Is Low Responsiveness of Income Tax Functions to Sectoral Output an Answer to Sri Lanka s Declining Tax Revenue Ratio?

Transcription

1 Is Low Responsiveness of Income Tax Funcions o Secoral Oupu an Answer o Sri Lanka s Declining Tax Revenue Raio? P.Y.N. Madhushani and Ananda Jayawickrema Deparmen of Economics and Saisics, Universiy of Peradeniya, Sri Lanka Keywords: Tax elasiciy; Srucural change; Declining ax raio Inroducion From 1950s o he presen, he economic srucure of Sri Lanka has changed significanly.six or seven decades ago he counry was highly dependen on is agriculural secor. Bu he GDP of he counry oday is mosly generaed in service and indusrial secors. Since domesic agriculure is a small scale subsisence secor, one may no expec high conribuion of ha secor o he ax revenue. Bu he growh of indusrial and service secors resuls in large corporae secor and raises he income of employees of he secors. Therefore, he expansion of services and indusrial secor is expeced o generae high posiive impac on he ax revenue, especially income ax revenue. This sudy focuses on long-erm ax elasiciies of income axes of Sri Lanka in relaion o changes in he secoral composiion of oupu. Tax buoyancy is an indicaor o measure he efficiency and responsiveness of revenue mobilizaion in response o growh in he ax base, GDP or naional income. A ax is said o be buoyan if ax revenue increases more han proporionaely in response o a rise in he ax base. If he ax revenue shows less responsiveness o ax base, ha ype of axes fails o generae enough revenue for he governmen in he long run. Ahmad (1994) finds ha he expansion in agriculural and service secors is no very imporan in he deerminaion of ax revenue. Bu he indusrial secor growh repors a posiive impac on ax revenue collecion of developing counries. Tancy and Zee (2001) 9

2 find he uncerainy of agriculural income and informal secor income as a reason for low ax collecion in developing counries. Indrarane (2003) cies ax evasion, ax avoidance and ax incenives as major reasons for he low responsiveness of axes o he naional income. Waidyasekar (2004) and Jayawickrama (2008) reveal ha he buoyancy of corporae income ax and goods and service axes are low and ha caused a decline rend in he average ax raio. There are many sudies ha focused on ax buoyancy of Sri Lanka using GDP as he ax base. No sudies used segregaion of ax based, as agriculure, indusry and services incomes o examine he responsiveness of axes o changes in secoral oupu. Objecive The objecive of his sudy is o examine he responsiveness of income ax funcions of Sri Lanka wih respec o changes in he composiion of GDP in order o find an answer o declining rend in he ax revenue raio. Mehodology This sudy esimaes personal income ax and corporae income ax funcions of Sri Lanka using annual daa from 1973 o In his sudy, ax revenue is assumed o be a funcion of ax base ( y ), average of marginal ax raes ( τ ) and some unobserved facors ( u ). In order o find he elasiciy parameer wih respec o ax base, he sudy uses he following logarihmic form in he esimaion: ln( ax ) = τ + u.(1) β 0 + β1 ln( y ) + β 2 The parameer β in (1) can be inerpreed as long-erm income (axbase) elasiciy of ax revenue. The income elasiciy of ax is imporan 1 in deermining he buoyancy of a ax or a ax sysem and ax buoyancy requires β in (1) o be greaer han one. This sudy separaes 1 y in (1) o 10

3 agr y, ind y and ser y which represen agriculural, indusry and service secor oupu respecively. Wih his modificaion (1) can be wrien as agr agr ind ind ser ser β 0 + β1 ln( y ) + β1 ln( y ) + β1 ln( y ) + β 2 ln( ax ) = τ + u...(2) agr In which β, 1 ind β1 ser and β are secoral-oupu elasiciy of ax revenue. 1 Since he regressan and regressor variables in (2) are non-saionary, he Leas Squares mehod of esimaion does no produce reliable esimaes for parameers of (2). However, he long-erm parameers of (2) can be derived from a dynamic model in a form of an auoregressive disribued lag (ADL) model. Since (2) conains I(1) variables, an I(0) error erm indicaes ha (2) is a coinegraing regression in which he leas squares parameer esimaes are super-consisen. The ADL ype dynamic model generaes long-erm (coinegraing) soluions wih sandard and F disribuions (Pesaran and Shin, 1998). Resuls and Discussion Equaion (2) is esimaed using personal income ax revenue and corporae income ax revenue as dependen variables. The real values of income axes and secoral oupu are obained by deflaing hem by he GDP deflaor. The ax raes used in he wo ax funcions are average personal income ax rae and he average corporae income ax rae. Tables 1 and 3 give he ADL esimaion of personal and corporae income ax funcions (see in Annexure A). The lag lengh of he esimaed models is se o one by AIC and SC model selecion crieria. The esimaed models were subjec o ess on auo-correlaion, auoregressive condiional heeroscedasiciy, normaliy of he error erm, heeroscesdasiciy and regression specificaion errors. Tables 2 and 4 give he long-erm relaionships of he relevan ax funcions. In he dynamic model of he personal income ax funcion, a level dummy was included for he year 2012 o hold an unprediced fall in 11

4 ax revenue. The model has an impressive predicive power and passes all diagnosic ess. The error correcion uni roo es saisic (= ) indicaes ha he saic soluion is a coinegraing one. As in Table 2, he average personal income ax rae has a highly significan posiive impac on he revenue of personal income ax. The real income of he agriculural secor is no a deerminan of he personal income ax revenue. This resul is quie meaningful as small scale agriculure does no influence income axaion. As indicaed by he responsiveness parameer of indusrial secor oupu, one percen increase in indusrial income will resul in 0.28 percen increase in personal income. On he oher hand, one percen increase in service secor income brings abou 0.58 percen increase in he personal income ax revenue. This impac is wice higher han he impac of he indusrial secor. This is also ineresing as mos of he income ax payers work in service providing eniies of he public and privae secors. Compared o service secor, he expansion of he indusrial secor does no have a large impac on personal income ax. To sum up, he above resuls indicae ha he personal income ax does no increase a he same rae of economic growh. Table 3 gives he resuls of he dynamic model of corporae income ax revenue. The model selecion crieria permi us limiing he lag lengh o one. The model explains more han 99 percen of he shor-erm variaion of he corporae income ax revenue. Dummy variables are used o conrol for unexplained flucuaions in 1980 and The diagnosic ess of model indicae ha here is no issue in he model. The agriculural secor is insignifican in explaining corporae income ax oo. The indusrial secor oupu has a highly significan posiive impac on he corporae income ax revenue. One percen increase in indusrial secor income increases corporae income ax by 0.55 percen. On he oher hand, service secor income has a marginal impac on corporae income ax, one percen increase in service secor oupu increases corporae income ax revenue by only 0.29 percen. 12

5 This effec is significan only a 10%. Therefore, corporae ax revenue shows a low responsiveness o service secor oupu. This low responsiveness of corporae income ax o service secor oupu could be a reason for declining ax raio in Sri Lanka. Since he service secor is he fases growing secor in he economy a presen, one may wonder why he expansion of servicessecor has only a marginal impac on corporae income ax revenue collecion. Conclusion The sudy finds ha agriculural secor expansion has no real impac on personal income ax and corporae income ax collecions. The nonresponsiveness of income axes o agriculural real oupu can be one reason for non-buoyancy and perhaps for he decrease in ax revenue o GDP raio. Personal income ax has a higher responsiveness o service secor income han he indusrial secor income. I may lead o one doub ha mos of he service secor income is disribued among is employees and more of indusrial secor income is accumulaed wihin corporaions as undivided profis. The accumulaion of corporae profis wihin companies may be a reason for relaively high responsiveness of corporae ax revenue o indusrial secor oupu. The less responsiveness of corporae income ax o service secor income quesions he srucure of he service providing companies in Sri Lanka. The non-accumulaion of corporae profis wihin service providing companies may be a reason for low responsiveness. As he service secor becomes he fases growing secor in Sri Lanka, his less responsive may resul in non-buoyancy and a declining ax revenue o GDP raio. The resuls of his sudy have imporan implicaions for ax policymakers. Aenion should be given o improve he efficiency in ax sysem in response o changes in he srucure of he economy. 13

6 References Ahmad, QM 1994, The Deerminans of Tax Buoyancy: An Experience from he Developing Counries, The Pakisan Developmen Review, vol. 33, no. 4, pp Indrarana, Y 2003, The Measuremen of Tax Elasiciy in Sri Lanka: A Time Series Approach, Saff Sudies (Cenral Bank of Sri Lanka), vol. 11, no.33, pp Jayawickrama, JMA 2008, An Examinaion of he Resiliency of Sri Lanka s Tax Sysem, Souh Asia Economic Journal, vol. 9, no. 2, pp Pesaran, MH and Shin, Y 1998, An Auoregressive Disribued Lag Modeling Approach o Coinergraion Analysis, In S. Seinar (ed.) Economerics and Economic Theory in he 20 h Cenury, The Ragnar Frisch Cenennial Symposium, Cambridge, Cambridge Universiy Press, pp Waidyasekara, DDM 2004, Curren Fiscal policy. In Kelegama, S. (ed.) Economic Policy in Sri Lanka: Issues and Debaes. Colombo, VijihaYapa Publicaion, pp Appendix Table 1: Regression Resuls: ADL(1,1) Model of Personal Income Tax Dependan variable, ln(pi) Variable Coefficien value Prob of value Consan ln[pi (-1)] ln[y agri ()] ln[y agri (-1)] ln[y ind ()] ln[y ind (-1)] ln[y ser ()] ln[y ser (-1)] τpi() τpi(-1) R ARCH 1-1 F [0.757] 14

7 es DW 2.09 Normaliy Chi [0.227] es n 40 Heero F es [0.597] AR 1-2 F es [0.216] RESET F es [0.844] es saiscs are given in brackes. Table 2: Solved Long-run Soluion of he Personal Income Tax Funcion Dependen variable, ln(pi) Variable Coefficien value Prob of value Consan ln[y agri ()] ln[y ind ()] ln[y ser ()] τpi() Long-run sigma Error Uni Roo *** Tes Wald es Chi (0.000) es saiscs are given in brackes. *** Significan a 1% Table 3: Regression Resuls: ADL(1,1) Model of Corporae Income Tax Dependan variable, ln(ci) Variable Coefficien value Prob of value Consan ln[ci (-1)] ln[y agri ()] ln[y agri (-1)] ln[y ind ()] ln[y ind (-1)] ln[y ser ()] ln[y ser (-1)] τci() τci(-1) R ARCH 1-1 F [0.229] 15

8 es DW 1.95 Normaliy Chi [0.469] es n 40 Heero F es [0.722] AR 1-2 F es [0.404] RESET F es [0.196] es saiscs are given in brackes. Table 4: Solved Long-run Soluion of he Personal Income Tax Funcion Dependen variable, ln(ci) Variable Coefficien value Prob of value Consan ln[y agri ()] ln[y ind ()] ln[y ser ()] τci() Long-run sigma Error Uni Roo * Tes WALD es Chi (0.000) es saiscs are given in brackes. *** Significan a 1%. 16