The revenue implications of trade liberalization in Tanzania

Transcription

1 Journal of World Economic Research 2014; 3(3): Published online Augus 10, 2014 (hp:// doi: /j.jwer ISSN: X (Prin); ISSN: (Online) The revenue implicaions of rade liberalizaion in Tanzania Manamba Epaphra Deparmen of Accouning and Finance, Insiue of Accounancy Arusha, Arusha, Tanzania address: To cie his aricle: Manamba Epaphra. The Revenue Implicaions of Trade Liberalizaion in Tanzania. Journal of World Economic Research. Vol. 3, No. 3, 2014, pp doi: /j.jwer Absrac: This paper examines he argumen ha rade liberalizaion depresses he impor duy revenue, and consequenly adversely affecs he oal ax revenue. The sudy is hough o be significan because Tanzania experiences difficuly in replacing impor duy revenue loss as a consequence of rade reform by srenghening is consumpion ax sysem. In he course of analysis, coinegraion analysis and error correcion modelling are employed over he 1979/ /10 period. The empirical resuls show ha impor duy revenue-o- raio is posiively relaed o ariff raes, implying ha a reducion in he ariff raes resuls in a significan loss of impor duy revenue. The resuls also show ha he removal of proecionis policies led o an increase in impor-o- raio which in urn led o rising shares of impor duy revenue in. Finally, he resuls generae some policy implicaions. The proper issue in ax design under rade liberalizaion, Tanzania needs o srenghen he domesic ax sysem and raise ax revenue wihou increasing ax raes by reinforcing ax and cusoms adminisraions so as o mainain fiscal sabiliy. Keywords: Impors, Impor Duy Revenue, Trade Liberalizaion, Tanzania 1. Inroducion Many represenaives of developmen insiuions and mos academics believe ha rade liberalizaion, combined wih oher economic and srucural reforms, simulae compeiion, enhance produciviy, and accelerae economic growh. Research by he World Bank, for example, shows ha per capia income in developing counries inegraing rapidly ino he global economy grows more han hree imes as fas as in oher developing counries (Dollar and Kraay, 2001). Rapid growh in many Lain American counries in he lae 1980s and 1990s came abou wih domesic policy liberalizaion and open-economy models, which reduced rade barriers. There is herefore, a wide and growing consensus regarding he need for all counries especially developing counries, o adop a liberalized rade regime. Tanzania being a member of World Trade Organizaion (WTO) could no avoid he increased global inegraion. However, a significan concern, as i conemplaes furher rade liberalizaion, is he poenial impac of rade liberalizaion on ax revenues, and he core issue being is abiliy o subsiue domesic sales axes such as VAT and excise axes for impor duies. Trade liberalizaion is mos likely o pose serious fiscal problems o Tanzania as i relies on inernaional rade axes for a large share of is oal revenue. Trade liberalizaion involves progressive eliminaion of ariffs and a he limi may push rade ax revenue o zero. Available evidence from previous sudies suggess ha revenue recovery has been a real issue for many developing counries. A bes, hey have on average recovered no more han 30 cens of each los dollar (Baunsgaard and Keen, 2005). This explains in par he hesiance of many counries o underake rade liberalizaion and he limiaions o heir susainabiliy. Many developing counries have implemened rade reforms while avoiding significan revenue losses. The prospecs for furher rade liberalizaion are likely o depend in par, on he exen o which hey will cause rade ax revenue o decline furher and wheher counries are able o deal wih such revenue losses. Arguably, he IMF and World Bank have unil now in almos all cases pressed rade liberalizaion up on low-income counries as par of srucural adjusmen by asking for he removal of ariffs. Thus, he main issue should no be he reducion of revenue from rade bu of oal ax revenue. If he fall in he rade

2 26 Manamba Epaphra: The Revenue Implicaions of Trade Liberalizaion in Tanzania axes had been easily compensaed by increase in oher axes, mos obviously by srenghening domesic indirec axes, hen hese would no be a problem. Tanzania is a good case sudy o analyze hese challenges since Tanzania s reliance on inernaional rade axes remains higher, relaive o many oher sub-sahara African counries (DiJohn, 2010). The main objecive of his paper is o examine he impac of rade liberalizaion on impor duy revenue in Tanzania. Specifically, he sudy aims a examining he effec of rade liberalizaion-proxied by a decline in colleced ariff raes and accompanying macroeconomic changes such as exchange rae, per capia and impors on impor duy revenue. I also examines he relaionship beween impor duy revenues and domesic ax revenues (domesic consumpion axes and income axes) o uncover wheher Tanzania can replace impor duy revenue loss from domesic sources. The sudy is imporan because here has been relaively lile empirical research on he impac of rade liberalizaion on ax revenue in Tanzania. Sudies of he exen o which impor duy revenue responds o rade liberalizaion and ha los rade ax revenues have been recovered from domesic sources reach somewha differen conclusions, reflecing differences in counries and ime samples, mehodologies and in he source of he revenue daa. The sudy fills his gap in informaion by piecing ogeher empirical evidence on some aspecs of he Tanzanian ax sysem. Tanzania has commied iself o a process of rade liberalizaion in he framework of he World Trade Organizaion. I is herefore, imporan o anicipae he impac of rade liberalizaion on ax revenue and ake he necessary acions in a imely fashion if i is concluded ha he impac could be significan. These issues need o be relaed o he specific circumsances of Tanzania. General answers as provided by previous sudies in oher counries are no likely o be very helpful for Tanzania since he srucure of he economies, he sophisicaion of he ax adminisraions; or poliical obsacles vary across counries. 2. Tax Revenue Concerns of Trade Liberalizaion Tanzania expresses concern ha eliminaing ariffs for non-agriculural producs would rigger fiscal insabiliy because of is high reliance on ariffs as source of revenue. Because ariff receips are colleced a specific locaions (i.e. cusoms clearance poins), hey are generally considered o benefi from lower collecion coss han mos oher axes. This migh explain why despie heir sub-opimaliy ariffs are frequenly used as revenue device by low-income counries (Aizenman, 1985). In Tanzania, he share of impor ax revenue in oal ax receips in amouned o more han 25 percen. Africa as a whole, inernaional rade axes generae on average 28 percen of oal curren revenues over he las decade (UNECA, 2004). This compares o 0.8 for high-income Organizaion for Economic Cooperaion and Developmen (OECD), and 23 percen for low-income counries (UNECA, 2004). Recognizing ha rade reform is vial o economic developmen and povery reducion; developing counry policymakers have lowered ariffs, and reduced non-ariff barriers, he argumen being ha when ariffs are very high or prohibiive, producs may be under-invoiced, smuggled, or no impored (Amponsah, 2002). When ariffs are reduced o reasonable raes, producs are more likely o ener a counry hrough formal means and have duies paid on hem. This may increase overall revenue, even if he average ariff rae is low. In addiion, when ariffs are reduced and he cos of impors declines, demand for impors is likely o increase. A leas in he shor-erm he volume of impors and herefore ariff revenue may, depending on elasiciies of demand, increase. However, Tanzania s ax revenue as percen of declined from 17.8 percen in 1979/80 o 9.8 percen in 1998/99 (Table 1). The IMF (2003) repors ha policy sequencing aribues o he revenue decline. For example, ariffs were lowered oo quickly before compensaory ax broadening measures and a srenghened ax adminisraion were in place (IMF, 2003). Wih ariff raes declining overime, revenues from impor duies as a share of oal revenue have also ended o decline, paricularly beween he second half of 1990s and he 2000s, bu hey remain imporan sources of revenue because hey are relaively easy o collec. Table 1 indicaes ha revenue from impor duy as a percen of oal revenue declined from 15.4 percen in 1996/97 o 8.4 in 2009/10 parly as a resul of he reducion in rade ariffs (African Developmen Bank Group, 2009). Decrease in oal ax revenue and impor duy revenue is closely linked o an overall rend owards rade liberalizaion-proxied, for example, by a decline in colleced ariff raes (Figure 1). I is widely recognised ha rade liberalizaion does no necessarily reduce revenue from rade axes when i involves cuing ariffs ha are iniially se for proecive reasons so as o increase rade volume by more han enough o offse he direc revenue loss from lower raes. Trade liberalizaion also may no reduce rade axes when i is accompanied wih reducing disoring exempions and addressing corrup adminisraion. However, rade liberalizaion mus reduce rade ax revenue because free rade ulimaely means no impor duy revenues. Colleced ariff raes are now low in Tanzania relaive o levels ha are likely o be revenue maximizing. Khary and Rao (2002) for insance, esimae he revenue maximizing ariff rae for low income counries o be in he order of 38 percen, while Ebrill e al. (1999) pu i a around 24 percen. I is noable ha wih he colleced ariff rae, around 6.2 percen beween 1997/98 and 2009/10, Tanzania has colleced ariff raes below revenue maximizing average ariff raes se by Khary and Rao (2002), and Ebrill e al. (1999). The colleced ariff rae was around 10 percen in 1996, whereas impor duy revenue -o- raio accouned for 2.1 percen. In 2010, he colleced ariff rae and impor duy revenue share in, respecively accouned for 8.4 and 1.4 percen, afer several

3 Journal of World Economic Research 2014; 3(3): years of flucuaions. This implies ha rade values are likely o prove much less responsible o furher lowering of ariff raes, and herefore ax revenue from rade may fall and consequenly a fall in governmen revenue. Tanzania has been unable o recover impor duy revenue loss from rade liberalizaion hrough srenghened domesic axaion, as a resul oal ax revenue as a percen of has on average declined in parallel wih impor ax revenue. These concerns are emerging ever more clearly as poenially significan obsacles o furher liberalizaion. However, oal ax revenue-o- raio improved in recen years parly due o improvemen in revenue collecions hrough excise duy and VAT on impors. Moreover, i is expeced ha, he efficiency gains from rade liberalizaion means an increase in aggregae real income, par of which could be capured as governmen revenue hrough income Year Tax Revenue ax and hence, minimizing he loss arises from ariff rae reducion. Several feaures however, characerize he income ax regime in Tanzania, which accouned for 28.7 percen of oal ax revenue and 4.1 percen of during he 1979/ /10 period. Firs, i has been mainly limied o he formal secor and he number of individuals subjec o personal income is small. Second, he effeciveness of rae progressiviy is undercu by high personal exempions and deducions ha benefi hose wih high incomes. The problem of exempions and deducions ends o narrow he ax base and negae effecive progressiviy. Thus, he decline of oal ax revenue as a percen of in Tanzania occurs no only as a direc consequence of declining impor duy revenue-o- raio, bu also as a resul of declining income axes as a percen of. Table 1. Various Tax Revenues as Percen of and of Toal Tax Revenue, 1979/ /10 Impor Duy Toal Tax Trade Taxes Sales/VAT on Impor Toal Tax Excise on Impor Toal Tax Sales/VAT Income Tax Oher Taxes 1979/ / / / / / / / / / / / / / / / / / / / / / / / / / / / ,4 2007/ / / Noes: Sales/VAT includes sales/vat on domesic and impor consumpion; Income Taxes include PAYE and corporae ax; Oher axes include domesic excise duy, oher income axes, oher impor charges, and oher domesic charges Source: Compued using daa from Bank of Tanzania: A Review of he Role and Funcions of he Bank of Tanzania ( and Tanzania Revenue Auhoriy: Naional Tax Saisics, %of %of Toal Tax %of Toal Tax %of Toal Tax

4 28 Manamba Epaphra: The Revenue Implicaions of Trade Liberalizaion in Tanzania Figure 1. High Income and Low Income Counries: Colleced Tariff Raes, Source: IMF TPIDB; Bank of Tanzania A Review of he Role and Funcions of he Bank of Tanzania, ; and Tanzania Revenue Auhoriy: Naional Tax Saisics, Lieraure Review 3.1. Theoreical Raionale for Trade Liberalizaion and Tax Revenue Considerable evidence has accumulaed overime ha rade liberalizaion is linked o higher raes of economic growh, which is is main objecive (Escolano, 1995). Trade liberalizaion simulaes a beer reallocaion of resources in erms of comparaive advanage and can also simulae long-run growh hrough he accumulaion of capial and echnology for developing counries. The link beween economic openness and growh is quie srong; for example, while developing counries in Lain America and sub-saharan Africa ha followed impor subsiuion indusrializaion sraegies experienced relaively lower growh raes, Eas Asian counries, ha employed expor-promoion policies, consisenly ouperformed oher counries. This enhanced growh leads o expansion of ax bases, wih a concomian expansion of revenue poenial. Overime he influence of economic growh on revenue may dominae he influence of shorer-erm changes in he rade regime. A higher per-capia income reflecing a higher level of developmen is held o indicae a higher capaciy o pay axes as well as a greaer capaciy o levy and collec hem (Chelliah, 1971). Alhough, rade liberalizaion is expeced o enhance economic efficiency and growh, i is widely recognized ha he effec of cuing ariff is ambiguous. On one hand, lower ariffs imply lower ax raes and hence smaller revenues. On he oher hand, he volume of impors ends o expand when ariffs are reduced and hence he ax base will follow sui. The normal argumen is ha raionalizing he ariff sysem reduces opporuniy and incenive for evasion so compliance increases, and so revenue may increase. Higher ariffs creae an incenive for imporers o evade ariff or seek exempions (Priche and Sehi, 1994). In urn, ax evasion affecs he produciviy of he ax sysem leading o a less han proporionae increase in ariff revenue. Which of he wo effecs is larger will depend on he demand elasiciy of impors. If he elasiciy is sufficienly high and ax compliance is high, hen revenue should increase. Trade liberalizaion is ofen accompanied by a devaluaion of nominal and real exchange raes. This raises he domesic value of impors, wih a posiive effec on rade ax revenues. Bu he domesic cos of governmen spending programmes will increase. Consumpion will swich from radable o non-radable goods, rade ax revenue will herefore decline and hose from domesic indirec axaion will increase. The overall effec of devaluaion is herefore ambiguous. Long-run effecs may be driven by enhanced growh performance as a resul of rade liberalizaion. If growh increases hen increased income levels will ranslae ino a large base for domesic direc axaion (UNECA, 2004). The rade liberalizaion measures and heir implicaions on he revenue can be posiive, negaive and neural depending on he naure of resricions and he characerisics of a paricular counry (Ebrill e al., 1999). When he rade regime is very resricive because of high ariff raes, rade volumes are likely o be reduced. Reducing resricions will resul in a srong increase in rade volumes. Bu when he rade regime is fainly already liberalized, furher reducion in resricions will no cause a sufficienly large increase in rade volumes o offse he lower ariffs. This effec is illusraed by a Laffer curve, racking he relaionship beween revenues and he rade resricions (average realized ariffs), being defined as impor duy revenue divided by impor value Empirical Evidence on Trade Liberalizaion and Tax Revenue The relaionship beween rade liberalizaion and ax revenue has been debaed in economic lieraure, especially in he las wo decades. In general, he revenue implicaions of rade liberalizaion are uncerain. Agbeyegbe e al. (2003), using a panel of 22 counries in sub-saharan Africa, over he period, repor ha he revenue impac of rade liberalizaion is limied. Overall, revenue ends o be leas affeced if he iniial posiion of he rade regime is highly resricive and if liberalizaion is accompanied by reforms in cusoms and ax adminisraion wih he aim o broaden ax base. Posulaing a firs-order dynamic panel model wih a number of explanaory variables, hey find evidence ha rade liberalizaion is linked o higher ax revenue and rade ax revenue. They also find ha he exchange rae is no generally srongly linked o revenues. By conras Khary and Rao (2002) find ha he impac of rade liberalizaion on revenue is significanly negaive. For he aggregae sample of 80 counries boh developed and developing counries over he period, and applying a number of regressors on ax revenue regression and rade ax revenue regression, Khary and Rao (2002) find ha domesic indirec ax revenue-o- and he rade-o- raios are significanly and negaively correlaed indicaing ha cusoms duies on radables are paricularly replaced by

5 Journal of World Economic Research 2014; 3(3): domesic indirec axes on radables. They also find ha here is srong evidence ha srucural facors play a significan role in deermining he rade ax revenue-o- raio across counries. For example, ax revenue-o- raio varies posiively wih he level of developmen while he share of rade axes in varies inversely wih he level of developmen. Moreover, reduced proecion adversely affecs overall ax revenue and rade revenue receips. In addiion, low income counry group face declining ax revenue-o- raio in he awake of rade reform, which is associaed wih subsanial increase in he deb-o- raio. They conclude ha rade liberalizaion in developing counries resuls in revenue losses due o inabiliy o compensae for forgone ariff revenue by raising higher revenue from domesic sources. Similarly, he Economic Commission for Africa (UNECA, 2004) sudy for fiscal implicaion of rade liberalizaion on African counries reveals ha rade liberalizaion is a poenial source of fiscal insabiliy for Africa counries because of heir high dependence on inernaional rade ax. UNECA (2004) applies he Generalized Mehod of Momens of Arellano and Bond (1991) o esimae boh ax revenue and rade ax revenue regressions for all African counries excep Democraic Republic of Congo, Erirea, Liberia, Libya, and Somalia over he period. The main finding of ha sudy is ha he decrease in rade ax revenue can be mached by an increase in revenue from domesic ax, and in paricular mos counries increased reliance on VAT. The sudy concludes ha sound macroeconomic environmen is criical o prevening fiscal disress during rade liberalizaion. By achieving a sound macroeconomic sance e.g. low and predicable inflaion, high growh ec. counries can progress oward fiscal sabiliy during liberalizaion period. Moreover, effecive insiuional reform is likely o help fiscal consolidaion. In a review of African counries experience of fiscal impac of rade liberalizaion, Fukasaku (2003) finds ha he overall impac of rade liberalizaion in sub-saharan Africa is ambiguous and depends on a mulipliciy of facs such as he naure and sequencing of reforms. Examining a daabase of 22 African counries, he finds ha rade liberalizaion has conribued o declines in he raio of rade-o-oal governmen revenue of more han 20 percen in Mauriius, more han 10 percen in Côe ďivoire and Senegal, and more han 5 percen in Cameroon, Tunisia, and Mozambique. In mos counries, domesic resources mobilizaion was somewha weaker. Similarly, Ebrill e al. (1999) find ha many developing counries have implemened rade reform while avoiding significan revenue loss. In some cases revenue increases because a counry s reducion of ariff and non-ariff barriers generaes more impors. Mexico and Chile winessed such revenue increases in shor erm, when hey implemened liberalizaion programs in he 1980s (WTO, 2003). Ebrill e al. (1999) sudy concludes ha rade liberalizaion could be ailored o avoid adverse consequences. For example, revenue objecives can be me when reducions in ariffs are accompanied by improved cusoms adminisraion and reducion in exempions, non-ariff barriers, and regulaions. Neverheless, in he long-erm, rade liberalizaion ineviably reduces he oal share of revenue derived from rade axes. To ensure ha revenue is mainained, domesic ax reforms mus accompany rade liberalizaion. Bu such reforms ake ime o develop and implemen. Evenually, however, rade liberalizaion mus reduce rade ax revenue simply because free rade ulimaely means no rade axes and many counries are now likely o be in his siuaion. Some counries have resored o oher sources of governmen revenue, such as income, excise, or value added axes (Ebrill e al., 1999). Paudel (2006) examines he revenue implicaions of World Trade Organizaion (WTO) regime and alernaive measures for revenue mobilizaion in Nepal for he period. He argues ha abou 25 percen of he impor duy is los wih he fulfillmen of WTO commimen. Paudel (2006) esimaes he ime series rade regression equaion applying variables such as impor duy as percen of, impor rade as percen of, real effecive exchange rae, un-weighed overall average ariff rae, and dummy variable as a proxy for rade liberalizaion. He observes ha impor-o- raio increased afer he rade liberalizaion whereas he average (colleced) ariff rae declined in he same period indicaing ha he increase in impor-o- raio may no be sufficien enough o offse he loss of revenue due mainly o declining rade axes in he shor-run. He concludes ha he higher impor-o- raio appears o compensae for he reducion in rade revenue arising from he rade liberalizaion in a long-run. Malanyane and Harmse (2001) apply he ordinary leas square on annual ime series daa from o esimae he produciviy of rade ax revenue and o esablish he effec of rade liberalizaion on ariff revenue in Souh Africa. The rade ax equaion is esimaed in linear logarihmic form where cusoms revenue is used as proxy for inernaional rade ax revenue and dummy variable for liberalizaion. Their main finding is ha cusoms revenue is highly producive. In addiion, rade liberalizaion has a significan influence on cusoms revenue and ha an increase in impor ax rae leads o a reducion in rade ax revenue. On he oher hand, Malanyane and Harmse (2001) find ha supporive macroeconomic policies are a prerequisie for successful rade liberalizaion. In differen sudies, Blejer and Cheasy (1990) and Tanzi (1989) conclude ha he ne impac of rade reform on revenue is an empirical maer. Rodrik (2001) argues ha he benefis of rade reform have been oversold and ha economic policy should focus on growh and no rade. Rodrik (2001) criizes he sandard rade heory and he argumen by Edward (1993) ha, sugges ha here is a srong empirical correlaion beween an open rade policy and economic growh since rade liberalizaion has a posiive effec on expors growh. However, Rodrik (1995) shows ha here is a robus negaive relaionship beween per capia income and he share of rade axes in oal ax

6 30 Manamba Epaphra: The Revenue Implicaions of Trade Liberalizaion in Tanzania revenue: an increase in per capia income of US$1,000 is associaed wih a reducion by 3.7 percen poins in he share of rade ax revenue. The usual explanaion is ha in he counries wih poor adminisraive capabiliies rade axes may be he easies way for governmens o raise revenue (Kuboa, 2000; and McLaren, 1998) Knowledge Gaps I is noed ha he empirical sudies have no resolved he debae over he impac of rade liberalizaion on ax revenue. The empirical findings have been mixed because of heir sensiiviy o se of counries and he period of analysis. While rade liberalizaion can help accelerae inegraion in he world economy and srenghen an effecive growh sraegy, i canno ensure is success. Oher elemens ha address binding consrains o growh such as macroeconomic managemen, rade relaed infrasrucure and insiuions are needed. Unlike Souh-Eas Asian counries, which are more compeiive and inegraed ino he global economy, moving from expor of raw maerials o a more dynamic inegraion focused on high echnology produc, Tanzania, and oher sub-saharan Africa counries have grea difficuly in dealing wih heir increasing impac of rade liberalizaion on ax revenues. For example, he economy of Tanzania is mainly characerized by low-income and based on subsisence agriculure, which is difficul o ax. Moreover, budgeary income is closely ied o inernaional rade revenue. From his viewpoin, rade liberalizaion could creae a considerable gap in revenue and migh lead o a worsening of major macroeconomic imbalances. Mos sudies on he impac of rade liberalizaion on ax revenue, however, are based on a large sample of developing counries, few focuses on specific emerging economy. To fill his gap his sudy examines he revenue implicaions of rade liberalizaion in Tanzania. This conribuion is par of debae and, on he basis of less developed counry seeks o add some ideas o he relaionship beween rade liberalizaion and ax revenue. 4. Mehodology and Daa Descripion 4.1. The Theoreical Model The share of inernaional rade in he economy is a measure of openness, which may also be a proxy for rade liberalizaion. Cerain facors of inernaional rade make i more suiable o axaion han domesic aciviies. In Tanzania, he inernaional rade secor is ypically he mos moneized secor of he economy and has gained increasing imporance in recen years. A posiive correlaion would be expeced beween he impor shares in he economy (IMPORT) and he impor duy revenue-o- raio (IDR) as adminisraive coss of he ax sysem o his secor should be lower han oher secors. The relaionship beween impors, rade liberalizaion, and impor duy, however is no direc. A counry ha axes impored goods wih very high ariffs discourage impors, ariffs may be above he level ha would maximize revenue from impor duies. In his case some reducion of impor duies, by encouraging more impors, would increase and no reduce impor duy axes. For some range of he rae of impor duy, a lowering of he rae will increase impors by so much as o more han compensae for he reducion in he ax rae. In his range, impors are elasic wih respec o he impor duy. There is a poin a which he level of impor duies is such as o maximize revenue from rade axes. Below his poin, however, he reducion in impor duies would reduce he revenue from impor axes. Apar from he share of impor in, here are oher several deerminans of impor duy revenue in Tanzania and oher low income counries; hough any generalizaion is difficul given he differences in he poliical and economic srucure across he counries. The economy of Tanzania is characerized by low per capia (pc). Per capia is considered o capure he level of developmen. Higher income counries end o have a more moneized economy and beer ax adminisraion bu less rely on rade axes, he coefficien of he per capia is, herefore, expeced o be negaive. This is considered o reflec feaures of Tanzania economy wih difficulies in adminisering an effecive domesic ax sysem and relaively small share of formal secor in he economy. Farhadian-Lorie and Kaz (1989) noe ha ariff axes are hisorically been a major source of governmen revenue during he early sages of economic developmen because hey are easier o collec han domesic income or consumpion axes, owing o he rudimenary saus of ax adminisraion, as well as he limied availabiliy of ax base. During he laer sages of developmen, however, collecion coss are expeced o fall, dependence on ariff axes o decline, and dependence on indirec axes o rise. The effec of macroeconomic policy on impor duy revenue is also capured in he real exchange rae (RER). Real exchange rae index is defined such ha an increase indicaes a real depreciaion and is coefficien could in general be expeced o be posiive, since a real depreciaion implies higher nominal value of impors and hence increased impor duy revenue. Neverheless an overvaluaion of he real exchange rae-ypically brough abou by expansionary financial policies-would be expeced o adversely affec overall economic aciviies, and hus o lower ax revenue. The general advice given by inernaional insiuions such as IMF and he World Bank o developing counries over pas few decades has been o replace rade axes wih domesic consumpion axes (CTAX), paricularly, VAT, and income axes (DTAX), paricularly corporae income ax. However, some recen lieraure challenges his convenional view, arguing ha he relaively large informal secor in developing counries may jusify a differen ax policy design. For example he income ax base in developing counries such as Tanzania is mosly comprised of wages of public secor employees because mos oher axpayers are self-employed or small businesses who evade pay of all, or mos of income ax. In addiion, axaion of

7 Journal of World Economic Research 2014; 3(3): personal capial income is easily evaded (IMF, 2005). Similarly, he relevan daa shows ha he performance of VAT was quie saisfacory in he iniial years bu subsequenly VAT collecion has remained sagnan a a cerain level. As a resul, VAT is unable o mee he objecives for which i was inroduced. The reasons behind his performance are many, such as a relaively small number of VAT ax-payers, exempion of agriculural producion and oher producs, and a weak monioring sysem. Therefore, here is no clear evidence wheher Tanzania is able o replace revenue loss from inernaional rade hrough domesic axes Concepual Framework Since rade liberalizaion enails reducing ariff raes, hen, he revenue implicaions would depend on he elasiciies of impor demand and supply wih respec o he ax raes. A sufficienly elasic response would imply ha revenue would increase as a resul a decrease in ariff raes. The resriciveness of he rade regime is proxied by he colleced ariff rae. The relaionship beween he colleced rae of rade axaion and impor duy revenue is assumed o be nonlinear as prohibiively high raes of he rade axaion may lead o declining impor duy revenue. A quadraic form is herefore applied o esimae he effecs of rade liberalizaion on impor duy revenue. The overall marginal effec of ariff rae on impor duy revenue is hen ( γ 1 + 2γ 2 CTR) and hence depends on CTR (see Ebrill e al., 999). The revenue-maximizing rae of rade axaion is obained by solving for CTR in he following equaion: ( γ 1 + 2γ 2 CTR) = 0, ha is CTR = γ 1 2γ 2 (see Khay and Rao, 2002). Many sudies on he impac of rade liberalizaion on revenue emphasize on policy decisions ha end o replace impor duy revenue wih domesically based axes such as domesic consumpion axes, boh excise on goods and VAT o offse any revenue loss from rade liberalizaion. Similarly, efficiency gains from rade liberalizaion means an increase in aggregae real incomes par of which could be capured as governmen revenue hrough he income ax. Impor duy revenue model herefore, predics ha Tanzania can compensae revenue loss from rade liberalizaion by raising revenue on domesic consumpion of goods and services and income axes. Thus, he model is specified as IDR = γ 2 ( CTR) + γ ( pc ) 0 + γ 1CTR + γ 2 3 log + γ IMPORT + γ CTAX γ DTAX 7 + ε 4.3. Definiions and Expeced Signs on Variable Parameers + γ RER Table 2 presens he definiions and expeced signs of variables used in he ax raio equaion. For esimaion, he colleced ariff rae, measured by he raio of impor duies o he value of impors, is considered as a proxy of degree of 4 liberalizaion. Wih his measure, a decline in he index is considered o indicae greaer rade liberalizaion. 1 The measure is based on how much ariff revenue is acually colleced; i may differ from official ariffs because of imperfec collecion and measuremen errors. Daa are available o consruc he index for fairly long ime periods for Tanzania. The appropriaeness of his measure is also discussed in Ebrill e al. (1999). The per capia variable is enered in logarihm, while he remaining variables are enered linearly because he ax revenue funcion is assumed o be nonlinear in he scale of he economy (as measured by per capia ) (see Kary and Rao, 2002; Ebrill e al.,1999). The sudy uses he raio of. includes income earned locally ha accumulaes o non-residens and excludes income received from abroad by residens, since local income accruing o non-residens ypically is axed while remiances from abroad ypically are no. Variable Impor Duy Revenue Colleced Tariff Rae Colleced Tariff Rae Square Per Capia Real Exchange Rae Value of Impor Domesic Consumpion Taxes Domesic Income Taxes Table 2. Descripions of Variables Abbrev. Definiion IDR CTR ( CTR ) 2 pc RER IMPORT CTAX DTAX Impor Duy Revenue as Percen of Impor Duy Revenue as Percen of Impor Value Colleced Tariff Rae Square (Included o Capure he Laffer Curve Effec) Log of Real Per Capia Calculaed from Nominal Exchange Rae and CPI Impor Value (Percen of ) Indirec (Consumpion) Taxes (Percen of ) Direc Taxes (Percen of ) Source: Consruced from Lieraure Review Expeced Sign γ 1 > 0 γ 2 < 0 γ 3 < 0 γ 4 > 0 γ > 0 5 γ 6 < 0 or γ 0 6 > γ 7 < 0 or γ 7 > Daa Sources and Time Series Characerisics of Daa Daa Sources Various caegories of ax revenues and daa on are obained from Bank of Tanzania: A Review of he Role and Funcions of he Bank of Tanzania ( ), and Tanzania Revenue Auhoriy: Naional Tax Saisics, Daa on series on macroeconomic variables such as real exchange rae, and rade balance (impors and expors) are 1 The colleced impor ariff rae (ariff revenues divided by impor value) is frequenly used as a proxy for he level of ariff proecion, bu changes in his rae do no capure oher imporan elemens of rade liberalizaion such as lowering non-ariff barriers (NTBs). For furher discussion, see Ebrill, Sosky and Gropp, 1999).

8 32 Manamba Epaphra: The Revenue Implicaions of Trade Liberalizaion in Tanzania obained from Bank of Tanzania, he Inernaional Moneary Fund (IMF): Inernaional Financial Saisics (IFS), World Trade Organizaion (WTO), Unied Naions Conference on Trade and Developmen (UNCTAD): Handbook of Saisics, 2006/07 and 2009/10, and he Economic Research Services (ERS)/ USDA: Inernaional Macroeconomic Daa se Time Series Characerisics of he Daa I is widely acceped ha regression of non saionary ime series may indicae significan relaionship even if he acual relaionship is spurious. For empirical analysis herefore, i is imporan o ake ino accoun he underlying properies of he processes ha generae he ime series variables, more specifically, wheher he variables under consideraion are saionary and have long-run relaionships. In his connecion, ess o deec non saionariy and deermine he order of inegraion of he variables in he model have o be conduced Saionariy and Uni Roo Tes Economic variables ha are non-saionary are characerized by a change in variance ha becomes infinie overime. In addiion, he series wanders wihou revering o he fixed mean. These characerisics resul in spurious relaions beween variables in economeric analysis. The use of ime series variables in esimaing economeric models requires ha a sochasic process generaing he daa series be saionary. The sochasic process is saionary if all he variables generaed have consan mean and consan variance over he ime. There are several mehods ha have been devised o es for uni roo in ime series variables. These mehods include Q-saisics, Ljung-Box saisics, weighed symmeric es, Phillips-Perron (PP), Dickey-Fuller (DF) and Augmened Dickey Fuller (ADF) ess. 2 The Dickey Fuller (DF), Augmened Dickey Fuller (ADF) and he Phillips-Perron (PP) uni roo ess are he commonly used mehods o es he saionariy of he ime series variables. In his sudy ADF and PP ess are used. The idea behind he ADF and PP uni roo ess is ha, he ADF and PP ess use differen mehods o conrol for higher-order serial correlaion in he series and have been widely used successfully in oher sudies. The ADF es makes a parameric correlaion for higher-order correlaion by assuming ha he series, say IDR follows auoregressive process and adjusing he es mehodology. Moreover, he ADF approach conrols for higher-order correlaion by adding lagged difference erms of he dependen variable (IDR) o he righ-hand side of he 2 The ADF es is virually he same as he DF es, excep he lag lengh has o be long in order o reflec he addiional dynamics ha canno be capured by he DF es and o ensure ha he error erm is whie noise. The ADF is i presened as Y = α + λ + δy 1 + φi y i + ε, where he lag lengh i i= 1 is se o ensure ha any auocorrelaion in Y is absorbed. This augmened specificaion is hen used o es he hypohesis ha: H 0 : δ = 0, H1 : δ 0. regression. Phillips-Perron es is a nonparameric mehod of conrolling for higher-order serial correlaion in a series. The PP makes a correlaion o he -saisics of IDR coefficien from he AR(1) regression o accoun for he serial correlaion in error erm (ε ). The correlaion is nonparameric since we use an esimae of he specrum of ε a frequency zero ha is robus o heeroskedasiciy and auocorrelaion of unknown form. There is evidence ha he Phillips-Perron es has more power han he augmened Dickey-Fuller es (Davidson and MacKinnon, 1993). If he variables are found o be I(1) saionary, he nex sep is o es wheher hey are coinegraed using he Engle Granger wo sep procedure, as discussed in he nex sub secion Coinegraion Analysis and Error Correcion Modelling Economic heory is mosly ineresed in equilibrium condiions and has lile o say abou he naure of economic configuraions in disequilibrium. However, he daa does no confirm ha equilibrium relaionships beween economic variables hold a all imes. To overcome his difficuly, economiss make a disincion beween he shor-run and he long-run. The appeal of coinegraion is ha i provides a formal framework for esing long-run models from acual ime-series daa. The coinegraion analysis allows non saionary daa o be used so ha spurious regression resuls are avoided. I also gives he chance o es he validiy of an economic heory. If a posulaed economic relaionship exiss, hen he variables under consideraion should be coinegraed. Tesing for coinegraion is, hus, a es for he exisence of he equilibrium relaionship posulaed. Granger and Engle (1987) wo-sep esimaion procedure and he Johansen (1988) procedure are wo procedures ha are popularly used o idenify and esimae he coinegraing vecors and he shor run adjusmen parameers. The former procedure involves normalizing he coineraing vecor on one of he variables, which makes he assumpion ha he corresponding elemen of he coinegraing vecor is non-zero. 3 The Johansen procedure is a mulivariae approach, he esimaion of which would consume a lo of degree of freedom. In his sudy he Engle-Granger wo-sep procedure is performed o es for coinegraion beween variables in a single equaion. The Engle-Granger ype of saic long-run regression has become a widely applied mehod since i was inroduced by Engle and Granger (1987). 3 Formally, Granger (1983) and Engle and Granger (1987) define ha an economic ime series is said o be inegraed of orderd, or I ( d ), if i requires o be differenced d imes o achieve saionariy. If we have an [ n 1] vecor of variables X, which I ( d ), he series in X are said o be coinegraed if here exiss a combinaion ( d b) I X ε such ha ε is β = and β is known as he coinegraing vecor. The saisical concep of coinegraion has a naural counerpar in economic heory. The basic idea is ha he exisence of coinegraion beween a se of economic variables implies ha he variables move ogeher in he long run, or in oher words are in long run equilibrium. In esing for coinegraion of economic ime series, one is also esing for presence (or absence) of a long run relaionship beween hem.

9 Journal of World Economic Research 2014; 3(3): Some sugges ha he esimaes of he Engle-Granger ype saic long-run ordinary-leas-squares (OLS) regression parameers are boh consisen and highly efficien (Sock, 1987). 5. Empirical Resuls and Discussion 5.1. Uni Roo Tes Tes for saionariy of variables before heir linear combinaion was conduced using boh Augmened Dickey-Fuller (ADF) and Phillips-Perron ess. Selecion of he numbers of lags was made using -saisics. The variables were esed for uni roo using 1 lag. This was done Variable Tes Saisic ADF-es ADF -value Table 3. Uni Roo Tes Resuls PP-es by including only a consan and also by including a consan and rend. Consan and rend opion was seleced because i is a general case. The whole process was done in levels, firs difference, and second difference. The resuls for uni roo es are presened in Table 3. The resuls in Table 3 indicae ha impor duy revenue (IDR), colleced ariff rae (CTR), domesic (consumpion) indirec ax revenue (CTAX), income ax revenue (DTAX) and impor value (IMPORT) are inegraed of order one I(1), whereas per capia (pc ) and real exchange rae (RER) are inegraed of order wo I(2), his suggess ha furher esimaion would be carried while in firs difference and second difference respecively in order o avoid spurious correlaion. Tes Criical Values PP Tes Criical Values I(d) 1% Level 5% Level -value 1% Level 5% Level IDR -4.31** ** I(1) CTR -3.88* ** I(1) Pc -4.37** * I(2) RER -5.63** ** I(2) IMPORT -3.14* ** I(1) CTAX -4.21** ** I(1) DTAX -4.91** ** I(1) Noes :(1) I(d)= Order of Inegraion (2) *(**) denoe rejecion of he null hypohesis a 5% & 1% criical values respecively (3) Included in he es equaion: (i) rend and inercep (ii) lagged difference (1). I(d) 5.2. Coinegraion Analysis According o Thomas (1993), if an equilibrium relaionship exiss, hen he disequilibrium error should flucuae abou zero (Figure 2). This implies ha saionariy of he error erm mus be obained a is iniial level. A more useful way of coinegraion is by using a single equaion error correcion model (SEECM), in which a saic regression model is esimaed Such a mehod allows for he possibiliy of including more han one independen variables in he saic regression (Charemza and Deadman 1992). Therefore, he Engle-Granger (wo-sep) single equaion procedure was used for coinegraion analysis. The use of a single equaion procedure was deemed appropriae, a leas wih respec o preserving he degree of freedom. Table 4 provides a summary of he resuls of he ess for coinegraion on he residuals of he long run impor duy revenue model. Overall, he resuls show ha he errors in he coinegraion regression are saionary. A comparison of he compued Augmened Dickey Fuller and Phillips-Perron (PP) es resuls ends o suppor coinegraion beween he ariff revenue and explanaory variables. Boh ess rejec he null hypohesis of no coinegraion. Table 4. Saic Model: Tess for Coinegraion beween Impor Duy Revenue (IDR) and Explanaory Variables Augmened Dickey Fuller Tes on Residuals Phillips-Perron (PP) Tes on Residuals ADF-es -value Tes Criical Value 1% Level 5% Level -3.05* I(0) PP-es -value Tes Criical Value 1% Level 5% Level -5.48** I(0) Order of Inegraion Order of Inegraion Noes (1) *(**) denoe rejecion of he null hypohesis a 5% and 1% criical values respecively (2) Included in he es equaion: (i) rend and inercep (ii) lagged difference (1)

10 34 Manamba Epaphra: The Revenue Implicaions of Trade Liberalizaion in Tanzania 5.3. Error Correcion Model 4 In order o capure he shor run relaionship beween he impor duy revenue and a series of explanaory variables, he error correcion model was esimaed. 5 The error correcion specificaion resrics he long run behaviour of he endogenous variables o converge o heir coinegraing relaionships while allowing a wide range of shor run ECT were obained dynamics. The error correcion erms ( ) from he solved saic long run equaion and lagged once, i.e. ECT 1. The impor duy revenue (IDR) equaion was herefore specified o include he error correcion model and he esimaion resuls are presened in Table 5. The model seems o be correc as he coefficien of he error correcion erm is negaive and saisically significan. The ECT 1 reflecs he aemp o correc deviaions from he long run equilibrium pah and is coefficien can be inerpreed as he speed of adjusmen Esimaion Resuls and Implicaions Esimaion resuls presened in Table 5 show ha adjused R-squared, which measures he goodness of fi of he variables, indicaes ha abou 93 percen of he variaions in he impor duy revenue is explained by he explanaory variables. Similarly, he F-saisic of 48.4 rejecs he null hypohesis ha all explanaory variables have coefficiens no differen from zero. Furhermore, Durbin-Wason saisic (DW) indicaes ha here is no auocorrelaion, as he DW saisic of 1.9 lies close o 2. Variable Table 5. Empirical Resuls Coefficien Sd. Error -Saisic Prob. C CTR *** ( CTR ) *** Log pc RER IMPORT *** CTAX ** DTAX *** ECT *** R-sq S.E. of reg Adj. R-sq F-saisic RSS Prob(F-saisic) DW Saisic Noes: *significan a 10 percen; **Significan a 5 percen; ***Significan a 1percen. 4 Error correcion model is resriced model ha has coinegraion resricions buil ino he specificaion, so ha i is designed for use wih non-saionary series ha are known o be coinegraed. The coinegraing erm is known as he error correcion erm (ECT -1) since he deviaion from he long-run equilibrium is correced gradually hrough a series of parial shor-run adjusmens. 5 According o Angle and Granger (1987), when coinegraion is esablished he nex sep is o represen a shor-run disequilibrium relaionship of he variables using an ECT. The resuls also show ha he coefficien of colleced ariff rae (CTR) is posiive and significan a 1 percen indicaing ha one percen reducion in colleced ariff rae may lead o abou 0.3 percen decline in impor duy revenue-o- raio. This confirms he hypohesis ha a reducion in he ariff raes resuls in a significan loss of impor duy revenue. The robus suppor is provided for he presence of a significan relaionship beween he impor duy revenue-o- raio and income ax-o- raio. The coefficien of income ax-o- raio (DTAX) is negaive and significan a 1 percen implying ha as income ax-o- raio (DTAX) increases, impor duy revenue-o- raio declines. This also indicaes ha reducion in impor duy revenue is parially replaced by income ax revenue. Despie he fac ha he income ax base is mosly comprised of wages of employees in he public secor, because all oher axpayers are self-employed or small businesses who evade paying all, or mos, of he income ax, he raio of income ax revenue as percen of has increased from 3.2 percen in 1993/94 percen o 5.3 percen in 2009/10. This may be due o he fac ha axing individual incomes and large corporaions does no involve significan adminisraive and compliance coss, because hey are per force required o comply wih sauory accouning requiremens. The coefficien of consumpion axes-o- raio is posiive and significan a 5 percen. The posiive relaionship beween indirec axes and impor duy revenue-o- raios may be due o he fac ha informal secor and exempions end o narrow he ax base; as a resul consumpion indirec axes (VAT and excises)-o- raio does no grow significanly o replace impor duy revenue loss. Moreover, his resul is conrary o he general advice of inernaional insiuions such as he IMF and he World Bank given o developing counries over he pas few decades, ha hese counries need o replace impor duy revenue wih domesic consumpion axes, paricularly value-added ax (VAT) and o mainain relaively high corporae income ax raes. The main reason for his conroversy is ha, while his assumpion is auomaically saisfied when an economy consiss of only he formal secor, i is no a plausible assumpion in he presence of a large informal segmen in he economy ha, by definiion, escapes VAT coverage. Furhermore, he macroeconomic environmen in Tanzania does no suppor revenue recovery as coefficiens of percapia (pc) and real exchange rae (RER) are insignifican. The coefficien of impor value as a percen of (IMPORT) is posiive and significan a 1 percen indicaing ha one percen increase in impor-o- raio may increase impor duy revenue-o- raio by 0.09 percen. This is consisen wih he heory ha higher impor-o- raio is seen o lead o rising shares of impor duy revenue in. However, due o he complex naure of Tanzania impor rade like exempions, informal rade, highly inelasic essenial impor, i is difficul o make conclusive remark on he magniude of he impac of impor on impor ax revenue boh as he percen of.