Tax Structure and Economic Indicators in Developing and High- Income OECD Countries: Panel Cointegration Analysis

Transcription

1 Avalable onlne at ScenceDrect Proceda Economcs and Fnance 7 ( 2013 ) Internatonal Conference on Economcs and Busness Research (ICEBR) 2013 Tax Structure and Economc Indcators n Developng and Hgh- Income OECD Countres: Panel Contegraton Analyss Imbarne Bujang a, Taufk Abd Hakm b *, Ismal Ahmad c a Faculty of Busness Management, Unverst Teknolog MARA, Locked Bag 71, Kota Knabalu, Sabah, Malaysa. b Arshad Ayub Graduates Busness School (AAGBS), Unverst Teknolog MARA, Selangor, Malaysa. c Faculty of Busness Management, Unverst Teknolog MARA, Selangor, Malaysa. Abstract Ths paper nvestgates the long-run relatonshp between tax structure and economc growth and the other economc ndcators va panel unt root tests and panel contegraton analyss. Panel unt root tests are carred out to determne the order of ntegraton of panel varables. In order to fnd out the long-run contegraton relatonshp, we employ the Kao resdual contegraton test. The results of panel contegraton test reveal that there s no long-run contegratng relatonshp between tax structure and both of GDP and gross savng (SAVING) n developng countres. Yet, there s an evdence of a strong contegratng relatonshp among tax structure and nternatonal trade s actvtes. Conversely, for hgh-ncome OECD countres, Kao s test suggests that there s a long-run contegratng relatonshp between components of tax revenue and GDP and gross savng, whle there s no evdence for mported and exported of goods and servces (nternatonal trade) The Authors. Publshed by by Elsever B.V. B.V. Selecton and peer-revew under responsblty of ICEBR of ICEBR Keywords: Developng countres; countres; Tax structure; Economc ndcators; Contegraton 1. Introducton For the past decades, the ssue and debate of taxaton and long-run growth has attracted the attenton of many researchers n the dfferent areas of studes. Emprcal studes conducted by Stokey and Rebello, 1995, and Mendoza et al., 1997, revealed that tax polcy has small effect on long-run growth. On the other hand, * Correspondng author. Emal address: taufkabdhakm@yahoo.com The Authors. Publshed by Elsever B.V. Selecton and peer-revew under responsblty of ICEBR 2013 do: /s (13)

2 Imbarne Bujang et al. / Proceda Economcs and Fnance 7 ( 2013 ) Bleaney et al., 2001, expressed that both of ncome and consumpton taxes have a clear effect on economc growth n the long-run. The lterature has wtnessed that taxaton also affect the other mportant economc ndcators. Avla and Strauch, 2008, who mplemented the endogenous growth model revealed that taxaton had negatve mpact on nvestment and led to slowdown the economc growth. Usng pooled tme seres and cross-sectonal analyses n South and East Asan, Kerr and Monsgh, (1998a), argued that there s a sgnfcant nverse relatonshp between tax-mx (drect and ndrect taxes) and the level of savng. More recently, Berger and Everaert, 2010, measured the mpact of labor taxes on unemployment n 16 OECD countres and dvded the countres nto European, Nordc and Anglo-Saxon countres. They concluded that labor taxes have postve relatonshp wth unemployment n European and Nordc countres. However, they dd not fnd any evdence to relate labor taxes and unemployment n Anglo-Saxon. Hence, two mportant ssues are addressed n ths study; (1) Do taxes have a long-run relatonshp wth economc growth and the other economc ndcators? (2) Is the long-run relatonshp between tax revenue and economc ndcators s consstent among developng and hgh-ncome OECD countres? The remander of the paper s organzed as follows: Secton 2 dscusses theoretcally how taxes affect the economc growth by adoptng the growth model. Secton 3 descrbes the data and methodology employed n ths study. Secton 4 provdes results of unt root and contegraton analyses for both developng and hghncome OECD countres. The last secton concludes. 2. Taxes n the growth model The neoclasscal growth model whch was ntroduced by Solow, 1956, suggested that the long-run growth rate s drven by the rate of techncal progress and populaton growth. The ncentve to nvest n human or physcal captal may be affected by productve government expendtures or dstortonary taxaton. However, n the long run, ths affects only the equlbrum factor ratos but not the growth rate. On the other hand, endogenous growth models whch was ntroduced by Barro, 1990, and Kng and Rebelo, 1990, predcted that the long run growth wll be affected by productve expendtures and dstortonary taxaton. Roháč, 2009, llustrated the growth rate of an economy as follows: y k m (1) where y on the left sde represent the change n output as fracton of GDP, k donates the change n the captal stock as a fracton of GDP and m s the percentage change n workforce or employee overtme. The resdual, μ stands for the overall change n economc productvty, whle both α and β measure the margnal productvty of captal and the elastcty of labour. Engen and Sknner, 1996, proposed fve ways on how taxes can nfluence the economc performance. The frst assumpton s nvestment, k can be reduced by mposng taxes on ncome, corporate and captal gans. Secondly, workng hours or workforce partcpaton, m also wll be nfluenced by the change n these taxes. Thrd, taxaton wll gve negatve mpact on the productvty growth, μ by attenuatng the venture captal nvestment to hgh technology ndustres and research and development. The fourth way how taxes can nfluence the movement of growth s the margnal productvty of captal, α can be nfluenced by taxes f the frm change the nvestments from hghly taxed to low tax nvestment whle fnally, taxaton can reduce margnal productvty of labour, β by dscouragng labours from workng n hgh productvty sectors but more tax burden.

3 166 Imbarne Bujang et al. / Proceda Economcs and Fnance 7 ( 2013 ) Data and methodology Our sample conssts of balanced panel data for 24 developng countres and 24 hgh-ncome OECD countres over the perod (early of the modern era). All data are gathered from World Bank Development Indcators (World Bank, 2012). The rate of change n a country s GDP (GDP), gross savng (SAVING), net nflow of foregn drect nvestment (FDI), mports and exports of goods and servces (IMPORT and EXPORT) and unemployment rates (UNEMP) are the economc ndcators chosen by ths study. The ndependent varables conssts of prmary sources of tax revenue whch ncludes the total tax revenue to GDP rato (TAX), taxes on ncomes, profts and captal gans (INCOMETX), taxes on goods and servces (GOODTX), taxes on nternatonal trade (INTERTX), other taxes (OTHERTX) and hghest margnal tax rates (MTR). We follow the measurement of these varables by Gober and Burns (1997) and Wdmalm (2001) whch dvdes tax structure wth total tax revenue (TAX). However, due to avod the bas towards fndng a contegraton relatonshp, MTR s not dvded wth TAX. The summary statstcs of varables for developng and hgh-ncome OECD countres s presented n Table 1. Table 1. Summary statstcs of varables n levels ( ) Varables Groups of countres Observatons Mean (%) Mn (%) Max (%) Std. Dev. GDP Developng FDI Developng SAVING Developng IMPORT Developng EXPORT Developng UNEMP Developng INCOMETX Developng GOODTX Developng INTERTX Developng OTHERTX Developng MTR Developng TAX Developng

4 Imbarne Bujang et al. / Proceda Economcs and Fnance 7 ( 2013 ) Data on tax structure and economc ndcators n developng vs. hgh-ncome OECD countres Table 1 compares the sources of tax revenue and economc ndcators among both of countres. To begn wth, as seen n the Table, the developng countres shows greater economc performance compared wth the hgh-ncome OECD countres, as measured by the annual growth rate of GDP. It ndcates that, even though most of the hgh-ncome countres have the advanced technologes and hgh standard of lvng, but recently, they faled to generate better economc growth. Nevertheless, the hgh-ncome OECD countres succeed on collectng the larger tax revenue compared wth the developng countres. As shown n the Table, the hghncome OECD countres collect more than double collected of the total tax revenue n the developng countres, as a fracton of GDP. It proves that the hgh-ncome OECD countres have the hghest per capta ncome and many bg companes wth large profts to be taxed compared wth developng countres. Among the hgh-ncome OECD countres, the man sources of tax revenue are the taxes on ncomes, profts and captal gans (28.99 per cent of total tax revenue) and taxes on goods and servces (27.84 per cent of total tax revenue). Taxes on goods and servces are even more mportant among developng countres (36.99 per cent of total tax revenue). Recently, developng and hgh-ncome OECD countres have encouraged the actvtes on mports and exports. Both groups of countres have reduced the barrers on nternatonal trade by mplement the lower tax rate on mported and exported goods and servces as well as free taxes on nternatonal trade. It shown by the large percentage of both mports and exports of goods and servces (43.01 per cent per cent and per cent per cent of GDP) whle low percentage on nternatonal trade tax (1.04 per cent and 6.21 per cent of total tax revenue) n both groups of countres. The lower fracton of GDP collected n tax revenue among the developng countres does not seem to be a result of ther choosng lower statutory tax rates. The hghest margnal tax rates for both groups of countres whch shows close same rate (28.67 per cent and per cent) proves that tax rates are not determne the total tax revenue collected. We suggest the lower fracton of GDP collected by these taxes are nfluenced by neffectve or over optmum level of tax rates whch employed by the polcy makers as well as lower per capta ncome and less of medum and large companes operated n developng countres compared wth the hgh-ncome OECD countres. 3.2 Emprcal model In order to nvestgate the long-run relatonshp between tax revenue and economc ndcators, the followng model was used: y ' ' t TX t t (2) where y represents the economc ndcators n the th country at tme t. The economc ndcators are y 1 = GDP, y 2 = SAVING, y 3 = FDI, y 4 = IMPORT, y 5 = EXPORT, and y 6 = UNEMP. The explanatory varables TX consst of the prmary components of tax revenue n both groups of countres. Ths study uses several panel unt root tests to determne the order of ntegraton of these varables. We adopt the more relable and well-behaved panel unt root tests such as the Levn et al., 2002; Im et al., 2003, and Fsher-type tests proposed by Maddala and Wu, 1999, and Cho, 2001, that based on augmented Dckey- Fuller (ADF) and Phllps-Perron (PP). Based on Monte Carlo results, Guterrez, 2003, reveals that the Kao resdual contegraton test s outperform Pedron s n the case of small tme seres dmenson of the panel. Thus, we employ Kao s, 1999, test to fnd out the long-run relatonshp between economc ndcators and tax structure n both developng and hgh-ncome OECD countres.

5 168 Imbarne Bujang et al. / Proceda Economcs and Fnance 7 ( 2013 ) Emprcal results 4.1 Panel unt root tests The analyss begns wth an examnaton of the ntegraton propertes of the panel varables ncluded n each model. Table 2 and Table 3 reports panel unt root tests estmate for developng and hgh-ncome OECD countres. As can be seen n Table 2, most of the varables do not exhbt unt roots n level form except for GDP, gross savng (SAVING), total tax revenue (TAX) and nternatonal trade s actvtes (IMPORT and EXPORT) n developng countres. The other varables are clearly statonary at levels. For hgh-ncome OECD countres, the test statstcs suggest that GDP, SAVING, EXPORT, unemployment rates (UNEMP) and hghest margnal tax rates (MTR) have a unt root whch mples nonstatonary. Table 2. Results of panel unt root tests for developng countres (levels) Varable LLC IPS Fsher (ADF) Fsher (PP) GDP FDI SAVING IMPORT EXPORT UNEMP INCOMETX GOODTX INTERTX OTHERTX MTR TAX *** *** (0.0005) (0.4260) *** *** *** *** *** *** *** *** (0.0009) (0.4228) (0.9991) (0.2582) *** (0.0004) ** (0.0134) *** *** *** (0.2189) (0.7695) *** ** (0.0139) (0.9999) (0.1672) * (0.0836) (0.2186) *** (0.0089) *** *** *** *** *** (0.0010) (0.7117) (0.6855) (0.5287) *** *** *** *** (0.1159) *** *** (0.9766) Notes: Values n parentheses are p-values. ***, ** and * ndcates rejectng the null hypothess of non-statonary at 1 per cent, 5 per cent and 10 per cent level, respectvely. Lag length (2) for LLC and IPS are determned by Akake Informaton Crteron (AIC), whle Fshertype tests are based on ADF and Newey-West. Constant and trend terms are ncluded n all tests. The probabltes for the Fsher-type tests are computed usng an asymptotc Ch-square dstrbuton.

6 Imbarne Bujang et al. / Proceda Economcs and Fnance 7 ( 2013 ) Table 3. Results of panel unt root tests for hgh-ncome OECD countres (levels) Varable LLC IPS Fsher ADF Fsher (PP) GDP FDI SAVING IMPORT EXPORT UNEMP INCOMETX GOODTX INTERTX OTHERTX MTR TAX *** *** (0.0080) *** * (0.0724) *** *** *** *** *** * (0.0580) *** *** (0.8108) *** (0.0037) (0.3262) (0.1078) *** (0.0045) *** *** (0.0002) *** *** (0.0027) *** *** (0.9217) *** * (0.0887) 100.5*** *** *** *** *** (0.1523) *** *** (0.9996) *** *** *** *** (0.5951) *** Notes: Values n parentheses are p-values. ***, ** and * ndcates rejectng the null hypothess of non-statonary at 1 per cent, 5 per cent and 10 per cent level, respectvely. Lag length (2) for LLC and IPS are determned by Akake Informaton Crteron (AIC), whle Fshertype tests are based on ADF and Newey-West. Constant and trend terms are ncluded n all tests. The probabltes for the Fsher-type tests are computed usng an asymptotc Ch-square dstrbuton. The same panel unt root tests were conducted by transformng all varables nto frst dfference as shown n Table 4 and Table 5. All the tests show that the varables are statonary at frst dfferences. Hence, we conclude that frst dfferenced varables are statonary so that panel varables are ntegrated of order one, I (1). Table 4. Results of panel unt root tests for developng countres (frst dfferences) Varable LLC IPS Fsher (ADF) Fsher (PP) GDP FDI SAVING IMPORT *** *** *** *** *** (0.0003) *** (0.0002) *** -7.1*** *** *** *** *** *** *** *** ***

7 170 Imbarne Bujang et al. / Proceda Economcs and Fnance 7 ( 2013 ) EXPORT UNEMP INCOMETX GOODTX INTERTX OTHERTX MTR TAX *** *** *** *** * (0.0538) *** *** *** *** (0.0093) ** (0.0427) *** *** (0.0006) *** (0.0018) *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** *** Notes: Values n parentheses are p-values. ***, ** and * ndcates rejectng the null hypothess of non-statonary at 1 per cent, 5 per cent and 10 per cent level, respectvely. Lag length (2) for LLC and IPS are determned by Akake Informaton Crteron (AIC), whle Fshertype tests are based on ADF and Newey-West. Constant and trend terms are ncluded n all tests. The probabltes for the Fsher-type tests are computed usng an asymptotc Ch-square dstrbuton. Table 5. Results of panel unt root tests for hgh-ncome OECD countres (frst dfferences) Varable LLC IPS Fsher ADF Fsher (PP) GDP FDI SAVING IMPORT EXPORT UNEMP INCOMETX GOODTX INTERTX ** (0.0111) *** *** (0.0080) *** *** *** (0.0002) *** *** *** *** *** *** *** *** ** (0.0438) *** *** *** *** (0.0006) *** *** *** *** * (0.0682) *** *** *** *** *** *** *** (0.0047) *** (0.0078) * (0.0943) *** *** ***

8 Imbarne Bujang et al. / Proceda Economcs and Fnance 7 ( 2013 ) OTHERTX MTR TAX *** *** *** *** *** (0.0027) *** *** *** *** *** *** *** Notes: Values n parentheses are p-values. ***, ** and * ndcates rejectng the null hypothess of non-statonary at 1 per cent, 5 per cent and 10 per cent level, respectvely. Lag length (2) for LLC and IPS are determned by Akake Informaton Crteron (AIC), whle Fshertype tests are based on ADF and Newey-West. Constant and trend terms are ncluded n all tests. The probabltes for the Fsher-type tests are computed usng an asymptotc Ch-square dstrbuton. 4.2 Panel contegraton analyss Snce all panel varables are found to be ntegrated of order one I (1), we can proceed wth the panel contegraton test usng frst dfferenced varables to fnd out the long-run relatonshp between tax structure and economc ndcators. Table 6 exhbts the results of Kao resdual contegraton test among the followng models: GDP, SAVING, FDI, IMPORT, EXPORT and UNEMP. Table 6. Results of Kao resdual contegraton test Model Regresson Developng Countres Countres GDP SAVING FDI IMPORT EXPORT UNEMP (0.1712) (0.3022) ** (0.0220) *** (0.0049) * (0.0695) * (0.0787) *** ** (0.0195) (0.1262) (0.1161) (0.2719) (0.1659) Notes: The value of t-statstc s determned by Augmented Dckey Fuller (ADF). Lag length (1) s computed by Newey-West bandwdth selecton and Bartlett kernel. Values n parentheses are p-values. ***, ** and * ndcates rejectng the null hypothess of no contegraton at 1 per cent, 5 per cent and 10 per cent level, respectvely. The test demonstrates mxed results for both groups of countres. The frst nspecton of Table 6 reveals that three man components of economc ndcators n developng countres, GDP, gross savng (SAVING) and unemployment rate (UNEMP) are faled to reject the null of no contegraton, even at the 10 per cent level. Yet, the reported statstcs suggest the presence of a strong contegratng relatonshp among tax structure and nternatonal trade s actvtes (IMPORT and EXPORT). In contrast, for hgh-ncome OECD countres, models of GDP and SAVING are succeeds to reject the null of no contegraton at the 10 per cent and 1 per cent level. Despte that, Kao s test suggests that there s no long-run contegratng relatonshp

9 172 Imbarne Bujang et al. / Proceda Economcs and Fnance 7 ( 2013 ) between tax structure and mports and exports on goods and servces. Nevertheless, test results support a contegraton relaton among tax structure and net nflow of foregn drect nvestment (FDI) for both developng and hgh-ncome OECD countres. Conversely, there s no evdence of contegraton exsts for model of UNEMP n both groups of countres. 5 Conclusons Ths paper nvestgates the long-run equlbrum relatonshp between components of tax revenue and economc ndcators. The result based on Kao resdual contegraton test reveals that there s an nconsstent long-run contegratng relatonshp between tax structure and economc ndcators for both groups of countres. For hgh-ncome OECD countres, we fnd that there s a long-run contegratng relatonshp between tax structure on three economc ndcators (GDP, gross savng and FDI). It ndcates that tax polces mplemented by hgh-ncome OECD countres were affectng the movement of long-run growth. Nevertheless, the results suggest that there s no evdence of long-run relatonshp among tax structure and the sze of nternatonal trade. In contrast, for developng countres, the reported statstc shows the presence of long-run contegratng relatonshp between taxes and nternatonal trade, but there s no evdence for GDP and gross savng. Thus, the fndngs n ths paper suggests that dfferent tax polces conducted by developng countres and hgh-ncome OECD countres have gven an opposte long-run effect on economc ndcators among these groups of countres. Addtonally, ths study concludes that tax can be an mportant tool to recover the current recesson or economc slowdown and contrbute to long-run growth and nternatonal trade n both groups of countres. Acknowledgement We thank the anonymous referee for the valuable comments and suggestons. The usual dsclamer apples. References Avla, D.R., Strauch, R., Publc fnances and long term growth n Europe: Evdence from a panel data analyss. European Journal of Poltcal Economy 24, p Barro, R.J., Government spendng n a smple model of endogeneous growth. The Journal of Poltcal Economy 98, p Berger, T., Everaert, G., Is the mpact of labour taxes on unemployment asymmetrc? Journal of Economc Dynamcs and Control 35, p Bleaney, M.F., Gemmell, N., Kneller, R., Testng the endogenous growth model: Publc expendture, taxaton and growth over the long-run. Canadan Journal of Economcs 34, p Cho, I., Unt root tests for panel data. Journal of Internatonal Money and Fnance 20, p Engen, E.M., Sknner, J., Taxaton and economc growth. Natonal Tax Journal 49, p Gober, J.R., Burns J.O., The Relatonshp between tax structure and economc ndcators. Journal of Internatonal Accountng, Audtng & Taxaton 6, p. l-24. Guterrez, L., On the power of panel contegraton tests: a Monte Carlo comparson. Economc Letters 80, p Im, K. S., Pesaran, M.H., Shn, Y., Testng for unt roots n heterogeneous panels. Journal of Econometrcs 115, p Kao, C., Spurous regresson and resdual-based tests for contegraton n panel data. Journal of

10 Imbarne Bujang et al. / Proceda Economcs and Fnance 7 ( 2013 ) Econometrcs 90, p Kerr, I., Monsngh, V., 1998a. Taxaton mx and tax polcy n developng economes. School of Economcs and Fnance Workng Paper No. 9801, Curtn Unversty, Perth, Western Australa. Kng, R.G., Rebelo, S., Publc polcy and economc growth: Developng neoclasscal mplcatons. The Journal of Poltcal Economy 98, p Levn, A., Ln, C.F., Chu, J., Unt root tests n panel data: Asymptotc and fnte-sample propertes. Journal of Econometrcs 108, p Maddala, G., Wu, S., A comparatve study of unt root tests and a new smple test. Oxford Bulletn of Economcs and Statstcs 61, p Mendoza, E.G., Mles-Ferrett, G.M., Asea, P., On the neffectveness of tax polcy n alterng long run growth: Harberger s superneutralty conjecture. Journal of Publc Economcs 66, p Roháč, D., Taxaton and economc growth: Reconclng ntuton and theory. Insttute for Research n Economc and Fscal Issues, p Solow, R.M., A contrbuton to the theory of economc growth. Quarterly Journal of Economcs, p Stokey, N.L., Rebello S., Growth effects of flat-rate taxes. Journal of Poltcal Economy 103, p Wdmalm, F., Tax structure and growth: Are some taxes better than others. Publc Choce 107, p World Bank., World Development Indcators. World Bank, Washngton, D.C. Appendx A. Lst of countres n sample Developng Countres Brazl Bulgara Chle Chna Colomba Egypt, Arab Georga Guatemala Inda Indonesa Jordan Kazakhstan Latva Lthuana Malaysa Moldova Morocco Pakstan Paraguay Peru Phlppnes Tunsa Ukrane Uruguay Countres Australa Austra Belgum Canada Czech Republc Denmark Estona Fnland France Germany Greece Hungary Iceland Ireland Italy Korea Republc Luxemburg Netherland Norway Poland Portugal Slovena Span Unted States of Amerca