Testing Twin Deficits and Saving-Investment Nexus in Turkey

Similar documents
Testing Twin Deficits and Saving-Investment exus in Turkey [ FIRST DRAFT] Abstract

NPTEL Project. Econometric Modelling. Module23: Granger Causality Test. Lecture35: Granger Causality Test. Vinod Gupta School of Management

Department of Economics University of Toronto

Fall 2009 Social Sciences 7418 University of Wisconsin-Madison. Problem Set 2 Answers (4) (6) di = D (10)

Data Collection Definitions of Variables - Conceptualize vs Operationalize Sample Selection Criteria Source of Data Consistency of Data

RELATIONSHIP BETWEEN VOLATILITY AND TRADING VOLUME: THE CASE OF HSI STOCK RETURNS DATA

Oil price volatility and real effective exchange rate: the case of Thailand

Applied Econometrics and International Development Vol- 8-2 (2008)

Energy Consumption- Growth Nexus in Saarc Countries: Using Cointegration and Error Correction Model

F-Tests and Analysis of Variance (ANOVA) in the Simple Linear Regression Model. 1. Introduction

V.Abramov - FURTHER ANALYSIS OF CONFIDENCE INTERVALS FOR LARGE CLIENT/SERVER COMPUTER NETWORKS

Stock Market Development And Economic Growth

John Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany

Capital Flow Volatility and Exchange Rates: The Case of India. Pami Dua and Partha Sen 1, 2

2. SPATIALLY LAGGED DEPENDENT VARIABLES

Econ107 Applied Econometrics Topic 5: Specification: Choosing Independent Variables (Studenmund, Chapter 6)

UNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 2017 EXAMINATION

Analysis And Evaluation of Econometric Time Series Models: Dynamic Transfer Function Approach

January Examinations 2012

Time Scale Evaluation of Economic Forecasts

The role of monetary policy in managing the euro dollar exchange rate

Export-Led Growth Hypothesis: Evidence from Agricultural Exports in Tanzania

New M-Estimator Objective Function. in Simultaneous Equations Model. (A Comparative Study)

Long-Run Relationship and Causality between Foreign Direct Investment and Growth: Evidence from Ten African Countries

Graduate Macroeconomics 2 Problem set 5. - Solutions

THE FORECASTING ABILITY OF A COINTEGRATED VAR DEMAND SYSTEM WITH ENDOGENOUS VS. EXOGENOUS EXPENDITURE VARIABLE

( t) Outline of program: BGC1: Survival and event history analysis Oslo, March-May Recapitulation. The additive regression model

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 4

Political Economy of Institutions and Development: Problem Set 2 Due Date: Thursday, March 15, 2019.

Volume 31, Issue 1. Are exports and imports cointegrated in India and China? An empirical analysis

In the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!") i+1,q - [(!

Robustness Experiments with Two Variance Components

Solution in semi infinite diffusion couples (error function analysis)

Journal of Chemical and Pharmaceutical Research, 2014, 6(5): Research Article

THEORETICAL AUTOCORRELATIONS. ) if often denoted by γ. Note that

Testing the Null Hypothesis of no Cointegration. against Seasonal Fractional Cointegration

TSS = SST + SSE An orthogonal partition of the total SS

On One Analytic Method of. Constructing Program Controls

The Relationship between Savings and GDP in Iran: An ARDL Approach for the case of Iran

Let s treat the problem of the response of a system to an applied external force. Again,

Variants of Pegasos. December 11, 2009

US Monetary Policy and the G7 House Business Cycle: FIML Markov Switching Approach

An Investigation of the Long-Run and Causal Relationships between Economy Performance, Investment and Port Sector Productivity in Cote d Ivoire

Advanced time-series analysis (University of Lund, Economic History Department)

Economic Growth, Export, and External Debt Causality: The Case of Asian Countries

A Nonlinear Panel Unit Root Test under Cross Section Dependence

Survival Analysis and Reliability. A Note on the Mean Residual Life Function of a Parallel System

ACEI working paper series RETRANSFORMATION BIAS IN THE ADJACENT ART PRICE INDEX

Real Exchange Rates In Developing Countries: Are Balassa-Samuelson Effects Present?

5th International Conference on Advanced Design and Manufacturing Engineering (ICADME 2015)

Bayesian Inference of the GARCH model with Rational Errors

Volume 30, Issue 4. Abd Halim Ahmad Universiti Utara Malaysia

International Parity Relations between Poland and Germany: A Cointegrated VAR Approach *

GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS. Youngwoo Ahn and Kitae Kim

THE PREDICTION OF COMPETITIVE ENVIRONMENT IN BUSINESS

Analysing the Relationship between New Housing Supply and Residential Construction Costs with the Regional Heterogeneities

Notes on the stability of dynamic systems and the use of Eigen Values.

Linear Response Theory: The connection between QFT and experiments

Economics Discussion Paper

Dynamic Team Decision Theory. EECS 558 Project Shrutivandana Sharma and David Shuman December 10, 2005

The Impact of SGX MSCI Taiwan Index Futures on the Volatility. of the Taiwan Stock Market: An EGARCH Approach

Online Supplement for Dynamic Multi-Technology. Production-Inventory Problem with Emissions Trading

Additive Outliers (AO) and Innovative Outliers (IO) in GARCH (1, 1) Processes

Appendix H: Rarefaction and extrapolation of Hill numbers for incidence data

Multivariate GARCH modeling analysis of unexpected U.S. D, Yen and Euro-dollar to Reminibi volatility spillover to stock markets.

Application of Vector Error Correction Model (VECM) and Impulse Response Function for Analysis Data Index of Farmers Terms of Trade

HEAT CONDUCTION PROBLEM IN A TWO-LAYERED HOLLOW CYLINDER BY USING THE GREEN S FUNCTION METHOD

Relative Efficiency and Productivity Dynamics of the Metalware Industry in Hanoi

Panel Data Regression Models

Robustness of DEWMA versus EWMA Control Charts to Non-Normal Processes

[Link to MIT-Lab 6P.1 goes here.] After completing the lab, fill in the following blanks: Numerical. Simulation s Calculations

Oligopoly with exhaustible resource input

CH.3. COMPATIBILITY EQUATIONS. Continuum Mechanics Course (MMC) - ETSECCPB - UPC

Cubic Bezier Homotopy Function for Solving Exponential Equations

Relative controllability of nonlinear systems with delays in control

ABSTRACT KEYWORDS. Bonus-malus systems, frequency component, severity component. 1. INTRODUCTION

FI 3103 Quantum Physics

CHAPTER 5: MULTIVARIATE METHODS

Comparison of Differences between Power Means 1

Estimations of the Price Transmission and Market Power in Canola Export Market: Implication to Canola Import of Japan

Critical Values for IPS Panel Unit Root Tests: A Response Surface Analysis Rajaguru, Gulasekaran

1 Constant Real Rate C 1

Time-interval analysis of β decay. V. Horvat and J. C. Hardy

Endogeneity. Is the term given to the situation when one or more of the regressors in the model are correlated with the error term such that

Existence and Uniqueness Results for Random Impulsive Integro-Differential Equation

By By Yoann BOURGEOIS and Marc MINKO

10. A.C CIRCUITS. Theoretically current grows to maximum value after infinite time. But practically it grows to maximum after 5τ. Decay of current :

An introduction to Support Vector Machine

Lecture Notes 4. Univariate Forecasting and the Time Series Properties of Dynamic Economic Models

Forecasting the Convergence State of per Capital Income in Vietnam

PhD/MA Econometrics Examination. January, 2019

On the linkages between stock prices and exchange rates: Evidence from the

Chapter 9: Factor pricing models. Asset Pricing Zheng Zhenlong

Lecture 6: Learning for Control (Generalised Linear Regression)

Approximate Analytic Solution of (2+1) - Dimensional Zakharov-Kuznetsov(Zk) Equations Using Homotopy

[ ] 2. [ ]3 + (Δx i + Δx i 1 ) / 2. Δx i-1 Δx i Δx i+1. TPG4160 Reservoir Simulation 2018 Lecture note 3. page 1 of 5

CHAPTER 10: LINEAR DISCRIMINATION

2 Aggregate demand in partial equilibrium static framework

Estimation of Cost and. Albert Banal-Estanol

US Monetary Policy and the G7 House Business Cycle: FIML Markov Switching Approach

Transcription:

MPRA Munch Personal RePEc Archve Tesng Twn Defcs and Savng-Invesmen Nexus n Turkey Ferda HALICIOGLU and Kasm EREN Isanbul Medenye Unversy Deparmen of Economcs, Yldz Techncal Unversy Deparmen of Economcs 017 Onlne a hps://mpra.ub.un-muenchen.de/8359/ MPRA Paper No. 8359, posed 30 December 017 04:7 UTC

Tesng Twn Defcs and Savng-Invesmen Nexus n Turkey Absrac Ths paper provdes furher evdence on he valdy of wn defcs and he Feldsen- Horoka hypoheses for Turkey durng he perod of 1987-004 usng bounds esng approach o conegraon. In order o explan he man deermnans of he curren accoun defcs n he long-run, he fscal balance and he domesc nvesmens are used n an economerc model.the conegraon ess ndcae he presence of a long-run relaonshp beween he curren accoun and budge defcs as well as he domesc nvesmens durng he esmaon perod. As a resul, s concluded ha he wn defcs hypohess and he Feldsen-Horoka puzzle are presen and Turkey appeared o be negraed no he world capal marke wh a low degree of capal mobly as less han 1/5 of s domesc nvesmen s fnanced hrough exernal funds. The augmened Granger-causaly ess sugges no causaly beween he curren accoun and budge defcs, boh n he shor-run and he long-run. The pos-sample varance decomposons sugges ha he domesc nvesmens are he man cause of curren defcs n he long-run. The paper also dscusses he polcy mplcaons of he emprcal resuls. Keywords: Twn defcs, Feldsen-Horoka hypohess, conegraon, Turkey JEL Classfcaons: C, F3, F36 Ferda HALICIOĞLU Deparmen of Economcs, Isanbul Medenye Unversy, Kadkoy, 34700 Turkey ferda.halcoglu@medenye.edu.r Co-auhor Kasım EREN Deparmen of Economcs, Yldz Techncal Unversy, Isanbul, 34349 Turkey erenk@yldz.edu.r 1

1. Inroducon In recen economc leraure, a new research neres has emerged, suggesng ha boh he wn defcs and Feldsen Horoka hypoheses may be used o explan he long-run deermnans of curren accoun mbalances. Fdrmuc (003) has ponoreed hs new drecon of research provdng frs heorecal underpnnngs and emprcal evdence from some EU (European Unon) counres. Accordng o hs new research area, he Feldsen-Horoka and he wn defcs hypoheses could be ncorporaed and esmaed emprcally n a sngle equaon n order o provde plausble exlanaons for he long-run deermnans of curren accoun mbalances. Feldsen-Horoka (1980) presens ha changes n domesc nvesmen are very sensve o changes n domesc savngs. Thus, here s a posve long-run relaonshp beween he rao of domesc nvesmen-gross domesc produc and he rao of domesc savngs gross domesc ncome. Ths smple sascal assocaon s also regarded as he exsence of he degree of nernaonal capal mobly. Meanwhle he wn defcs hypohess s defned as a posve long-run relaonshp beween he curren accoun and fscal balance. Fdrmuc (003) esed boh hypoheses for 1 OECD counres durng 1970-001 and concluded ha he wn defcs and Feldsen-Horoka hypoheses exsed only n he case of Hungary and Poland. The sudy of Fdrmuc (003) recenly has been adoped by a few researchers. Marnhero (008) repors a hgh degree of capal mobly and a rejecon of wn defcs hypohess durng he perod of 1974-004 for Egyp. Alnas and Taban (011) argues ha he wn defcs and Feldsen Horoka hypoheses are vald n he case of Turksh daa for he perod of 1974-007. Usng he annual daa of 1976-010 for Paksan, Khan and Saeed (01) confrms he exsence of he wn defcs and he Feldsen - Horoka hypoheses. Bagher e al. (01) presens a weak suppor for he hypoheses n Iran

over he perod of 1971-007. Erdem e al. (016) fnds evdence for he hypoheses n he case of Turkey usng annual daa of 1960-014 1. Turkey has been mplemenng a se of economc reforms o ransform s mpor-subsung economc srucure o marke economy srucure snce 1980. The lberalzaon of money and foregn exchange markes were relavely faser and effecve n comparson o reducng he sze of governmen n economy. To hs end, followng he neres rae lberalzaon n 1987, he capal accoun was also lberalzed n 1989 whch paved he way for foregn savngs o conrbue o he domesc savngs gap. Durng he decade of he 1990s and he begnng of he 000s, Turkey has faced boh severe curren accoun and budge defcs a he same me. Alhough he budge defcs were aken under conrol followng he ausery programme whch was pu no mplenaon n 001 followng he economc crses of 1999 and 001, he curren accoun defcs are sll seen o be runnng que hgh n he las decade. The objecve of hs paper s as follows: o es emprcally he valdy of he wn defcs and he Feldsen Horoka hypoheses usng Turksh me seres daa durng 1987-004 and mplemenng a dynamc conegraon procedure as well as esablshng he casualy ess beween he varables boh whn and ou of sample perods. The am of hs sudy s o nvesgae he man deermnans of he curren accoun mbalances n Turkey and provde some polcy gudelnes for he polcy makers. Ths sudy dffers from Alnas and Taban (011) and Erdem e al. (016) on wo dfferen accouns; frsly, hese sudes cover some perods before1989 n whch Turkey dd no have fnancal accoun lberalzaons. Therefore, usng exernal savngs for domesc nvesmens was no a fnancal opon and he auhors dd no ake hs no accoun n her economerc model. Moreover, hs sudy exends he causaly analyss beyond he ou of sample perods. 1 As a passng noe, here are numerous sudes emprcal relang o wn defcs hypohess n Turkey such as Ay e al (004), Yay and Tasan (007), Yaprakl (010), Bola e al. (011), Kayhan e al. (013), Calk e al. (015) wh nconclusve resuls for he hypohess. 3

The remander of hs paper s organzed as follows: he nex secon presens a bref concepual framework and economerc mehodology. Secon 3 dscusses he emprcal resuls and he las secon concludes.. Emprcal Model and Mehodology Fdrmuc (003) ses ou he relaonshp beween budge and curren accoun balances usng he naonal accouns as follows: Y = C + I + G + X M (1) where Y s ncome, C s prvae consumpon, I s prvae nvesmen, G s publc consumpon, X s expors and M s mpors. Eq.(1) can be arranged as folllows: X M = Y C G I = S I () Eq. () suggess ha he exernal accoun has o equal he dfference of naonal savngs and nvesmens. Thus, he curren accoun s drecly relaed o savng and nvesmen n he economy, mplyng ha as nvesmen s encouraged as a resul, exernal accoun wll be negavely affeced. However, a conracon n prvae or publc consumpon wll have a posve mpac on curren accoun balance as hey ncrease naonal savngs. As we separae publc g S from prvae savngs p S, hen publc savngs are relaed o fscal budge whch are defned as (T - G) n whch T s ax ncome. Smlarly, prvae savngs are defned as S p = Y T C. Thus, we wre ou Eq. () as follows: 4

X M = ( Y T C ) + ( T G ) I = S + S I (3) p g Eq. (3) suggess ha f prvae savngs equal nvesmen, he curren accoun and fscal budge are drecly nerrelaed or wnned. The concep of wn defcs hypohess fnds s heorecal bass n he Mundell-Flemng open economy model and he Keynesan absorpon heory. The former approach argues ha an ncrease n budge defcs wll cause an ncrease n domesc neres rae above he world rae whch leads o capal nflows and exchange rae apprecaon and, n urn, leaves he counry s curren accoun n defcs. The laer heory demonsraes ha a rse n budge defcs nduces domesc absorpon whch resuls n an ncrease n mpors and a decrease n expors; hus, he curren accoun defcs occurs. On he oher hand, he Rcardan Equvalence Hypohess (REH) of Barro (1974) dcaes ha he curren accoun and budge balance are no relaed, mplyng ha he governmen ax polcy has no mpac on prvae spendng and naonal savngs. Accordng o Feldsen and Horoka (1980) n a world of perfec capal mobly, he fnancng of domesc nvesmen s no relaed o domesc savngs, f he domescs savngs and nvesmens are no correlaed mplyng hgh capal mobly. However, a number of emprcal research sudes have found reverse resuls leadng o he concluson of a puzzle ha a hgh poron of domesc nvesmen s sll fnanced from domesc savngs, especally n developed counres. On he oher hand, some researchers, such as Coakley e al. (1996) and Sachsda and Caeano (000), argue ha he Feldsen-Horoka hypohess does no necessarly mply capal mobly and should be regarded as an ndcaon of subsuon beween exernal and domesc savngs. Therefore, seems ha he debae over wheher savng-nvesmen co-movemen as an ndcaon of capal mobly s sll unresolved. 5

Eq. (3) also mples ha here s a long-run relaonshp beween he curren accoun, he budge defc and oal nvesmen. Therefore, s possble o esmae Eq. (3) by a regresson model, as proposed by Fdrmuc (003). x α ( + ε (4) m = 0 + α1 g ) α The lower case leers n Eq. (4) ndcae ha he varables are expressed as a share of GDP n whch (x m) sands for he curren accoun, ( g) represens he governmen budge balance and he nvesmen rao s defned as. A posve sgn for he coeffcen of fscal balance (.e. α 1 > 0) and a negave sgn for he coeffcen of nvesmen (.e. α < 0) are expeced ndcang ha a budge defc and hgh nvesmen deerorae he curren accoun. If boh slope coeffcens are equal o one, hen s assumed ha he counry s perfecly negraed no he world economy, mplyng ha boh budgeary and nvesmen expendures are fnanced on he world fnancal marke. If he coeffcen of budge balance s posve, resuls n a wn defc whch also mples non exsence of he REH. On he oher hand, f he coeffcen of nvesmen s relavely close o uny or hgher han uny, ndcaes he valdy of he Feldsen Horoka hypohess and, f he coeffcen of nvesmen s relavely close o zero, mples he exsence of he Feldsen Horoka puzzle. The shor-run dynamc adjusmen process of he long-run relaonshp n Eq. (4) may provde useful polcy recommendaons. I s possble o ncorporae he shor-run dynamcs no Eq. (4) by expressng n an error-correcon model as suggesed n Pesaran e al. (001). ( x m) β ( x m) 4 1 n1 1 = 1 = 0 = 0 + β ( g) 1 + β 6 1 + v n = β0+ β ( x m) + β ( g) + β3 5 n3 + (5) 6

Ths approach, also known as auoregressve-dsrbued lag (ARDL), provdes he shor-run and long-run esmaes smulaneously. Shor-run effecs are refleced by he esmaes of he coeffcens aached o all frs-dfferenced varables. The long-run effecs of he explanaory varables on he dependen varable are obaned by he esmaes of β 5 -β 6 ha are normalzed on β 4. The ncluson of he lagged-level varables n Eq. (5) s verfed hrough he bounds esng procedure, whch s based on he Fsher (F) or Wald (W)-sascs. Ths procedure s consdered as he frs sage of he ARDL conegraon mehod. Accordngly, a jon sgnfcance es ha mples no conegraon hypohess, (H 0 : all β4 o β = 6 0 ), agans he alernave hypohess, (H 1 : a leas one of β4 o β 6 0 ) should be performed for Eq. (5). The F/W es used for hs procedure has a non-sandard dsrbuon. Thus, Pesaran e al. (001) compue wo ses of crcal values for a gven sgnfcance level wh and whou a me rend. One se assumes ha all varables are I(0) and he oher se assumes hey are all I(1). If he compued F/W-sasc exceeds he upper crcal bounds value, hen he H 0 s rejeced, mplyng conegraon. In order o deermne wheher he adjusmen of varables s oward her long-run equlbrum values, esmaes of β 4 -β 6 are used o consruc an error-correcon erm (EC). Then lagged-level varables n Eq. (5) are replaced by EC -1 formng a modfed verson of Eq. (5) as follows: n1 n n3 0 + β1 ( x m) + β ( g) + β3 + λec 1 = 1 = 0 = 0 ( x m) = β + µ (6) Eq. (6) s re-esmaed one more me usng he same lags prevously. A negave and sascally sgnfcan esmaon of λ no only represens he speed of adjusmen bu also Dfferen applcaons of he ARDL approach o conegraon can be found n he followng sudes: Bahman e al. (017), Bahman e al. (016), Halcoglu and Keenc (016), Durmaz (015), Tayeb (014), Halcoglu and Karaas (013), Halcoglu (013), Pachs (01), Cak e al. (011), Dell Anno and Halcoglu (010), Halcoglu (007). 7

8 provdes an alernave means of supporng conegraon beween he varables. Pesaran e al. (001) conegraon approach has some mehodologcal advanages n comparson o oher sngle conegraon procedures. Reasons for he ARDL are: ) endogeney problems and nably o es hypoheses on he esmaed coeffcens n he long-run assocaed wh he Engle-Granger (1987) mehod are avoded; ) he long and shor-run coeffcens of he model n queson are esmaed smulaneously; ) he ARDL approach o esng for he exsence of a long-run relaonshp beween he varables n levels s applcable rrespecve of wheher he underlyng regressors are purely saonary I(0), purely non-saonary I(1), or muually conegraed; v) he small sample properes of he bounds esng approach are far superor o ha of mulvarae conegraon, as argued n Narayan (005). The Granger represenaon heorem suggess ha here wll be Granger causaly n a leas one drecon f here exss a conegraon relaonshp among he varables n equaon (1), provdng ha hey are negraed order of one. Engle and Granger (1987) cauon ha he Granger causaly es, whch s conduced n he frs-dfferenced varables by means of a VAR, wll be msleadng n he presence of conegraon. Therefore, an ncluson of an addonal varable o he VAR sysem, such as he error correcon erm would help us o capure he long-run relaonshp. To hs end, an augmened form of he Granger causaly es nvolvng he error correcon erm s formulaed n a mulvarae ph order vecor error correcon model. [ ] + + + = = p EC g m x L g m x L 3 1 1 3 1 33 3 31 3 1 13 1 11 1 3 1 ) ( ) ( ) (1 ) ( ) ( ) (1 ω ω ω δ δ δ φ φ φ φ φ φ φ φ φ θ θ θ (7) ) 1 ( L s he lag operaor. EC -1 s he error correcon erm, whch s obaned from he longrun relaonshp descrbed n Eq. (1), and s no ncluded n Eq. (7) f one fnds no conegraon amongs he vecor n queson. The Granger causaly es may be appled o

Eq. (7) as follows: ) by checkng sascal sgnfcance of he lagged dfferences of he varables for each vecor; hs s a measure of shor-run causaly; and ) by examnng sascal sgnfcance of he error-correcon erm for he vecor ha here exss a long-run relaonshp. As a passng noe, one should reveal ha Eq. (6) and (7) do no represen compeng error-correcon models because Eq. (6) may resul n dfferen lag srucures on each regressors a he acual esmaon sage; see Pesaran e al. (001) for deals and s mahemacal dervaon. All error-correcon vecors n equaon (7) are esmaed wh he same lag srucure ha s deermned n unresrced VAR framework. Esablshng Granger causaly s resrced o essenally whn sample ess, whch are useful n dsngushng he plausble Granger exogeney or endogeny of he dependen varable n he sample perod, bu are unable o deduce he degree of exogeny of he varables he beyond he sample perod. To examne hs ssue, he decomposon of varance of he varables may be used. The varance decomposons (VDCs) measure he percenage of a varable s forecas error varance ha occurs as he resul of a shock (or an nnovaon) from a varable n he sysem. Sms (1980) noes ha f a varable s ruly exogenous wh respec o he oher varables n he sysem, own nnovaons wll explan all of s forecas error varance (.e., almos 100%). By lookng a VDCs polcy makers gaher addonal nsgh as o wha percenage (of he forecas error varance) of each varable s explaned by s deermnan. 3. Resuls Annual daa over he perod 1987-004 were used o esmae Eq. (5) and (6) by he ARDL conegraon procedure of Pesaran e al. (001) 3. Varable defnon and sources of daa are ced n he Appendx. 3 The daa span for hs sudy was nally seleced as 1987-01 and bu afer several economerc rals, he mos meanngful resuls confrmng he exsence of he wn-defcs hypohess were obaned from he perod of 1987-004. 9

Correlaon marx and graphs of he varables n Eq. (4) are provded below n Table 1 and Graph 1, respecvely n order o presen he prelmnary relaonshps beween he varables. Table 1. Correlaon Marx (x-m) (-g) (x-m) 1.000 0.09-0.145 (-g) 0.09 1.000 0.701-0.145 0.701 1.000 Consderng he correlaon marx n Table 1, s clear ha here s a weak posve correlaon beween he curren accoun and fscal balance. I s also seen ha he here exss a negave relaonshp beween he curren accoun balance and nvesmen level. These smple sascal relaon dd no hold n he oher perods. I s very lkely ha because of hese pror resuls, when he ARDL conegraon was mplemened n oher perods such as 1980-01 and 1987-01, we could no fnd any long-run relaonshps beween he varables apar from he perod of 1987-004. Thus, our economerc resuls are lmed wh only he seleced esmaon perod. 10

Graph 1: Plo of he varables 30 0 10 0-10 -0 1987 199 1997 00 004 CABY BDY INVY In Graph 1, CABY sands for curren defcs whch s n he mddle, BDY represens he budge defcs whch s a he boom, and INVY sands for he nvesmen whch s a he op. All varables are refleced as a percenage of GDP. To mplemen he Pesaran e al. (001) procedure, one has o ensure ha none of he explanaory varables n equaon (1) s above I(1). Three ess were used o es un roos n he varables: Augmened Dckey-Fuller (henceforh, ADF) (1979, 1981), Phllps-Perron (henceforh, PP) (1988), and Ello-Rohenberg-Sock (henceforh, ERS) (1996). Un roo ess resuls are dsplayed n Table o warran mplemenng he ARDL approach o conegraon as he varables are n he combnaon of I(0) and I(1). 11

Table. Un roo resuls Varables ADF PP ERS (x m) 4.99 * 5.68 * 4.3 * ( g) 3.08.57 3.14 3.71 * 4.6 * 1.88 (x m) 5.98 * 3.95 * 3.95 * ( g) 4.83 * 4.83 * 4.83 * 5.70 * 5.71 * 5.71 * Noes: The sample level un roo regressons nclude a consan and a rend. The dfferenced level un roo regressons are wh a consan and whou a rend. All es sascs are expressed n absolue erms for convenence. Rejecon of un roo hypohess s ndcaed wh an asersk. sands for frs dfference. Crcal values are smulaed for sample sze. Ths s done auomacally by he economerc sofware Table 3. The resuls of F and W ess for conegraon. Panel A: The assumed long-run relaonshp: F / W (( x m) ( g), )) F-sasc 95% LB 95% UB 90% LB 90% UB 1.70 4.77 6.19 3.71 4.91 W-sasc 65.1 14.33 18.57 11.14 14.73 If he es sasc les beween he bounds, he es s nconclusve. If s above he upper bound (UB), he null hypohess of no level effec s rejeced. If s he below he lower bound (LB), he null hypohess of no level effec canno be rejeced. Crcal values are smulaed for sample sze. F and W esng crcal values are smulaed for he small sample sze. Ths s done auomacally by he economerc sofware On esablshng a long-run conegraon relaonshp amongs he varables of Eq. (4), a wo-sep procedure o esmae he ARDL model was carred ou. Frs, n search of he opmal lag lengh of he dfferenced varables of he shor-run coeffcens, Schwarz Bayesan Creron (SBC) was ulzed and n he second sep, he ARDL model was esmaed. The resuls of SBC based ARDL model s dsplayed n Panel A, B, and C of Table 4. The resuls of long-run coeffcens are presened n Panel A of Table 4, whereas he shor-run esmaes are repored n Panel B of Table 4. Fnally, Panel C of Table 4 demonsraes he shor-run dagnosc es resuls. The overall regresson resuls are sasfacory n erms of dagnosc ess. The shor-run dagnoscs obaned from he esmaon of Eq. (5) sugges ha he 1

esmaed model s free from a seres of economerc problems such as seral correlaon, funconal form, normaly, and heeroscedascy. Table 4. ARDL conegraon resuls. Panel A: Esmaed long-run coeffcens Dependen varable: (x m) Regressor Coeffcen Sandard error T-rao ( g) 0.55 * 0.095.680-0.165 ** 0.084 1.998 Consan 4.446 **.354 1.888 Panel B: Error correcon represenaon resuls. Dependen varable: (x m) Regressor Coeffcen Sandard error T-rao (x m) -1 0.670 * 0.17 5.68 ( g) 0.040 0.100 0.407 ( g) -1-0.48 * 0.103.403-0.59 * 0.10 5.756-1 0.33 * 0.091.54 EC -1-1.55 * 0.05 7.414 Panel C: Dagnosc es resuls. R 0.9 F-sasc 34.8 * χ SC(1) 0.034 χ FF (1) 4.07 RSS 5.4 DW-sasc 1.89 χ () 1.68 χ (1) 0.90 N H *, **, and, *** ndcae, 1%, 5%, and 10% sgnfcance levels respecvely. RSS sands for resdual sum of squares. T-raos are n absolue values. χ SC, χ FF, χ N, and χ H are Lagrange mulpler sascs for ess of resdual correlaon, funconal form ms-specfcaon, non-normal errors and heeroskedascy, respecvely. These sascs are dsrbued as Ch-squared varaes wh degrees of freedom n parenheses. The crcal values for χ (1) = 3. 84 and χ () = 5. 99 are a 5% sgnfcance level. The overall resuls confrm he exsence of he wn defcs phenomenon for Turkey durng he esmaon perod, snce he coeffcen of he fscal accoun s posve and sascally sgnfcan. The esmaed governmen budge defc, 0.55, suggess ha for each 1 % ncrease n budge defcs, here resuls 0.55% rse n curren accoun defcs n he longrun. Smlarly, he coeffcen of nvesmen, -0.165, mples ha abou less han 1/5 of domesc nvesmens are fnanced from world fnancal markes. As hs value s relavely close o zero, also an ndcaon of he exsence of he Feldsen Horoka puzzle, whch suggess ha Turkey s fnancal negraon o he world markes s lmed durng he esmaon perod despe he wave of globalzaon n he decades of 1990s and 000s. These 13

resuls are n lne wh he sudy of Alnas and Saban (011) for Turkey. The speed of adjusmen parameer s 1.5, suggesng ha when he curren accoun balance equaon s above or below s equlbrum level, adjuss by 76% whn he frs year. The full convergence o s equlbrum level akes less han one year. Table 5. Resuls of Granger causaly F-sascs (probably) Dependen Varable (x m) ( g) EC -1 (-sasc) (x m) - 1.3 (0.09) 0.89 (0.15) -0.66 (1.1) ( g) 0.6 (0.61) - 0.79 (0.56) 0.44 (1.4) 1.06 (0.3) - Causaly nference : none * ** and ndcae 5 % and 10 % sgnfcance levels, respecvely. The probably values are n brackes. The opmal lag lengh s and s based on SBC. Granger-causaly resuls ndcae ha here exss no Granger-causaly beween he curren accoun and budge balance. Ths may be also be nerpreed as he confrmaon of he REH n he shor-run. However, hs nerpreaon would be flawed as he conegraon resuls ndcae he reverse suaon n he frs nsance. Granger-causaly resuls conradc he resul of Alnas and Saban (011). Table 6. Decomposon of Varance Percenage of forecas varance explaned by nnovaons n: Years Curren accoun defcs Budge defcs Invesmen 0 1.000 0.000 0.000 1 0.79 0.05 0.44 0.71 0.057 0.30 3 0.731 0.055 0.1 5 0.718 0.061 0.19 10 0.716 0.066 0.16 Noes: Fgures n he frs column refer o horzons (.e., number of years). All fgures are rounded o wo decmal places. The covarances marces of errors from all he VECMs appeared o be very small and approachng zero suggesng ha he combnaons of all he varables n hese models are lnear. Therefore, he orohogonal case for he varance decomposons are appled. 14

Table 6 provdes he summary resuls for he VDCs. As for he VDCs, a subsanal poron of he varance of curren accoun defcs (7.9%) s explaned by s own nnovaons n he shor-run, for example, a he wo-year horzon. In he long-run, for example, a he en-year horzon, he poron of he varance of curren accoun defcs slghly decreases from 71.6% mplyng ha oher varables explan abou 7 % of he shocks n he curren accoun defcs. The pos-sample VDCs also ndcaes ha 1.6% of he shocks n he curren accoun defcs s due o nnovaons n nvesmen a he en year-horzon, emphasng he fac ha nvesmen s he man cause of he curren accoun defcs n he long-run. 4. Conclusons Ths sudy esed he valdy of he wn-defc and he Feldsen Horoka hypoheses for Turkey and concludes ha only he former hypohess s vald for he perod of 1987-004 as far as he conegraon ess are concerned. Ths paper also fnds ha he Feldsen Horoka puzzle s presen as Turkey appeared o be negraed no world fnancal markes relavely wh a low degree capal mobly and durng he esmaon perod less han 1/5 of s nvesmens s fnanced wh foregn savngs despe he consderable amoun of globalzaon. However, he augmened Granger-causaly ess or he VDCs dd no ndcae any sgnfcan causaly beween he curren and fscal accouns. Therefore, our resuls are no conclusve o confrm ha here exss wn a defc phenomenon n Turkey. The VDCs are presened here o sugges ha one of he man deermnans of he curren accoun defcs n he long-run s he level of nvesmens. Turkey has been pursung a growh polcy n he las decade, where foregn savngs are subsued for domesc savngs and prvae consumpons are encouraged. I would be approprae o recommend ha Turkey should connue o manan s floang exchange rae regme and allow s currency o deprecae faser o reduce he curren accoun 15

defcs n he shor-run bu hs polcy would no be effcen n he long-run. Therefore, he expor promoon polces along wh he srucural reforms should be desgned o overcome he curren accoun defcs. Smlarly, ax ncreases would allevae he fscal defcs n he long-run. Moreover, he long erm-savng polcy n he form of penson funds should be wdened wh compulsory measuremens o reduce furher he savng gaps. 16

References ANDRES. F., HALICIOGLU, F. (011) Tesng he Hypohess of Naural Sucde Raes: Furher Evdence from OECD Daa, Economc Modellng, Vol.8, No.1, pp.-6. ALTINTAŞ, H. AND S. TABAN (011): Twn Defc Problem and Feldsen-Horoka Puzzle Hypohess n Turkey: ARDL Bound Tesng Approach and Invesgaon of Causaly, Inernaonal Research Journal of Fnance and Economcs, 74, 30-45. AY, A., KARAÇOR, Z., MUCUK, M., AND S. ERDOĞAN (004): Büçe Açığı-Car Đşlemler Arasındak Đlşk: Türkye Örneğ, Selçuk Ünverses SBE Dergs, 1, 75-8. BAGHERI, F., KESKHARAN, S., AND F.D. HAZRATI (01): Twn Defcs and Feldsen-Horoka Puzzle n he Case of Iran, Journal of Socal and Developmen Sudes, 3(5), 167-171. BAHMANI-OSKOOEE, M., HALICIOĞLU, F., AND S. BAHMANI (017): Do Exchang rae Changes Have Symmerc or Asymmerc Effecs on he Demand for Money n Turkey?, Appled Economcs, 49(4), 461-470. BAHMANI-OSKOOEE, M., HALICIOĞLU, F., AND S.W. HEGERTY (016): Mexcan Blaeral Trade and he J-curve: An Applcaon of he Nonlnear ARDL Model, Economc Analyss and Polcy, 50, 3-40. BARRO, R. J. (1974): Are Governmen Bonds Ne Wealh? Journal of Polcal Economy, 8, 1095-1117. BOLAT, S., BELKE, M., AND O. ARAS (011): Türkye de Đkz Açık Hpoeznn Geçerllğ: Sınır Tes Yaklaşımı, Malye Dergs, 161, 347-364. ÇATIK, A.N., MARTIN, C., AND A.O. ÖNDER (011): Relave Prce Varably and he Phllps Curve: Evdence from Turkey, Journal of Economc Sudes, 38(5), 546-561. ÇATIK, A.N., GÖK, B., AND U. AKSEKĐ (015) A Nonlnear Invesgaon of he Twn Defcs over he Busness Cycle: Evdence from Turkey, Economc Sysems, 39(1), 181-196. COAKLEY, J., KULASI, F., AND R. SMITH (1996): Curren Accoun Solvency and he Feldsen Horoka Puzzle, Economc Journal, 106, 60-67. DELL ANNO, R. AND F. HALICIOĞLU (010): An ARDL Model of Recorded and Unrecorded Economes n Turkey, Journal of Economc Sudes, 37(6), 67-646. DICKEY, D. A. AND W. A. FULLER (1981): Lkelhood Rao Sascs for Auoregressve Tme Seres wh a Un Roo, Economerca, 49, 1057-107. 17

DICKEY, D. A. AND W. A. FULLER (1979) Dsrbuon of he Esmaors for Auoregressve Tme Seres wh a Un Roo, Journal of Amercan Sascan Assocaon, 40, 1-6. DURMAZ, N. (015): Indusry level J-curve n Turkey, Journal of Economc Sudes, 4(4), 689-706. ENGLE, R. F. AND C.W.J. Granger (1987): Conegraon and error correcon: represenaon, esmaon and esng Economerca, 55, 51-76. ELLIOT, G., ROTHENBERG, T., AND J. STOCK (1996): Effcen ess for an auoregressve un roo, Economerca, 64, 813-836. ERDEM, E., KÖSEOĞLU, A., AND A.G. YÜCEL (016) Tesng he Valdy of Feldsen- Horoka Puzzle: New Evdence from Srucural Breaks for Turkey Theorecal and Appled Economcs, 3(),17-6. FELDSTEIN, M. AND C. HORIOKA (1980): Domesc Savng and Inernaonal Capal Flows, The Economc Journal, 90, 314-39. FIDRMUC, J. (003): The Feldsen-Horoka Puzzle and Twn Defcs n Seleced Counres, Economcs of Plannng, 36, 135-15. HALICIOGLU, F. AND N. KETENCI (016): The Impac of Inernaonal Trade on Envronmenal Qualy: he Case of Transon Counres, Energy,109,1130-1138. HALICIOGLU, F. (013) Dynamcs of Obesy n Fnland, Journal of Economc Sudes, 40(5), 644-657. HALICIOGLU, F. (01) Temporal Causaly and Dynamcs of Crme n Turkey Inernaonal Journal of Socal Economcs, Vol. 39, No.9, pp.704-70. HALICIOGLU, F. AND C. KARATAS (013): A Socal Dscoun Rae for Turkey, Qualy and Quany, 47(), 1085-1091. HALICIOGLU, F. (007): The J-curve Dynamcs of Turksh Blaeral Trade: A Conegraon Approach, Journal of Economc Sudes, 34(), 103-119. HALICIOGLU, F. (004) The Gbson Paradox: An Emprcal Invesgaon for Turkey European Research Sudes, Vol.7, Nos: 1-, pp.111-119. HALICIOGLU, F. (001) An Economerc Analyss of Foregn Drec Invesmen Flows no Turkey from Major Global Regons: 1975-1999 Conference Paper, presened a METU Inernaonal Conference n Economcs V, Ankara. HALICIOGLU, F. (1999) The Black Economy n Turkey: An Emprcal Invesgaon The Revew of Polcal Scences of Ankara Unversy, Vol.53, No.1, pp.147-158. 18

ĐYĐDOĞAN, P, (013): The Twn Defcs Phenomenon n Turkey: An Emprcal Invesgaon, Journal of Busness, Economcs and Fnance, (3), 36-45 KAYHAN, S., BAYAT, T., AND B. YÜZBAŞI, (013): Governmen Expendures and Trade Defc n Turkey: Tme Doman and Frequency Doman Analyses, Economc Modellng, 35, 153 158 KHAN, M.A. AND S. SAEED (01): Twn Defcs and Savng-Invesmen Nexus n Paksan: Evdence from Feldsen-Horoka Puzzle, Journal of Economc Cooperaon and Developmen, 33, 1-36. MARINHEIRO C.F. (013): Rcardan Equvalence, Twn Defcs and he Feldsen- Horoka Puzzle n Egyp, Journal of Polcy Modellng, 30, 1041-1056. NARAYAN, P.K. (005): The Savngs and Invesmen Nexus for Chna: Evdence from Conegraon Tess, Appled Economcs, 37, 1979-1990. PATTICHIS, C. (01): Exchange Rae Effecs on Trade n Servces, Journal of Economc Sudes, 39(6), 697-708. PESARAN, M. H., SHIN, Y., AND R.J. SMITH (001): Bounds Tesng Approaches o he Analyss of Level Relaonshps, The Journal of Appled Economercs, 16, 89-36. PHILLIPS, P.C.B. AND P. PERRON (1988): Tesng for a Un Roo n Tme Seres Regresson, Bomerka, 75, 335-346. SACHSIDA, A. AND M.A.R. CAETANO (000): The Feldsen-Horoka Puzzle Revsed, Economcs Leers, 68, 85-88. SIMS, C. A. (1980): Macroeconomcs and Realy, Economerca, 48, 1-48. TAYEBI, S.K. AND M. YAZDANI (014): Fnancal Crss, Ol Shock and Trade n Asa, Journal of Economc Sudes, 41(4), 601-614. YAPRAKLI, S. (010): Türkye de Esnek Dövz Kuru Rejm Alında Dış Açıkların Belrleycler: Sınır Tes Yaklaşımı, Ankara Ünverses SBF Dergs, 65(4), 141-163. YAY, G.G. AND H. TAŞTAN, (007): Đkz Açıklar Olgusu: Frekans Alanında Nedensellk Yaklaşımı, Đsanbul Ünverses SBF Dergs, 37, 87 111 19

Appendx Daa defnon and sources Daa are colleced from wo dfferen sources: Inernaonal Fnancal Sascs of Inernaonal Moneary Fund (IMF) and Turksh Mnsry of Fnance (TMoF). (x m) s he dfference beween expors and mpors as % of GDP. Source: IMF. ( g) s he dfference beween ax revenues and governmen expendures as % of GDP. Source: TMoF. s he oal nvesmens as % of GDP. Source: IMF. 0