables Y, X and W which can all be endogenous. Each ime series variable has lins wih is pas values. The variables can wrien in he marix noaion as Z =[ Y, X, W ], where he vecor may be expressed as: Z µ = α 0 + α1 Z 1 + α2 Z α Z + (4.1) Defining =1-L, where L is he lag operaor, he above can be formulaed in a vecor error correcion model (VECM) as follows: Z = Π = α + 1 Z 1 + Π 2 Z Π i Z 1 i= 1 + α Z + Π Z + µ 1 + α where Π = 1 α α.... ), i = 1,2,..., 1 Z i ( 1 2 α The П marix is 3 3 due o he fac ha we assume hree variables in Z. The marix conains he informaion regarding he long run relaionships among he variables concerned. If he marix has a full ran, all he elemens in he vecor Z are saionary. On he oher hand, a zero ran indicaes he absence of any co-inegraion and he model reduces o VAR in he firs difference. In case he ran is posiive bu less han full, here exiss co-inegraion. To es wheher here exiss co-inegraion among he variables or oherwise, wo mehods (es saisics) are used ha deermine he ran of coinegraion space, due o Johansen (1988) and Johansen and Juselius (1990). The procedures are based on he proposiions abou eigenvalues. This mehod ess he null hypohesis ha ran of he marix П is r agains he alernaive hypohesis ha he ran is r+1. Thus, according o he null hypohesis, here are r co-inegraing vecors. The es saisics is based on he characerisic roo (Eigenvalues). To es how many of he numbers of he characerisics roos are significanly zero his es uses he following saisics: ( r, r + 1) = T ln (1 λr 1) max + + µ λ (4.3) The es saisic is based on he maximum eigenvalue (maximum eigenvalue saisic) The second mehod is based on a lielihood raio es abou he race of he marix (race saisic). This saisic is considers wheher he race is increased by adding more eigenvalues beyond he rh eigenvalue. The null hypohesis in his case is ha he number of coingraing vecors is less han or equal o r. This saisic is calculaed by: n λ ( r ) = T ln (1 λ ) (4.4) race i = r + 1 r + 1 The usual procedure is o wor downwards and sop a he value of r which is associaed wih a es saisic ha exceeds he displayed criical value. 12 Anoher imporan aspec is o selec he appropriae model regarding he deerminisic componens in he mulivariae sysem. I means ha wheher an inercep and/or a rend should ener eiher he shor run and long run model, or boh models. Granger Causaliy Tes If a pair of series is coinegraed hen here mus be Granger-causaliy in a leas one direcion, which reflecs he direcion of influence beween series. Theoreically, if he curren or lagged erms of a (4.2) 12 See Aserious Revised Ediion 2006, chap 17, p

6 ime-series variable, say X, deermine anoher ime-series variable, say Y, hen here exiss a Grangercausaliy relaionship beween X and Y, in which Y, is Granger caused by X. The Shor Run Analysis The shor run dynamics are examined using he Error Correcion Model (ECM). I explains changes in he dependen variable in erm of changes in he explanaory variables as well as deviaions from he long run relaionship beween he variables and is deerminans. The model follows from general o specific approach in economeric modeling which bes fis he given daa se. The co-inegraion of any wo variables implies ha here is some adjusmen process which prevens he error erm o ener in he long run relaionship. However, if he variables are co-inegraed over ime, hen an Error Correcion Model (ECM) is appropriae. According o Engle and Granger (1987), he co-inegraed variables have an ECM represenaion, which has he advanage of incorporaing boh he shor-run and long run relaionship beween he variables in he same regression. Daa and Variables Availabiliy of adequae and reliable daa is very imporan for consequenial analysis. The validiy of resuls depends on sufficien and consisen daa. We have done our umos effor for he collecion of reliable and consisen daa se for our research. We have used annual daa se of Paisan for he period This daa se is rerieved from differen daa sources. Mos of he daa is colleced from Federal Bureau of Saisic, Annual Repors of he Sae Ban, Paisan Economic Survey and Yearboo of Inernaional Financial Saisic (IFS) published by he IMF. The main variables used were Curren Accoun Balance, Budge Balance, and Invesmen ec. RESULTS AND INTERPRETATION As discussed in he previous secion, we have adoped a hree-sep procedure in esing he hree hypoheses under consideraion. Firs we apply he Augmened Dicey Fuller (ADF) uni roo es o chec he saionariy and order of inegraion of differen economic variables used in his sudy. Nex we resor o he Johansen s Maximum Liehood procedure o es for he long run co-inegraion among he variables. Finally, he Error Correcion Model is employed o see he causaliy beween he crucial variables. In he following lines, we discuss he findings and analyze he relevan resuls. Uni Roo Tess A he ouse, we es for he saionariy of all he ime series variables involved wih all he hree hypoheses of his sudy. we have used he variables in level and also in erm of heir raios o GDP for esing of he Twin Defici Hypohesis and he F-H puzzle. We use he ADF model o chec for he saionariy and he order of inegraion. The resuls are projeced in Table 5.1 below. Table 5.1: The ADF Tes y = 1 + i= 1 = µ + δ y β y + ε Variable Levels Firs Difference Conclusion ADF T-es ADF T-es I(1) GDP raios Curren Accoun defici * I(1) Budge balance * I(1) Budge Defici * I(1) Curren Accoun balance * I(1) Invesmen ** I(1) Noe: he Macinnon criical values of significance a he 1%, 5% and 10% are , and respecively. The superscrips *, **, *** shows significance a 1%, 5% and 10% respecively. i i 530

7 I is eviden from he es ha he series have become saionary by aing he firs difference, which implies he presence of uni roos in he original daa. In oher words, all he variables are nonsaionary a level and saionary a firs difference or hey are inegraed of same order I (1). This validaes our proposiions ha he variables concerned are indeed co-inegraed and a long run relaionship holds among hem. Wih his informaion a hand, we proceed furher o examine he naure of long run and shor run relaionships among he variable. The Feldsein and Horioa Puzzle Finally, we concenrae on exploraion of he F-H puzzle wih reference o Paisan economy by following he model proposed by Fidrmuc (2003). The model acles he win defici issue and Feldsien Horioa puzzle simulaneously. To be specific, we are going o esimae he model as discussed earlier and given in equaion 3.14, reproduced below: x m = 1 + β 2 g ) β 3 β ( inv + ε (3.14) Esimaed resuls of Fidrmuc s (2003) model specificaion are given below: x m = ( g ) inv The TD hypohesis implies a posiive coefficien for he budge balance. The empirical resuls also reveal a posiive coefficien which leads o accepance of he TD hypohesis. These resuls indicae no rejecion of he TD hypohesis and rejecion of F H puzzle for Paisan. The coefficien of (inv) is no equal o one (uniy) so his implies ha here is perfec capial mobiliy which means ha here is no F- H puzzle presen in Paisan. The co-inegraion resuls are repored in Table 5.2. Table 5.2 Johansen s Maximum Lielihood Tes for Twin Deficis & F-H Puzzle (The budge defici, curren accoun defici and invesmen are expressed as %GDP) Λ-Trace Eigen-value H 0 H 1 Trace 5% Criical Value Λ- Max H 0 H 1 Max 5% Criical Value R 0 r> R 0 r> R 1 r> R 1 r> R 2 r> R 2 r> Noe: The lag lengh of one is used in he VAR. Criical values for race and maximum lielihood ess are due o Oserwald-Lenum (1992). The esimaion is carried by assuming only an inercep and no rend in he coinegraion equaion. The above resuls reveal ha he null hypohesis of no co-inegraion among he ime series variables is rejeced a 5% level of significance by he eigen-value max and race saisics. The resuls spea of he exisence of one co-inegraing vecor and confirm he long run relaionship among rade balance, budge defici and capial formaion in he economy. There is a large degree of capial mobiliy as FH puzzle does no hold in Paisan. Bu Paisan is no perfecly inegraed ino he world economy. Since increase in he governmen s ne borrowing requiremens could be financed by he exernal financing. 531

8 CONCLUSIONS The objecive of his paper was o find any evidence any evidence of Feldsein-Horioa (F-H) puzzle in he presence of Twin deficis for Paisan. The esimaed empirical resuls showed no evidence in favor of he Feldsein-Horioa puzzle for Paisan. This is because our economy is no perfecly inegraed ino he world economy. Moreover he degree of inernaional capial mobiliy has never been perfec. Despie hese facors, he domesic invesmen in Paisan (paricularly in he public secor) has never depended on domesic saving or inernal borrowing only. Foreign assisance has always played an imporan role in financing he naional developmen programs. Thus he hisorical relaionship beween domesic invesmen and saving is very poor in Paisan. As such, he empirical resuls do suppor he validiy of win defici hypohesis bu rejec he F-H puzzle in case of Paisan. REFERENCES Caporale, G.M., Panopoulou, E. and Piis, N., (2003). The Feldsein-Horioa puzzle revisied: a Mone Carlo sudy, Cenre for Inernaional Capial Mares, London Meropolian Universiy, and London Coaley, Jerry, Farida Kulasi & Ron Smih, (1998), The Feldsein Horioa Puzzle and Capial Mobiliy: A Review, Inernaional Journal of Finance and Economics, 3: Feldsein, M. & C. Horioa. (1980), Domesic savings and inernaional capial flows, The Economic Journal, 90, Feldsein, M. (1983), Domesic saving and inernaional capial movemen in he long run and he shor run, European Economic Review; 21: Founas, S., & C. Tsouis, (2000). Twin deficis, real ineres raes, and inernaional capial mobiliy, Naional Universiy of Ireland (Galway), Woring Paper No. 49. Sachsida A. and M. A.R Caeano, (2000), The Feldsein-Horioa puzzle revisied, Economics Leers; 68: Levy, D. (2003), Is he Feldsein-Horioa puzzle really a puzzle?, Advances in Inernaional Economics and Finance. London: Kluwer Academic Publishers. Khedhiri, Sami, &Rimel, Hebiri, (2005), An Empirical Invesigaion of he Feldsein-Horioa puzzle in MENA Counries, Journal of Economic Lieraure. 532