Financial Development and Tax Revenue Nexus in Turkey

Transcription

1 Journal of Economics and Developmen Sudies March 216, Vol. 4, No. 1, pp ISSN: (Prin), (Online) Copyrigh The Auhor(s). All Righs Reserved. Published by American Research Insiue for Policy Developmen DOI: /jeds.v4n1a1 URL: hp://dx.doi.org/1.1564/jeds.v4n1a1 Financial Developmen and Tax Revenue Nexus in Turkey Selçuk Akçay 1, İsa Sağbaş 2 & Gökhan Demiraş 3 Absrac This sudy explores he nexus beween financial developmen (banking and non-banking) and direc ax revenue in a mulivariae framework in Turkey for he period 26 o 214, employing monhly daa. We examined he long run equilibrium relaionship beween financial developmen and ax revenue using wo differen co-inegraion ess namely Johansen and Juselius (199), and Haemi-J (28). The resuls of he co-inegraion ess indicae ha direc ax revenue and financial developmen are co-inegraed. The Vecor Error Correcion Model (VECM) used o invesigae he shor run and long-run dynamic relaionship beween financial developmen and direc ax revenue. The resuls reveal ha banking and non-banking financial developmen Granger cause direc ax revenue in he long run. Only he banking secor Granger causes direc ax revenue in he shor run. Keywords: financial developmen; ax revenue; srucural break; co-inegraion; causaliy; Turkey JEL Classificaion: B26; C22; H71; O53 I. Inroducion In he las decade, Turkey has implemened a rade oriened growh model, and has achieved considerable economic performance. Over he las en years, macroeconomic sabiliy has been secured, he inflaion rae has reduced o a single-digi number, he ineres rae has decreased significanly, public deb as a percenage of GDP has diminished, public balance has been achieved and a high primary surplus has been aained. A presen, Turkey is in he op weny larges economies in he world. The srong economic growh has generaed considerable governmen revenue. According o he World Bank (214), beween 2 and 212 general governmen revenues in Turkey increased from abou 27 percen of GDP o almos 33 percen. In erms of ax revenues, beween 24 and 212, direc ax revenue grew by percen while, indirec ax revenue grew by 12.2 percen. The increase in ax revenue can be aribued o several facors including sound fiscal policy, economic growh and financial developmen. There are hree hypoheses abou he linkage beween financial developmen and economic growh in lieraure. Firs, he supply-leading hypohesis holds ha by faciliaing he allocaion of resources, capial accumulaion and diffusion of echnology, financial developmen promoes long-run economic growh. Second, he demand-pulling hypohesis ascerains ha economic growh causes financial developmen. Third, according o he feedback hypohesis financial developmen and economic growh muually influence each oher. In oher words, hey are complemenary. 1 Afyon Kocaepe Universiy, Deparmen of Economics, 327 Afyonkarahisar, Turkey. 2 Afyon Kocaepe Universiy, Deparmen of Public Finance, 327 Afyonkarahisar, Turkey. 3 Afyon Kocaepe Universiy, Deparmen of Economics, 327 Afyonkarahisar, Turkey.

2 14 Journal of Economics and Developmen Sudies, Vol. 4(1), March 216 The financial developmen may influence direc ax revenue in several ways. Firs, economic growh leads o an expansion of axable economic aciviies, which in urn, increases direc ax revenue. Second, economic growh brings prosperiy and booss he demand for goods and services which raises new invesmens. As a resul he income ax base will increase which conribues o direc ax revenues. Third, boh financial developmen and economic growh migh discourage he spread of shadow economy. Finally, financial developmen could direcly increase ax revenues as i faciliaes racking and collecion of axes (see, Bose e al., 212 and Capasso and Jappeli, 213). The purpose of his sudy is o explore he co-inegraion and causaliy beween financial developmen and oal direc ax revenue in Turkey. Financial developmen is associaed wih differen macroeconomic variables in many empirical sudies. However, very few sudies have examined co-inegraion and causal relaionships beween financial developmen and direc ax revenue. Looking a he ax revenue effecs of he financial developmen is a relaively new idea. Taha e al. (213) invesigaed he causal relaionship beween financial sysem aciviies and direc ax revenue for Malaysia and found ha sock marke aciviies Granger causes direc ax revenue. The conribuion of his sudy is wofold. Firs, o our knowledge, his is he firs aemp o probe co-inegraion and causal relaionship beween financial developmen and direc ax revenue in a mulivariae framework in Turkey. Second, his sudy considers boh linear and nonlinear co-inegraion ess. The res of he sudy is oulined as follows. Secion 2 presens model, daa and mehodology, Secion 3 repors on empirical findings, and Secion 4 concludes. 2. Model, daa and mehodology Using he variables proposed by Taha e al. (213) he following model 1,2 is exploied o invesigae he causal relaionship beween direc ax revenue 3 and financial developmen. lnd = f (lncb, lnbi, lnpb ) (1) where d, cb, bi and pb represen he oal direc ax revenue, he oal value of credi provided by commercial banks o he privae secor, Borsa Isanbul s sock marke index (BIST 1), and ousanding privae secor bonds marke, respecively and ln denoes naural logarihm. In his sudy, financial developmen is decomposed ino banking and non-banking financial aciviies. The oal value of credi provided by commercial banks o he privae secor is used as a proxy for banking secor financial developmen. The sock marke index and he ousanding privae secor bonds are used as proxies for nonbanking financial developmen. Monhly ime series daa from January 26 o December 214, for a oal of 18 observaions are used in his sudy. Daa availabiliy forced us o choose his paricular ime period. All he series are in nominal Turkish Lira obained from wo sources. While direc ax, credis, and privae secor bonds daa aken from he Cenral Bank of Turkey s elecronic daa delivery sysem (EDDS) he daa for he sock marke index (BIST 1) is obained from Borsa Isanbul s websie ( Prior o analysis, o reduce he problem of heeroscedasiciy, all of he series are ransformed ino a naural logarihm scale. The daa used in his paper is no seasonally adjused for wo reasons. Firs, seasonal adjusmen enails he smoohing of daa. Second, i has undesirable effecs on procedures involving srucural breaks. Johansen and Juselius (JJ) (199) and Haemi-J (HJ) (28) co-inegraion ess are used o deermine he long-run equilibrium relaionship among variables. While JJ es ignores srucural breaks HJ assumes wo srucural break poins. In he case of HJ procedure, srucural poins are no known in advance bu heir dae is deermined by daa. The JJ es assumes ha a long run equilibrium relaionship is no influenced by he inernal and exernal developmens. The HJ co-inegraion es considers hree models namely Level Shif (C), Level Shif and Trend (C/T) and Regime Shif (C/S). The regime shif (C/S) version of he Haemi J. es can be expressed as follows: y ' ' 1D1 x 1D1 x ' ' ' 1D1 2D2 x 1D1 x 2D2 x y (2) (3)

3 Akçay, Sağbaş & Demiraş 15 In Eq. (2) and (3) D 1 and D 2 denoe dummy variables defined as D 1 = if n 1 1 if n 1 and D 2 = if n 2 1 if n 2 here τ 1 and τ 2 represen unknown parameers belongs o se (,1) indicaing he relaive iming of srucural break poins. The bracke signifies ineger par. See Haemi-J. (28) for he deailed explanaions regarding he esimaion procedures. To ascerain he direcion of causaliy (shor-run and long-run) beween four variables, we use he vecor error correcion model (VECM) specified in Eq.(4) ln d 1 ln cb 2 (1 L) ln bi 3 ln pb 4 s i 1 (1 L) 11i 21i 31i 41i 12i 22i 32i 42i 13i 23i 33i 43i 14i 24i 34i 44i ln d ln cb ln bi ln pb i i i i ( ECT ) Where L is he lag operaor and (1-L) is he difference operaor, s is he lag lengh, ECT -1 is he error correcion erm which represens disequilibrium beween financial developmen and ax revenues, є 1, є 2, є 3 and є 4 are serially uncorrelaed residuals. For insance, if he coefficien of he error erm (π 1) is saisically significan in Eq. (4), one can conclude ha he direcion of causaliy is from cb, bi, and pb o d in he long run. On he oher hand, α 12i, α 13i, and α 14i indicae ha cb, bi, and pb Granger cause d in he shor run. 3. Findings The empirical analysis of he sudy conains hree seps, namely uni roo, co-inegraion and he Granger causaliy ess. Deermining he inegraion order of variables is a prerequisie for he co-inegraion ess. For his purpose, ADF and Zivo Andrews (ZA) uni roo ess are used as he firs sep of he empirical verificaion procedure. The resuls of he ZA es show ha srucural breaks are mosly clusered around he year 28 and 29, which refers o global financial crisis of 28. The resuls abulaed in Table 1 and Table 2 indicaes ha all he variables are saionary as of heir firs differences. Therefore, all of he series are inegraed in an order of one, I(1). Table 1: ADF uni roo es resuls Tes lnd lncb lnbi lnpb ADF c& c& c& c Level -2.1(11) -3.1(3) -2.39() -2.34() Firs difference -6.79(11)*** -3.42(2)** -1.4()*** -9.63()*** Order of inegraion I(1) I(1) I(1) I(1) Noes: *** and ** indicae significance a 1% and 5% levels respecively. c& denoes consan and rend; c denoes consan. The lag lenghs are given in parenhesis and based on SBC. EViews 9. was used for all compuaions. Table 2: Zivo-Andrews es resuls lnd lncb lnbi lnpb c& c& c& c Poin Poin Poin Poin Level -3.18(12) 29:2-4.83(6) 28:1-5.45(3) 29: (1) 21:5 Firs difference -7.65(11)*** 29: (3)*** 29: ()*** 29:3-1.5()*** 28:8 Order of inegraion I(1) I(1) I(1) I(1) Noes: Criical values a 1% and 5% significance level are 5.57 and 5.8 respecively (Zivo and Andrews, 1992). c& denoes consan and rend, c denoes consan. The lag lenghs are given in parenhesis. EViews 9. was used for all compuaions. (4)

4 16 Journal of Economics and Developmen Sudies, Vol. 4(1), March 216 The exisence of coinegraion is invesigaed in he second sep. The JJ and HJ co-inegraion ess require non-saionary series wih idenical orders of inegraion. To his end, JJ and HJ co-inegraion ess are used, and he resuls are repored in Table 3 and 4 respecively. The resul of JJ es shows ha all he series are co-inegraed which means ha here is a long run equilibrium relaionship among he variables. In he case of HJ es, while Ζ * and Ζ α * ess can rejec he null hypohesis of no co-inegraion a a 1% level of significance for models C, C/T and C/S, bu modified ADF es canno rejec he null hypohesis of no co-inegraion a 5% level of significance for all models in HJ procedure. Furhermore, he resul of HJ es provides evidence of srucural breaks vary beween May 27 and Sepember 211. Provided ha all variables are I(1) and co-inegraed, one can expec causaliy a leas in one direcion. Table 3: JJ Coinegraion es resuls Hypohesis Alernaive Trace Criical Value Max Criical Value Prob. Saisics 5% Saisics 5% Prob. r = r= *** *** r 1 r= ** ** r 2 r= r 3 r= Noes: *** and ** indicae significance a 1% and 5%level respecively; r denoes number of co-inegraing vecors. Opimal lag lengh is 2 deermined by Schwarz Informaion Crierion (SIC). EViews 9. is used for all compuaions. Table 4: Haemi-J. coinegraion es resuls Models ADF TB 1 / TB 2 Ζ * TB 1 / TB 2 Ζ α * TB 1 / TB 2 C (11) 29:1 / 211: *** 27:5 / 28: *** 29:7 / 211:1 C/T (11) 28:12 / 211: *** 27:5 / 28: *** 28:6 / 29:5 C/S -5.7(12) 29:6 / 21: *** 27:12 / 28: *** 28:2 / 28:4 Noes: Criical values are available in Haemi-J (28:p51). TB denoes break poin. *** implies significance a 1%. Gauss 1. is used for all compuaions. Lasly, Granger causaliy es based on he VECM is conduced and shor run and long run resuls are documened in Table 5. Of he four equaions only he error correcion erms (ECT -1, ECT -2) in he direc axequaion (Δlnd ) are saisically significan. This resul confirms unilaeral causaliy running from banking and non banking financial developmen o direc ax revenue in he long run. Conrary o he resuls of Taha e al. (213), our resuls indicae ha no only he sock marke bu also commercial bank credis o he privae secor and privae bond marke influence direc ax revenue. Unidirecional causaliy running from banking, sock marke and privae secor bonds marke o direc ax revenue in he long run indicaes ha banking and non-banking financial aciviies maer for direc ax revenue in Turkey. On he oher hand, in he shor run, bi-direcional causaliy beween direc ax revenue and he banking secor is also found. The unidirecional causaliy running from banking o ax revenue boh in he shor and he long run (srong causaliy) show ha he banking secor promoes direc ax revenue. Dependen variable Δlnd - Δlncb 5.26* [.71] Δlnbi Table 5: VECM resuls Sources of causaion Shor run Long run Δlnd Δlncb Δlnbi Δlnpb ECT 1, -1 ECT 2, [.783] 5.45* [.65] [.224] 3.82 [.148].53 [.763] [.149] 1.63 [.441] [.563] *** (-6.31) -.21 (-1.16).114 (.94).88*** (5.58) -.13 (-1.4) -.32 (-.49) Δlnpb - [.235] [.489] [.886] (.67) (.21) Noes: ***, and * indicae significance a 1% and 1% levels respecively. Figures in parenhesis represen he -saisics. Opimal lag lengh is 2 deermined by Schwarz Informaion Crierion (SBC). The join significance of he coefficiens is ascerained by he Wald es. Figures in brackes represen p-value of he Chi-sq es. EViews 9. is used for all compuaions.

5 Akçay, Sağbaş & Demiraş 17 The sabiliy in he coefficiens of he esimaed model (Δlnd ) is checked using a cumulaive sum (CUSUM) and a cumulaive sum of squares (CUSUMQ) sabiliy ess ha employ recursive residuals. The plos of CUSUM and CUSUMQ saisics from he Eq.(4) for he dependen variable (Δlnd ) are depiced in Figure 1. Since he wo saisics are confined wihin he 5% criical bounds, he esimaed coefficiens in he error correcion model are sable over he analysis period, and herefore resuls can be used for policy recommendaions CUSUM 5% Significance CUSUM of Squares 5% Significance 4. Conclusion Figure 1: Plo of CUSUM and CUSUMQ: Dependen variable: Δlnd Using monhly daa from January 26 o December 214 as well as a mulivariae model, his sudy examined he linear and nonlinear co-inegraion and causaliy beween financial developmen (boh banking and nonbanking) and direc ax revenue in Turkey. The empirical findings of he sudy show ha he financial marke developmen and direc ax revenue are co-inegraed. Furhermore, he Granger causaliy ess based on VECM confirmed ha financial developmen Granger cause direc ax revenue in he long run. However, he shor run causaliy es resuls show ha only he banking secor Granger cause direc ax revenue. In summary, hese resuls conclude ha in he las decade, financial developmen has played an imporan role in he direc ax revenue collecion in Turkey. From a policy poin of view, findings of he sudy sugges ha auhoriies in Turkey should coninue policies ha faciliae financial developmen. Noes 1 We include index of indusrial producion as a proxy of economic aciviy in he model, however, i is inegraed order of wo, I(2) and herefore his variable is excluded from he model. 2 Due o he unavailable monhly GDP daa we could no deflae each variable by GDP. 3 We also ried o esimae he model when dependen variable is indirec ax revenue, however, i is found ha indirec ax revenue series is inegraed order of wo, I(2) and herefore we canno proceeded o analysis.

6 18 Journal of Economics and Developmen Sudies, Vol. 4(1), March 216 References Bose, N., Capasso, S. and Wurm, M.A. (212). "The impac of banking developmen on he size of shadow economies" Journal of Economic Sudies, 39(6): doi:1.118/ Capasso, S. and Jappelli, T. (213). "Financial developmen and he underground economy" Journal of Developmen Economics, 11: doi:1.116/j.jdeveco Haemi-J, A. (28). "Tess for coinegraion wih wo unknown regime shifs wih an applicaion o he financial marke inegraion" Empirical Economics, 35(3): doi:1.17/s Johansen, S. and Juselius, K. (199). "Maximum likelihood esimaion and inference on co inegraion wih applicaions o he demand for money" Oxford Bullein of Economics and Saisics, 52(2): doi:1.1111/j mp5223.x Taha R., Colombage, S.R.N., Maslyuk, S., and Nanhakumar, L. (213). "Does financial sysem aciviy affec ax revenue in Malaysia? Bounds esing and causaliy approach" Journal of Asian Economics, 24: doi:1.116/j.asieco World Bank. (214). "Turkey Public Finance Review, Turkey in Transiion: Time for a Fiscal Policy Pivo?", Repor No TR. Washingon, DC: The World Bank Zivo, E., Andrews and Donald, W. K. (1992). "Furher evidence on he grea crash, he oil- price shock, and he uniroo hypohesis" Journal of Business & Economic Saisics, 1(3): doi:1.18/