Do Tax Implications Change the Fisher Effect for the Turkish Economy?

Transcription

1 Loyola Universiy Chicago Loyola ecommons Topics in Middle Easern and Norh African Economies Quinlan School of Business 9-- Do Tax Implicaions Change he Fisher Effec for he Turkish Economy? Kenan Lopcu Çukurova Universiy Nuran Coşkun Mersin Universiy Süleyman Değirmen Mersin Universiy Recommended Ciaion Lopcu, Kenan; Coşkun, Nuran; and Değirmen, Süleyman, "Do Tax Implicaions Change he Fisher Effec for he Turkish Economy?" (). Topics in Middle Easern and Norh African Economies. 79. hp://ecommons.luc.edu/meea/79 This Aricle is brough o you for free and open access by he Quinlan School of Business a Loyola ecommons. I has been acceped for inclusion in Topics in Middle Easern and Norh African Economies by an auhorized adminisraor of Loyola ecommons. For more informaion, please conac ecommons@luc.edu. This work is licensed under a Creaive Commons Aribuion-Noncommercial-No Derivaive Works. License. he auhors

2 Topics in Middle Easern and African Economies Vol. 5, No., Sepember DO TAX IMPLICATIONS CHANGE THE FISHER EFFECT FOR THE TURKISH ECONOMY? Kenan Lopcu Nuran Coşkun Süleyman Değirmen Prepared for Presenaion a he nd Annual Meeing of he MEEA, ASSA January 4-7,, San Diego, CA Absrac This sudy invesigaes he validiy of boh he convenional and ax-adjused Fisher effecs using ime series mehods such as he ARDL bounds es, and Gregory-Hansen (G-H) coinegraion es. To compare he convenional and ax-adjused Fisher effecs, we use wo differen ime series daa for ineres raes: ) ineres raes adjused for axes, and ) ineres raes no adjused for axes. Using monhly changes in quarerly and annual ineres and inflaion raes, boh he G-H and ARDL bound ess suppor he convenional and ax-adjused Fisher effecs. However, he magniude of he inflaion coefficien in he long-run relaionship ends o decline for he ax adjused Fisher effec. Keywords: Fisher Effec, Taxaion, Time Series Mehods JEL Code: C, E4 Inroducion The exisence of he Fisher effec in a counry is imporan in deciding wheher implemened economic policies are susainable. When a counry experiences deviaions from is argeed inflaion rae due o shocks o he sysem, he nominal ineres rae could be a significan insrumen for inflaion argeing, given he coinegraion beween he nominal ineres rae and inflaion. Thus, for counries like Turkey, which aims a price sabiliy, he nominal ineres rae is a significan insrumen for long-run inflaion argeing. The empirical validiy of he Fisher effec has been resudied exensively in he lieraure in he las several decades, mosly due o he developmen of ime series mehods Çukurova Universiy, Deparmen of Economerics, klopcu@cu.edu.r Mersin Universiy, Deparmen of Economics, ncoskun@mersin.edu.r Mersin Universiy, Deparmen of Economics, suleymandegirmen@gmail.com

3 Topics in Middle Easern and African Economies Vol. 5, No., Sepember dealing wih non-saionary daa. Researchers now have various coinegraion (Engle- Granger, Johansen) and VAR echniques in heir hands o es he long-run relaionship beween individually non-saionary series (Akins and Coe, ). However, poenial shorcomings of hese sudies are ) ha hese echniques require ha individual series be non-saionary; and ) mos of hese sudies es he exisence of a convenional Fisher effec raher han a ax-adjused Fisher effec. The purpose of his sudy, hen, is o es if he convenional and ax-adjused Fisher effecs are valid for Turkey, one of he mos dynamic economies in he classificaion of developing counries. This sudy also aims o conribue o he curren lieraure by including ax adapaion in he nominal ineres rae for a developing counry case. To ha end, we use wo differen ime series daa for ineres raes: ) ineres raes adjused for axes, and ) ineres raes no adjused for axes. Time series mehods such as he ARDL bounds es of Pesaran, Shin and Smih () and coinegraion ess wih a srucural break are used o deermine if consisen resuls for he exisence of he Fisher effec can be reached. In conclusion, for he pos financial liberalizaion era in Turkey when ax adapaion has been especially implemened, declines in he inflaion rae are associaed wih declines in he nominal ineres rae. Therefore, in conras o he findings for developed counries (see Akins & Coe, ), his sudy finds suppor in favor of boh ypes of Fisher effecs in Turkey: he convenional and he ax-adjused Fisher effec. Theoreical Framework and Lieraure Review Fisher (9) hypohesized ha he nominal ineres rae was equal o he sum of he real ineres rae and he expeced inflaion rae. He claimed ha he nominal ineres rae is comprised of he real ineres rae and he expeced inflaion of he same period. Addiionally, he Fisher effec assers ha here is a linear relaionship beween he nominal ineres rae and he expeced inflaion, and assumes ha he real ineres rae does no change in he long run. If he real ineres rae is no affeced by moneary imbalances ha affec inflaion in he long erm, his will cause a relaionship beween inflaion and he nominal ineres rae, leading o he likely exisence of coinegraion beween he nominal ineres rae and inflaion. To es he Fisher effec we use a sandard model ofen encounered in he lieraure (e.g. Harrison ()): in real in E (inf ) According o he raional expecaions;

4 Topics in Middle Easern and African Economies Vol. 5, No., Sepember E (inf ) inf = e - where E ( e ) = Hence, he average expecaional error is zero. Addiionally, raional expecaions sugges ha he predicion error is uncorrelaed wih all he variables in he informaion se a he ime of he predicion. Thus, subsiuing he realized inflaion rae for he expeced one he Fisher equaion and our model is hen; in real in inf More generally, his equaion should be wrien as follows for developing counries; ( i ) ( r )( ) i r r The r erm is imporan for developing counries which have suffered from high inflaion raes. Fisher (9) examined he relaionship beween nominal ineres raes and he rae of inflaion for he U.S. and he U.K. using annual daa over he period for he US, and he 8 94 period for he U.K. Though Fisher (9) assered ha he nominal ineres rae and inflaion should affec one anoher evenly, he failed a presening his relaionship empirically as his resuls indicaed significan deviaions from he Fisher effec. Mundell (96) aribued Fisher s empirical resuls o he inflaionary process by assering ha he inflaionary process reduces he real rae hrough he wealh effec. Similarly, Tobin (965) saed ha nominal ineres raes should adjus less han one o one due o inflaionary pressures on he real ineres rae. Darby (975) and Feldsein (976), on he oher hand, argue ha he nominal ineres rae would adjus more han one for one ( /( ) ) o he expeced inflaion rae due o he ax effec. Shome, Smih and Pinkeron (988) argued ha risk-averse invesors need a premium o compensae hem for any risks. Carmichael and Sebbing (98) assumed ha nominal ineres raes on financial asses can be considered consan over ime and ha he real ineres rae moves inversely wih inflaion. Engsed (996) and Crowder and Hoffman (996) found suppor for he ax-adjused Fisher effec for OECD counries and he U.S., respecively. Carr, Pesando and Smih (976), in conras, could no find any evidence ha suppors he ax-adjused Fisher effec for Canada while Akins and Coe () found evidence supporive of he convenional Fisher effec for he U.S. and Canada, bu rejeced he ax-adjused Fisher effec for Canada and provided mixed resuls for he U.S. Evans and Lewis (995) and Mishkin (99) employed

5 Topics in Middle Easern and African Economies Vol. 5, No., Sepember Engle and Granger s (987) bivariae coinegraion es. Neverheless, Mishkin s (99) resuls did no suppor he Fisher hypohesis, while Evans and Lewis (995) resuls suppored he Fisher hypohesis only when regime shifs in he expeced inflaion process were aken ino accoun. Since 994 Turkey has been applying income ax on nominal ineres incomes. In he even of a ax applicaion on nominal ineres incomes, he examinaion of he Darby- Feldsein effec is imporan in erms of policy proposals in counries wih inflaion argeing whose main insrumen is shor erm ineres raes. For Turkey, Turgulu (4), Kesriyeli (994), Şimşek and Kadılar (6) and Köse e al () obained findings supporive of he Fisher effec in Turkey. Addiionally, Kasman e al (6) found resuls in favor of he presence of he Fisher effec in a large majoriy of counries, including Turkey. All he menioned sudies for Turkey, however, excluded he ax adjusmen procedure, which led us o examine he ax adjused Fisher effec for Turkey. Daa This sudy examines he validiy of he Fisher effec in Turkey over he period of 99: hrough :. Ineres raes are proxied by quarerly and annual deposi raes and are from inernaional financial saisics (IFS). Two differen ime series are used for nominal ineres raes: i) Ineres raes adjused for axes ii) Ineres raes no adjused for axes The quarerly inflaion rae is measured as he monh o monh percenage change in he consumer price index (CPI) muliplied by 4. The annual inflaion is measured as he yearly change in he CPI muliplied by. Turkey has implemened income ax on ineres earnings since 994. Tax raes are obained from he decisions of he Council of Minisers. Because no publicly available daa exis for double axaion and ax immuniies hey canno be aken ino accoun. So, he ax adjused ineres raes are only imperfec proxies of he heoreical ones. Neverheless, hey are also he only ones ha can be derived from he exising daa sources. Variable definiions, he period covered and he daa sources are given in Table. Quarerly and annual inflaion, ineres rae and he ax adjused ineres rae series are shown in figures and, respecively. According o he graphs, changes in ineres raes follow he changes in he inflaion rae. This is paricularly rue for he crises periods of 994 and. These are also he periods where he ineres raes end o be negaive or circa zero. I is also noiceable ha he declines in he inflaion rae are associaed wih declines in nominal ineres raes in he las decade. 4

6 Topics in Middle Easern and African Economies Vol. 5, No., Sepember Table : Daa Variables Definiion Source Period inf Quarerly inflaion rae. (APR) IFS 99:-:8 in Quarerly nominal ineres rae. (APR) IFS 99:-:8 in Tax adjused quarerly nominal ineres rae. IFS 99:-:8 (APR) inf Annual inflaion rae. (APR) IFS 99:-: in Annual nominal ineres rae. (APR) IFS 99:-: in Tax adjused annual nominal ineres rae. (APR) IFS 99:-: ax ( ) ax ( ) Tax raes for quarerly deposi revenue. Council of Miniser Decisions Tax raes for annual deposi revenue. Council of Miniser Decisions 99:-:8 99:-: Figure: INF INT TINT 5

7 Topics in Middle Easern and African Economies Vol. 5, No., Sepember Figure : INF INT TINT Mehodology We sar he analysis firs by invesigaing he order of inegraion of he individual series via ADF and KPSS uni roo ess. The empirical validiy of he Fisher effec has been resudied exensively in he lieraure, hanks o he developmen of various ime series mehods such as Engle-Granger and Johansen coinegraion ess and VAR echniques. However, hese echniques require ha individual series be non-saionary. Therefore, using he ARDL bound es o deermine long-run relaionships for he Fisher effec provides an advanage since his echnique does no require he nominal ineres rae and he inflaion rae o be inegraed a he same order. On he oher hand, due o he likely changes in he inflaion - ineres rae relaionship in he long run, he G-H coinegraion es which allows for he possibiliy of a srucural break is used. ADF and KPSS Uni Roo Tess To es he ime series properies of individual series we sar wih he ADF and KPSS ess. We perform he ADF es, saring wih he larges model which includes rend and drif, and es he significance of rend and hen drif (φ, φ ess) as well. If he null of uni roo 6

8 Topics in Middle Easern and African Economies Vol. 5, No., Sepember and insignifican rend or drif canno be rejeced, we move o he nex model. Lag lenghs are deermined saring wih he maximum lengh of and reducing hem unil we find a significan -value for he individual lag. Since he ADF ess have low power, we also use he KPSS ess o corroborae he resuls. For he KPSS ess boh he rend and drif models are used o es he null of no uni roo, using Newey-Wes Bandwidh crieria wih Barle kernel. However, neiher of he above ess akes ino accoun a poenial srucural break when esing for a uni roo. For ha, we urn o he Zivo-Andrew (Z-A) (99) uni roo es. Zivo-Andrews Uni Roo Tes (Z-A) Uni roo ess can be biased if he srucural breaks are no accouned for. Z-A (99) proposed a es ha deermines he break in he series endogenously. The individual series has a uni roo under he null and is rend saionary wih a srucural break a a ime T B (<T B <T) under he alernaive. The es is based on a procedure ha rims a small percenage of he daa a wo ends and calculaes he -value for each of he remaining observaions as being he poenial break poin. The observaion wih he smalles -value, he leas favorable oucome for he null, hen, is chosen as he break poin. Z-A (99) provided he criical values for he es. Three differen models are proposed for he es. Model A permis one ime change in he level of he series: y A A DU A A ( ) y k j c A j y j Model B allows one ime change in he slope of he rend funcion occurring a ime T B: y B B DT * B B ( ) y k j c B j y j Finally, Model C permis a oneime change boh in he level and he rend of he series. y C C DU C C ( ) DT * C ( ) y k j c C j y j DU(λ), DT*(λ) are dummy variables, represening he change in he level and he rend of he series:, TB, DU ( ),oherwise fracion. * DT ( - TB, TB, ) TB T, and λ is he break,oherwise 7

9 Topics in Middle Easern and African Economies Vol. 5, No., Sepember ARDL Bound Tes Approach Pesaran, Shin and Smih (PSS) () developed an ARDL based es o deec he long-run relaionship beween he variables regardless of he individual series being I() or I(). The mos imporan advanage of using his mehod for esing he validiy of he Fisher effec is ha i does no require he nominal ineres rae and he inflaion series o be inegraed a he same order. The model o invesigae he long-run level relaionship is as follows: Y c c Y c X m i c Y yi i m i c xi In order o apply he bound es, firs, he order of appropriae lag lengh is deermined from a VAR model, using crierion such as AIC, SIC, or HQ. The model should be free of auocorrelaion a he seleced lag. So, he addiional lags are added if needed. The null of no long-run relaionship can be esed by an F es, resricing c =c =. The disribuion of he es saisic is non-sandard and canno be compared wih he sandard F disribuion, bu he criical values were provided in PSS (), If he es saisic is below he lower bound, he null canno be rejeced; if i is above he upper bound he null is rejeced. In cases where he es saisic is beween he lower and upper bound, no cerain conclusion can be reached regarding he exisence of a long-run level relaionship. Once he exisence of a long-run level relaionship is concluded, he variables in he ARDL model are allowed o have differen lag lenghs and he model is reesimaed o deermine he long and shor-run relaionships. The X i e long-run coefficiens can be esimaed using he ARDL model: Y m i Y yi i n i X xi i To deermine he shor-run relaionship beween he variables, he error correcion represenaion of he ARDL model is used: Y m i Y yi i n i X xi i ecm Where ecm - represens he long-run equilibrium, and he differences of X and Y sand for he shor-run dynamics. Gregory- Hansen Coinegraion Tes Gregory and Hansen (G-H) (996a, b) proposed a residual based procedure for esing he null of no coinegraion agains he alernaive of coinegraion wih a srucural break. The ime of he break poin is decided endogenously. The dummy variable, specifying he iming of he srucural change, is defined as follows (Gregory Hansen, 996a, p. -): 8

10 Topics in Middle Easern and African Economies Vol. 5, No., Sepember,if n,if n,, where n is he number of observaions, τ (,) indicaes he iming of he srucural change poin and [ ] denoes he ineger par. The four models proposed are as follow: Model : Level Shif (C) Model y T y, =,, n This model allows a break in he inercep, where µ represens he inercep before he break, and µ represens he change in he inercep a he ime of he break. Model : Level Shif wih rend (C/T) Model y T y, =,, n This model akes he break ino accoun in he inercep in he presence of a deerminisic rend. Model : Regime Shif (C/S) Model y T y y, =,, n. T The model does no include a deerminisic rend, bu permis a break in he inercep as well as in he slope coefficiens. µ and µ are as in Model, T and T, on he oher hand, represen he slope coefficiens before he regime shif and he changes in he slope coefficiens afer he regime change, respecively. Model 4: Trend and Regime Shif (C/T/S) Model y T y y, =,, n The model was proposed by G-H (996b) and is he larges model which allows a T break in he level, rend and slope coefficiens. µ, β, and and slope coefficiens, respecively, before he regime change, µ, β and denoe he changes in he parameers afer he regime change. T represen he inercep, rend T, in conras, The es is based on a procedure ha rims a small percenage of he daa a wo ends and calculaes he ADF, and he Phillips (987) Z α and Z saisics for each of he remaining observaions as being he poenial break poin. The observaion wih he smalles es-value, he leas favorable oucome for he null, hen, is chosen as he break poin. Hence he hree es saisics proposed are as follows (Gregory and Hansen 996a, b). ADF * * min ADF( ), Z min Z ( ) T T, Z * min Z ( ) T 9

11 Topics in Middle Easern and African Economies Vol. 5, No., Sepember Empirical Resuls Uni Roo Tess: The resuls for he ADF and KPSS uni roo ess are given in Table. All he ADF and KPSS ess indicae ha all he variables are non-saionary a convenional significance levels, excep in and in which are significan a he 5% level, using he larges model for he ADF es. However, he null of no uni roo is rejeced for boh of hese variables by he KPSS ess. So, we conclude ha all he variables are inegraed order of one according o he resuls. Table : ADF and KPSS Uni Roo Tess Series ADF KPSS Trend&Drif Drif None Trend&Drif Drif k k k K k inf in in inf in in Criical Values % %5 % k is he number of he lags and he bandwidh parameer for he ADF and KPSS ess, respecively In he case of he Z-A es, he effec of a poenial srucural break on he individual series degree of inegraion is invesigaed for all he hree models. The resuls are presened in Table. Firs of all, all he break daes for all he series occur in or around he wo crises years of 994 and for Turkey. Secondly, none of he ess can decisively rejec he null of uni roo (a % significance level), excep he resuls for in and in from Model B. Noneheless, Sen () suggesed ha Model C is preferable o he oher models. Alhough he resuls from Model C sugges ha in, in and inf are saionary wih a srucural break a he 5% level, he null of uni roo canno be rejeced a he % significance level, even afer allowing a break boh in he level and rend. So, he conclusion based on he ADF and he KPSS ess regarding he degree of inegraion of all he variables may sill very well be valid.

12 Topics in Middle Easern and African Economies Vol. 5, No., Sepember Table : Zivo- Andrews Uni Roo Tes Series Model A Model B Model C K Tes T B DU K Tes T B DT K Tes T B DU DT Prob. saisic Prob. saisic Prob. saisic Prob. Inf in in Inf in in Criical Values % %5 % k is he number of he lags deermined by a significan - raio reduced from and T B is he break poin. Table 4-a : ARDL (4,) for in Regressor Coefficien Sandard Error T-Raio[Prob] inf [.] C [.] 95% Lower Bound 95% Upper Bound 9% Lower Bound 9% Upper Bound F-saisic ARDL(4,4) 4.67 ARDL(4,4): Serial Correlaion:CHSQ() =.4[.44] F(,6) =.96946[.479] Table 4-b: ARDL(4,) for in Regressor Coefficien Sandard Error T-Raio[Prob] inf [.] C [.] 95% Lower Bound 95% Upper Bound 9% Lower Bound 9% Upper Bound F-saisic ARDL(,).55 ARDL(,):Serial Correlaion:CHSQ() = [.69] F(,) =.6865[.7] Table 4-c: ARDL(5,) for in Regressor Coefficien Sandard Error T-Raio[Prob] inf [.] C [.9] 95% Lower Bound 95% Upper Bound 9% Lower Bound 9% Upper Bound F-saisic ARDL(,) 7.84 ARDL(,):Serial Correlaion:CHSQ() =.8[.5] F(,) =.9655[.54]

13 Topics in Middle Easern and African Economies Vol. 5, No., Sepember Table 4-d: ARDL(,) for in Regressor Coefficien Sandard Error T-Raio[Prob] inf [.] C [.] 95% Lower Bound 95% Upper Bound 9% Lower Bound 9% Upper Bound F-saisic ARDL(,) ARDL(,):Serial Correlaion:CHSQ() =.544[.568] F(,) =.84999[.599] Coinegraion Tes Resuls The ARDL bound es resuls and he esimaed long-run level relaionships are displayed in ables 4-a o 4-d. The calculaed F-values are above he upper bounds in all cases, indicaing he exisence of a long-run relaionship beween he nominal ineres rae and he expeced inflaion rae. The LM ess as well indicae ha he errors are free of serial correlaion a he 5% significance level. Once he coinegraion relaionship is verified, each model is reesimaed, allowing he righ-hand variables o have differen lag lenghs. In all he cases he poin esimaes for he expeced inflaion coefficiens are slighly less han uniy, bu he 95% confidence inervals conain uniy. Thus, he ARDL resuls suppor boh he convenional and ax adjused Fisher effec for Turkey. In order o ake ino accoun a poenial srucural break in he long-run relaionship, we use he G-H coinegraion es. The resuls are provided in Table 5. The G-H ess provide a subsanial suppor for coinegraion wih a srucural break. In paricular, using a -monh inflaion and ineres rae series, all hree ess saisics rejec he null of no coinegraion for all he srucural change models and for boh he ax adjused and non-adjused ineres raes. For he -monh inflaion and ineres rae series, he ADF ess furher verify he exisence of coinegraion for all he models excep he level shif (C) and he rend and he regime shif (C/T/S) models. For he models where he null of no coinegraion is rejeced in favor of coinegraion wih a srucural break, we esimae he following regressions o see how he coinegraion relaionships may have changed over ime. C C C : in dummy E (inf ) / T : in rend dummy k E (inf ) / S : in dummy E (inf ) dummye (inf k ) k

14 Topics in Middle Easern and African Economies Vol. 5, No., Sepember C / T / S : in dummyrend 4 rend dummy E (inf ) 4dummyE (infk ) where in would replace in when he model is esimaed for he ax adjused Fisher effec. Table 5 :Gregory Hansen Tes Resuls Model k ADF Break Poin Z Break Poin Z a Break Poin C (in,inf) -7.7*.658 (7) -6.57*.658 (7) -76.*.658 (7) C/T (in,inf) -7.4*.76 (54) -6.55*.76 (54) -74.5*.76 (54) C/S (in,inf) -7.6*.576 (67) -8.4*.9 (5) -.47*.9 (5) C/T/S (in,inf) -7.47*.47 () -7.9*.47 () -88.6*.47 () C (in,inf) -7.55*.65 (69) -6.6*.669 (6) -77.8*.669 (6) C/T (in,inf) -7.47*.76(54) -6.5*. (5) -7.7*. (5) C/S (in,inf) -7.7*.584 (9) -8.*.9 (5) -.5*.9 (5) C/T/S (in,inf) -7.7*.476 () -6.96*.47 () -8.57*.47 () C (in,inf) (49) (55) (55) C/T (in, inf) 9-5.*.55 (59) (55) (55) C/S (in, inf) 9-5.7*.59 (8) (46) (46) C/T/S (in, inf) 9-5.7*.498 (5) (5) (5) C (in, inf) *.74 (79) (55) (55) C/T (in, inf) 9-5.9*.5498 (64) (55) (55) C/S (in, inf) *.759 (8) (46) (46) C/T/S (in, inf) (5) (4) (4) C C/T C/S C/T/S Observaio n %/5/ %/5/ %/5/ %/5/ -5. / -4.6 / / / / / -4,68-6. / -5.5 / -5.4 Criical Values Corresponding Break Daes -5. / -4.6 / / / / / -4,68-6. / -5.5 / / / / / / / / / -5. Dae Observaion Dae Observaion Dae Observaion Dae 99: 99: 994: 994:4 994: :7 994: 995:4 995:7 997: :4 999:6 999:7 :5 : :7 4: 4: 4: 5: The resuls in Table 6 show ha he effec of inflaion on he nominal ineres rae is quie close o one boh for he convenional and ax-adjused Fisher effec a leas in one of he regimes. When he sandard errors are aken ino consideraion, he presence of he Fisher effec canno be rejeced. Endogenously deermined break daes follow he 994 and economic crises in general. Typically for breaks following he 994 crisis he impac of inflaion on he nominal ineres rae increases significanly and he Fisher effec holds. This

15 Topics in Middle Easern and African Economies Vol. 5, No., Sepember can be seen, for example, for he ax-adjused and non-adjused quarerly ineres raes (in and in) C/S models wih break poins 9 and 67. Before he break poin 67, one percenage poin increase in he inflaion rae causes a.6 percenage poin increase in he nominal ineres rae. However, afer he break, a one percenage poin increase in inflaion causes a rise of.9 percenage poin, almos one o one, in he ineres rae. This can be explained by he fac ha he 99s in Turkey is he high inflaion period wih rising inflaionary expecaions, especially afer 994. Table 6: Esimaed Coinegraion Regressions wih a Srucural Break for Seleced Models Model Coefficien Sd. Error -Sa Prob. GH- (67) 995:7 C / S : in dummyinf GH- (5) 994: C / S : in dummyinf GH- (9) 997:8 C / S : in dummyinf GH- (5) 994: C / S : in dummyinf GH- (59) 994: C / T : in inf GH-4 (5) :5 dummy inf dummy inf dummy inf C / T / S : in dummyinf dummy inf dummy rend dummyrend inf 5 GH- (79) 4: dummy rend C : in dummy inf GH- (64) 995:4 C / T : in inf GH- (8) 5: C / S : in dummyinf 4 dummy rend dummy inf E-4 6.6E-4 5.5E- 6.45E-9.E- 7.84E-.76E-.6E-5 8.8E E-7 6.9E-.85E-.6E-6 7.E-4.7E-.88E-6.E-6.4E-.5E-7.58E E-6 6.E-.99E E E- 5.4E-8.6E E-9.96E E-7 4.6E- 6.5E-9..4E

16 Topics in Middle Easern and African Economies Vol. 5, No., Sepember An opposie picure arises for he break poins around or afer. In his case he impac of inflaion on he ineres rae becomes weaker afer he break. This can be aribued o decreasing inflaionary expecaions as he las decade is a relaively lower inflaion period. See for example he C/S model for he ax-adjused and non-adjused annual ineres raes (in, in) wih break poins 8 and 5. Conclusion This sudy invesigaed he convenional and ax adjused Fisher effec for Turkey o humbly sugges a policy for an incumben governmen. Because he real ineres rae for developed small open economy counries is closer o he world ineres rae in comparison o developing counries, i is reasonable o expec ha he Darby and Feldsein effec should apply for he developed counries. Mundell (96) and Tobin (965) indicaed ha inflaion would affec he real moneary sabiliy when he economic agens prefer holding oher asses raher han money due o he increase in he anicipaed inflaion. This would lower he nominal raes insanly. Wih regard o he Mundell-Tobin effec for ax implicaions, he welfare loss caused by he inflaionary process can be he reason for he endency of inflaion coefficiens o be lower han uniy in he Fisher equaion. In his sudy, using he CPI-based inflaion rae and he ax-adjused and non-adjused deposi raes, we invesigaed he validiy of he Fisher effec wih he ARDL bounds es and he Gregory-Hansen coinegraion es. The resuls from he ARDL mehod are, in general, supporive of boh he convenional and ax-adjused Fisher effecs, bu he magniude of he inflaion coefficiens ends o decline for he ax adjused Fisher effec. The G-H ess provide subsanial evidence for he exisence of he coinegraion relaionship beween he expeced inflaion and he nominal ineres rae. However, abou half he coefficiens for he inflaion rae in he coinegraion relaionships are less han uniy and do no significanly change for he ax adjused Fisher effec, a fac in conradicion o he Darby-Feldsein effec, bu compaible wih Mundell-Tobin effec. Real ineres raes in Turkey have been above he world ineres rae in he pos financial liberalizaion era. Hence, invesors may coninue o prefer invesing in TL asses as long as real ineres raes are higher han he res of he world. According o he resuls from he G-H coinegraion regressions, ax implicaions on he nominal ineres rae do no affec he saving behavior. Therefore, public revenues (axes) should be obained wih direc raher han indirec axaion for price sabiliy and lower inflaion. 5

17 Topics in Middle Easern and African Economies Vol. 5, No., Sepember I should be noed ha deposi ineres raes do no accoun for excepional cases such as ax immuniies and double axaion. Savings deposis are preferred by small invesors raher han professional ones. Therefore, he resuls here should be compared wih resuls using oher measures of nominal ineres raes such as Treasury bill raes. Furhermore, he G- H es was designed o accoun for only a single srucural break. The resuls migh be biased, if here are muliple breaks, given he wo imporan economic crises (994, ) in he las wo decades, i very likely ha he Fisher equaion in Turkey had suffered more han one srucural break (see Figure, p. 6). So, he resuls here should be inerpreed wih cauion. Fuure research is needed o scruinize he Fisher effec furher wih muliple srucural breaks and he alernaive measures of ineres raes. 6

18 Topics in Middle Easern and African Economies Vol. 5, No., Sepember References: Akins, F. and Coe, P.J. (), An ARDL bounds es of he long-run Fisher effec in he Unied Saes and Canada, Journal of Macroeconomics, 4(), Carr, J., Pesado, J.E. and Smih, L.B. (976), Tax effecs, price expecaions and he nominal rae of ineres, Economic Inquiry, 4, Carmichael, J., & Sebbing, P. W. (98). Fisher's paradox and he heory of ineres. The American Economic Review, Crowder, W.J. (997), The long-run Fisher relaion in Canada, Canadian Journal of Economics, (4), 4-4. Crowder, W.J. and Hoffman, D.L. (996), The Long-run relaionship beween nominal ineres raes and inflaion: The Fisher equaion revisied, Journal of Money, Credi, and Banking, 8(), -8. Crowder, W.J. and Wohar, M.E. (999), Are ax effecs imporan in he longrun Fisher relaionship? Evidence from he municipal bond marke, Journal of Finance, 54(), 7-7. Dickey, D. A. and Fuller, W. A. (979), Disribuion of he esimaors for auoregressive ime Series wih a uni roo, Journal of he American Saisical Associaion, 74, Darby, M.R. (975), The financial and ax effecs of moneary policy on ineres raes, Economic Inquiry, (), Engle, R. F., and Granger, C. W Co-inegraion and error correcion: represenaion, esimaion, and esing. Economerica, 55() Engsed, T. (996), Non-saionariy and ax effecs in he long-erm Fisher Hypohesis, Applied Economics, 8(7), Evans, M. D. D, and Lewis K. K. (995) "Do long-erm swings in he dollar affec esimaes of he risk premia?" Review of Financial Sudies 8 (), Fama, E.F. (975), Shor-erm ineres raes as predicors of inflaion, American Economic Review, 65(), Feldsein, M. (976), Inflaion, income ax and he rae of ineres: A heoreical analysis, American Economic Review, 66, Fisher, I. (9), The Theory of Ineres, New York: Macmillan. Gregory, A.W. and Hansen, B. E. (996), Residual-based ess for coinegraion in models wih regime shifs, Journal of Economerics, 7,

19 Topics in Middle Easern and African Economies Vol. 5, No., Sepember Gregory, A. W., Nason, J.M. and Wa, D.G. (994), Tesing for srucural breaks in coinegraed relaionships, Journal of Economerics, 7(), 4. Harrison, M. (), Valuing he Fuure: he social discoun rae in cos-benefi analysis, Ausralian Governmen Produciviy Commission. Kasman, S., Kasman, A. and Turgulu, E. (6), Fisher hypohesis revisied: a fracional coinegraion analysis, Emerging Markes Finance Trade, 4, Köse, N., Emirmahmuoğlu, F. and Aksoy, S. (), The ineres rae- inflaion relaionship under inflaion argeing regime: The case of Turkey, Journal of Asian Economics, 84, -. Kesriyeli, M. (994), Policy regime changes and esing for he Fisher and UIP hypoheses: The Turkish evidence, The Cenral Bank of he Republic of Turkey, 94. Mishkin, F.S. (99), Is he Fisher effec for real? A reexaminaion of he relaionship beween inflaion and ineres raes, Journal of Moneary Economics,, Mishkin, F.S. (), Inflaion argeing in emerging marke counries, American Economic Review, 9,5-9. Mundell, R. (96), Inflaion and real ineres, Journal of Poliical Economy. 7, 8-8. TCMB. (6), Türkiye Cumhuriyei Merkez Bankası Bilançosu Açıklamalar, Rasyolar ve Para Poliikası Yansımaları, TCMB, Ankara. Perron, Pierre. (989), The grea crash, he oil price shock, and he uni roo hypohesis, Economerica 57, 6 4. Pesaran, M. H., Shin, Y. and Smih, R. J. (), Bound esing approaches o he analysis of long-run relaionships, Journal of Applied Economerics, 6, Phillips, P. C. B. (987), Time series regression wih a uni roo, Economerica, 55, 77-. Sen, A. (), On uni roo ess when he alernaive is a rend break saionary process, Journal of Business and Economic Saisics, (), Shome, D., Smih S., and Pinkeron, J. (988) The purchasing power of money and ineres raes: A re-examinaion, Journal of Finance, -5. Şimşek, M. and Kadılar, C. (6), Fisher ekisinin Türkiye verileri ile esi, Doğuş Üniversiesi Dergisi, 7, 99-. Tobin, J. (965), Money and economic growh, Economerica,,

20 Topics in Middle Easern and African Economies Vol. 5, No., Sepember Tobin, J. (949), Taxes, saving, and inflaion, The American Economic Review, 9(6), -. Turgulu, E. (4), Fisher hipoezinin uarlılığının esi: Parçalı durağanlık ve parçalı koinegrasyon analizi, D.E.İ.İ.B.F Dergisi, 9(), Zivo E. and Andrews D. W. K. (99), Furher evidence on he Grea Crash, he oil-price shock, and he uni-roo hypohesis, Journal of Business & Economic Saisics, (),

21 Topics in Middle Easern and African Economies Vol. 5, No., Sepember Appendices Table 7 : : Esimaed Coinegraion Regressions wih a Srucural Break Model Coefficien Sandard Error GH- (7) 4: C : in dummy inf GH- (54) 994:6 C / T : in inf GH- (67) 995:7 C / S : in dummyinf GH- (5) 994: C / S :in dummyinf GH-4 () 999:4 C / T / S : in dummyinf dummy rend dummy inf dummy inf dummyrend inf 5 dummy rend Tes Saisic Prob. 5.7E-6.8E- 4.88E-4 6.4E-8.5E-4.59E-7 5.5E-.6E-4 6.6E-4 5.5E- 6.45E-9.E- 7.84E-.76E-.6E-5.7E E-.68E-7.7E E-5 Table 8 : : Esimaed Coinegraion Regressions wih a Srucural Break Model Coefficien Sandard Error GH- (69) 4: C in : dummy inf GH- (6) :7 C : in dummy inf GH- (54) 994:6 C / T : in inf dummy rend Tes Saisic Prob..8E-57.67E- 5.7E-6 6.5E-6.E E-.E-45 9.E-.7E- 6.E-

22 Topics in Middle Easern and African Economies Vol. 5, No., Sepember Table 9 : : Esimaed Coinegraion Regressions wih a Srucural Break Model Coefficien Sandard Error GH- (59) 994: C / T : in inf GH- (8) 99: C / S : in dummyinf GH-4 (5) :5 dummyinf dummy rend C / T / S : in dummy inf dummyrend inf 5 dummy rend Tes Saisic Prob..E-6.4E-.5E-7.58E- 6.6E- 8.58E-4.68E- 4.64E E-6 6.E-.99E E-4.77 Table : : Esimaed Coinegraion Regressions wih a Srucural Break Model Coefficien Sandard Error GH- (79) 4: 5: C : in dummy inf GH- (64) 995:4 C / T : in inf GH- (8) 5: C / S : in dummyinf dummy rend dummy inf Tes Saisic Prob. 6.5E- 5.4E-8.6E E-9.96E E-7 4.6E- 6.5E-9..4E-5.85