Proximity, Regional Integration and Weak Trade among African Countries Perspective from SADC

Transcription

1 Proxmty, Regonal Integraton and Weak Trade among Afrcan Countres Perspectve from SADC Chukwuma Agu Afrcan Insttute for Appled Economcs, Enugu Anthona Achke Department of Agrcultural Economcs, Unversty of Ngera, Nsukka TRAPCA Trade Polcy Forum 2010 Abstract A lot of efforts are beng put nto untng Afrca through regonal ntegraton. Smply put, regonal ntegraton s the vogue. But t seems not much consderaton s gven to economc underpnnngs and potental gans from the efforts. Beng contguous should help Afrcan countres, but whether such stance s supported by emprcal analyss s questonable. In ths work therefore, we try to show what drves trade among Afrcan countres usng South Afrca, the SADC and the rest of Afrca as nstance. We then try to explore mplcatons of ths for regon ntegraton n the regon.

2 I. Introducton Increasngly, trade among contguous countres s organzed along regonal trade agreements (RTAs). However, Afrca s experence n regonal ntegraton s severally consdered to have fallen short of the optmal. Whle RTAs have multpled n the contnent, ntra-group trade among partcpatng countres has been ether epleptc or stagnant snce Intally weak commtments by ntegratng countres contrbuted to these poor outcomes; but there s the probablty that the poor outcomes have n turn mpacted on commtment, leadng to a vcous cycle n regonal trade protocols. The ultmate losses are on output growth and welfare. The story s slghtly dfferent n some other parts of the developng world. Whle Afrca s share of global trade has dpped between 1980 and 2008, Asa s trade share (drven manly by Chna, Inda and South East Asa) has doubled over the same perod. Lkewse, whle ntra-group trade among Asan countres has ncreased substantally, that of Afrca has remaned epleptc or stagnant over the perod. Interestngly, whle regonal ntegraton has a long hstory n Afrca, t s stll tryng to take root n developng Asa. South Afrca s unquely placed among developng Afrca to trade the most wth ts neghbours beng the largest economy n sub Saharan Afrca (SSA). It has a far more dversfed economy than the rest of SSA and boasts hgher output and better nfrastructure than all ts neghbours. Gven the poston of theory on proxmty and regonal ntegraton effects on trade, the country should have replaced most European, North Amercan and Asan sources of mports for other Afrcan countres. Unlke most other neghbourng Afrcan countres, both n SADC and other parts of Afrca, ts exports are very dversfed and conssts of a number of goods needed by poor Afrcan countres. However, as shown n Fgure 1, after nearly three decades of regonal ntegraton n SADC, South Afrca s major export destnatons are stll countres far out of Afrca. In fact, South Afrca s trade structure s such that there s hgh ntra-ndustry trade, wth the country mportng and exportng same types of goods from other developed economes and mportng lttle from Afrca. Meanwhle, t has spent nearly two decades n ntensve regonal trade formaton negotatons and nsttutonalzaton wth other 13 countres n Southern Afrca and has the oldest Customs Unon n the World wth another 4 countres. Wth ts unque geographcal poston at the end of the southern hemsphere, there s need to queston conventonal wsdom on the mpacts of proxmty and regonal ntegraton efforts on South Afrca s trade structure. Ths paper ntends to nvestgate ths paradox. It holds the assumpton that the negatve effects of poor trade facltaton measures and weak output base on trade volumes between South Afrca and ts neghbours may more than offset the meager gans accruable from regonal ntegraton and proxmty. It nvestgates ths poston usng exports from South Afrca to a sample of 23 Afrcan countres 10 of whch are n the SADC regonal trade groupng. The analytcal framework s the standard gravty model, augmented wth selected measures of trade facltaton constructed by the authors from World Economc Forum Global Compettveness Report and the World Bank s Dong Busness Index. Defntely, there seems to be reason to queston the poston of gravty theory on dstance for countres at the level of development as those n our sample, especally gven South Afrca s unque hstory and trade structure. But more crtcal for the purpose of ths study s the queston of the overall mpact of the multpled regonal ntegraton efforts ongong n many parts of Afrca. Have they mpacted trade? Can they really 2

3 mpact trade? If the major reason for establshng regonal trade groupngs s to mprove trade, why have Afrcan regonal ntegraton arrangements (RIAs) not done so? What are the mnmum condtons that may be needed to mprove overall mpacts of RIAs n Afrca and how can those be acheved? Is there a role for trade facltaton or do we just take standard specfcaton of the gravty model as beng enough to explan trade among contguous countres? The paper does not ntend to answer all the questons; t smply sets out to rase ssues for further debate on ths crtcal subject. Fgure 1: Trends n SA Export II. Facts Regonal Integraton and Trade Facltaton n Afrca: Some Stylzed Intra-group trade n Afrca s doubtless affected by hstorcal tes. Two regons prncpally led ntra-group trade growth n Afrca between 1980 and 2005: Eastern Afrca and French West Afrca (the UEMOA). Wthn each of these groups, consttuent countres share a lot of cultural affnty. French West Afrcan countres, for example, share long hstorcal tes among themselves and between them and France. There are lttle language barrers to trade and other cultural tes are much stronger than they are n other parts of the contnent. Lkewse, Swahl remans a general East Afrcan language. Even though Englsh language s the lngua franca n these places, there s lttle doubt the communal dentty through a common language has helped n promotng trade wthn the regon. Economc tes among these countres have also been longer than n some other parts of the contnent. For example, the East Afrcan Communty (EAC) and the UEMOA have had longer hstores of economc relatonshp arsng from socal nteractons than the more recent and artfcal Economc Communty of West Afrcan States (ECOWAS) and Economc and Monetary Communty of Central Afrca (CEMAC). But there are also multple oblgatons, membershps and allegances. Presently, there exst multple unservced (and ndeed unservceable) vsons, nterests, expectatons and 3

4 responsbltes vested on regonal ntegraton projects wthn the contnent. Commtments and tmelnes are not always respected by member countres of regonal blocs n Afrca. Consequently, t s not clear that regonal ntegraton n Afrca has helped to substantally mprove trade among Afrcan countres. It s acknowledged n the lterature that membershp n many RTAs can complcate admnstratve procedures rasng trade facltaton costs. Multplcty of rules from dfferent RTAs strans nsttutons charged wth admnsterng trade agreements on such ssues as customs procedures and techncal standards. For Afrcan countres wth weak nsttutons and capabltes for such complex admnstratve requrements as mposed by multple commtments, the mpacts on overall trade and development can be dauntng. It s not surprsng then that conclusons on the mpact of regonal ntegraton on ntra-group trade n Afrca have been mxed. For example, whle Cernat (2001) workng on SADC, COMESA and ECOWAS - found that Afrcan RTAs have postve mpact on ntra-rta trade, World Bank (2000) and Yeats (1998) concluded that Afrcan regonal blocs are potentally more trade dvertng than others and have doubtful non-economc benefts. Ths s partly because as shown n Table 1, regonal ntegraton arrangements n Afrca (as n a number of other regons of the world) often have mandates beyond merchandse trade. Table 1: Other Mandates of Selected Regonal Integraton Arrangements RIA/Item NAFTA EU- South Afrca MERC OSUR CARI COM AFTA SADC COM ESA Standards Yes No Yes Yes Yes Yes Yes Transport No No Yes Yes Yes Yes Yes Customs Yes No Yes Yes Yes Yes Yes cooperaton Servces Yes No Yes Yes Yes No Yes Intellectual Yes Yes No No No Yes No property Investment Yes No Yes Yes Yes Yes Yes Dspute Yes Yes Yes Yes No Yes Yes settlement Labour Yes No Yes Yes No Yes Yes Competton Yes Yes Yes Yes No Yes Yes Source: World Bank 2005 Regonal ntegraton projects the world over are heavly mpacted by the state of nfrastructure among ntegratng countres. And t s known that nfrastructure s weak n many parts of Afrca. Ths ncludes both human and physcal nfrastructure. Transportaton network across countres n Afrca s probably one of the leanest globally and communcaton nfrastructure s skeletal and costly. Socal nfrastructure and nsttutons for the effectve actualzaton of the goals of regonal trade ntatves are also weak. Cultural barrers, ncludng language also stand n the way of trade agreements. 4

5 Fact s, most regonal ntegraton arrangements (RIAs) n Afrca reflect not just a desre to ncrease trade, but more so, the need for mproved barganng power and reduced poltcal tensons among contguous countres. SADC for example orgnated n the 1980s as a coalton opposed to aparthed n South Afrca and only wth progress n the frst objectve was the creaton of free trade area gven precedence. Ostensbly n reacton to the relatvely hgh number of conflcts n the regon, many other Afrcan RIAs are as actve n conflct resoluton as n trade lberalzaton. Consequently, whle ncreases n merchandse trade may be meager, partcpatng countres of these RIAs evaluate the success of ther efforts n terms of the extent to whch conflct has been reduced (World Bank, 2005). Despte the above though, t remans a fact that the prncpal objectve of RIAs s economc ncreasng trade and potentally mprovng welfare among partcpatng countres reman strategc objectves. Even wth sgnfcant successes n non-economc objectves, the extent to whch an RIA equally mproves economc well-beng among partcpatng countres s consdered crtcal. Besdes mprovng trade among partcpatng countres, effectve trade facltaton, whether led by regonal or natonal reforms, s expected to mprove overall trade wth the rest of the world and potentally ncrease the share n global trade of a regonal economc groupng. For nearly all Afrcan regonal economc communtes (RECs) though, there has been a race to the bottom n global trade share snce early 1980 as shown n Fgure 2. Indeed, trade share has become so mnscule that there s hardly any Afrcan regonal groupng wth global trade share up to 1 percent as opposed to SADC s and COMESA s trade share of nearly 3 percent and 2.4 percent respectvely n the md 1960s. Fgure 2: World Trade Share among Regonal Integraton Arrangements 3.5 World Trade Share Among Afrcan RECs Percent CEMAC % World Trade COMESA % World Trade ECCAS % World Trade ECOWAS % World Trade SADC % World Trade Source data: UNCTAD It s not a stretch of magnaton to propose that poor trade facltaton measures alongsde a weakenng output base have largely contrbuted to the dwndlng trade fortunes n Afrca. For example, constructng a domestc nfrastructure measure from a pool of three ndcators overall 5

6 nfrastructure level, qualty of electrcty nfrastructure and qualty of the educatonal system from the World Economc Forum s (WEF) Global Compettveness Report (GCR), Fgure 2 shows the relatve performance of Afrcan (and SADC) countres relatve to other selected regons of the world. As can be quckly observed from Fgure 3, the gap between Afrca and other regons n terms of domestc nfrastructure keeps wdenng by the year. Between 2005 and 2007, Afrca s average fell closer to the score of 3 whle other regons mproved to almost a score of 6 (the lowest score s 1 whle the hghest score s 7). The same could be sad of several other ndcators and measures n most benchmarkng results nvolvng the regon. Fgure 3: Performance of Selected Regons n Domestc Infrastructure Measure n the GCR Source: WEF GCR, Varous Issues III. The Lterature Emprcal studes on trade facltaton, even though growng, are very recent and generally nadequate (Njnkeu et al 2008). Maur (2008), underscorng the relatonshp between regonal ntegraton and trade facltaton, takes a backward assessment of the mplcatons of regonal ntegraton on trade facltaton. Acknowledgng that most regonal trade agreements presently ncorporate trade facltaton dmensons, he tred to evaluate the mplcatons of trade facltaton reforms at the three levels they are usually undertaken natonal, regonal and multlateral. The am of the paper was to nvestgate how regonal ntatves can contrbute to trade facltaton reform, wth the ntenton of determnng where the optmal level of polcy nterventon les among the three. The work assumes that n some nstances regonal solutons, as opposed to natonal or multlateral ones, can be qute more helpful n solvng trade facltaton challenges among contguous countres. To do so, he apples a market falure test that determnes whether a regonal publc nterventon wll delver a welfare maxmzng reform, partcularly when natonal remedes to market falures are nadequate. Secondly, followng Sauve and Zampett (2000), he apples a subsdarty test that tres to match the poltcal jursdcton of the 6

7 respondng government to the economc doman of benefts that accrue from the publc good whch market has faled. For the frst test, he nssts that regonal solutons should be sought when the falng markets correspond to some well-defned set of natons. For the second, he observes that transacton costs to provson of regonal trade facltaton are optmzed when the most approprate partcpants partake n such provson (Arce and Sandler, 2002). He concludes that the need to address coordnaton and capacty falures, whch can occur when dsparate natonal governments ndependently tackle regonal trade facltaton challenges, there s need for care n the choce of operatonal platform for delverng on regonal trade facltaton reforms. In ths respect, regonal collaboraton wll not exactly delver the same results as regonal ntegraton. The work attempted to explan why some outcomes of ntegraton-led trade facltaton reforms have not delvered on ths theoretcal deal. Furtherng the debate on regonal facltaton measures vs-à-vs unlateral lberalzaton and/or facltaton programmes, Ptgala (2005) studed the seven South Asan countres n SAPTA Bangladesh, Bhutan, Inda, the Maldves, Nepal, Pakstan, and Sr Lanka 1. The author frst selected varous defntons of natural tradng partner ncludng Lpsey-Summers volume of trade, geographcal proxmty and trade complementarty. On the volume of trade crteron, the work fnds that only Bhutan and Nepal whch have strong tradng lnks wth Inda on account of beng landlocked qualfy for such characterzaton. The other countres have overall very low ntra-group shares to qualfy as natural tradng partners. He equally demonstrated that the countres do not qualfy on the bass of the second crteron geographcal proxmty. He showed that the SAPTA countres all demonstrate tendency to trade ntensvely wth partners outsde the regon. Ths was ether due to comparatve endowments or long hstorcal, relgous, cultural or other afflatons. On trade complementarty, usng revealed comparatve advantages (RCA) ndces, he found that Inda and Pakstan have ndces greater than one, ndcatng that ther exports complement the mports of a number of countres n the regon. RCA ndces of other countres n the regon showed lmted complementarty. The study concludes that the countres n SAPTA can be characterzed only moderately as natural tradng partners. Gven that much of the factors that led to ths characterzaton are entrenched n the hstory, prevous polces and other mpedments n the regon, the author concludes that the trade structures that have evolved among these neghbours may not facltate rapd ncrease n ntra-group trade. In contrast, he thnks there s evdence that prevous unlateral trade lberalzaton efforts by ndvdual countres yelded more pronounced postve results n boostng ntra-group trade. He therefore encouraged that these countres should contnue the process of unlateral trade lberalzaton as t would more lkely facltate export dversfcaton. Ths way, they would hopefully also evolve new comparatve advantages and complementartes that would support regonal ntegraton. Denns (2006) also assesses ntra-group trade and trade facltaton measures, wth a slghtly dfferent bent from Ptgala s work. He studes the Mddle East and North Afrca (MENA) regon and focuses on the development prospects of trade agreements and facltaton measures that have been put n place n the regon. The assessment of the mpact of trade agreements was twopronged; an analyss of the welfare gans accruable from the myrad regonal trade agreements that are poppng up n the regon and a comparson of these wth the gans accruable from trade 1 SAPTA members have n prncple agreed to fully mplement a free trade area transformng the group to SAFTA (South Asan Free Trade Area) begnnng 2006 and full mplementaton completed between 2009 and

8 agreements wth the European Unon. The study employs the Global Trade Analyss Project (GTAP) computable general equlbrum model wth modfcatons to capture trade facltaton through techncal progress n tradng actvtes n lne wth Hertel et al (2001). Ths modfcaton allows for nputtng hgher ndrect (ceberg) costs owng to longer and more cumbersome transt processes n trade between any two countres. He equally ncorporates new nsghts from Fox et al (2003) and OECD (2003) allowng for drect (tax component) cost mputaton. For the latter, he uses estmates from a survey by Zarrouk (2003) that calculated neffcences n trade facltaton as a proporton of the value of traded goods. The work also assumed that each day saved n shppng tme be equal to 0.5 percentage pont reducton n ad valorem tarffs (as found by Hummels, 2001 probt estmaton). The work fnds that ntra-group ntegraton and ntegraton wth the EU have postve mpact welfare n the regon. However, the welfare gans from ntegratng wth the EU were about twce those from wthn-group ntegraton. Addng trade facltaton measures helped to further shore up the gans to about three tmes ther orgnal values. For example, the welfare gans from ntegratng wth the EU were estmated to ncrease by 0.82 percent of GDP (from $1.8 bllon to $7.2 bllon) whle gans to ntra-group ntegraton were estmated to ncrease by 0.1 percent of GDP (from $913 mllon to US$3 bllon) Baller (2007) examned the trade facltaton queston from a more mcro and sectoral perspectve. Gven the poston of several frm-level surveys that dentfy techncal regulatons, rules of orgn and customs procedures as key non-tarff barrers, hs research focuses on greater enqury nto the nature of and solutons to techncal regulatons (termed techncal barrers to trade n the WTO). He pcked two ndcators wthn the body of techncal regulatons the effects of TBT lberalzaton n the form of harmonzaton and mutual recognton agreements (MRA) for testng procedures. The study frst structured a theoretcal poston derved from Meltz (2003) to formalze expected mpacts of harmonzaton and MRAs programmes and appled the theory to the data. The emprcal work n turn examnes sectoral mpacts of TBT lberalzaton on both partcpatng and excluded countres usng a two-stage gravty model. The major value that the work attempted to add s to stuate the emprcal analyses on more frmly grounded theoretcal analyss a major gap n gravty models. Identfyng two potental channels for ncreased trade flows from endogenous frm selecton process (an ntensve margn reflectng the volume of exports by each exportng frm and an extensve margn reflectng the proporton of frms exportng), the work analyzed surveys of telecoms and medcal devces ndustres. The frst part of the fndngs on MRAs s consstent wth apror expectatons namely that they do have postve mpact on both export probabltes and trade volumes for partner countres. The mpact of harmonzaton on the other hand was not as sgnfcant for ntegratng countres and dd not seem to matter for excluded developng countres, but was postve and huge on excluded developed countres. Equally, the probablty that harmonzaton would brng n new exportng frms was hgher than the probablty that exstng frms wll ncrease volume of ther exports. Some other works use general equlbrum models (the most pronounced of whch are GTAP CGE models). Some n ths category nclude APEC (1999), Fox et al (2003), OECD (2003) and Hertel and Keeney (2005). APEC (1999) blazed the tral n ths respect, wth the work on ts member countres. It fnds that the members ncome can be boosted by up to 0.4 percent of ts 1997 value (about $75 bllon) by mprovng trade lberalzaton and facltaton measures. Fox et al (2003) accounted for both drect (tax) and ndrect (ceberg) costs of trade facltaton costs between Mexco and the US and fnd $1.8 bllon and $1.4 bllon annual welfare gans for the Mexcan and US economes respectvely. In turn OECD (2003) puts estmated gans from 8

9 reductons n trade transactons costs of 1 percent of value of global trade at some $40 bllon annually. Hertel and Keeney (2005) feedng estmates from Wlson et al (2004) nto ther model note that facltaton-nduced gans from $150 bllon ncrease n global merchandze trade s about $110 bllon per annum. IV. Methodology a. The Basc Model The basc estmaton framework for ths work shall be the gravty model. The gravty model has come to be a popular formulaton for statstcal analyses of blateral flows of goods and servces between any two countres. Proposed by Tnbergen n 1962, t follows the basc Newton s law of unversal gravtaton n physcs that defnes the attracton between any two objects. In lne wth Newton s ntuton, standard representaton of the gravty equaton n trade relatons assumes the amount of trade between countres to be an ncreasng functon of the sze of each country represented by ts output, and a decreasng functon of the obstacles to trade represented by the dstance between the two countres. For the latter, the usual measure s the dstance between the captal ctes of the countres under consderaton, vewed as the economc centres. Snce Anderson (1979), a number of studes have tred to gve an economcally plausble theoretcal dervaton and justfcaton for the gravty model. As the efforts go, so do the assumptons underpnnng the dervaton. However, the basc consderatons are farly the same and we brefly outlne same underneath followng Head (2003). If the proporton of own ncome that a country j spends on all goods from all sources be gven by M j and the share of ncome that t spends on goods from a partcular country be s j. Clearly, s j wll range between 0 and 1, be postvely related to country s product varety (n) and qualty (u) as well as be negatvely related to the dstance between countres and j (Dj). Relatonally, that mples s j = g( η, µ, D e ) g( µ, η, D e e j ej ) (1) Anderson s approach to defnng g µ, η, D ) s to assume a sngle good for all countres (.e. ( e e ej equaton n to 1) and allowng market preferences u to vary. However, Bergstrand (1985) rather uses the Dxt and Stgltz model of monopolstc competton between dfferentated but symmetrc frms to assume that the varety coeffcent n vares wth ncome M whle equatng preferences u to 1 (.e. assumng t s same across countres). Head (2003) showed that for as long as the goods from the same country are of the same average qualty and subject to the same transport costs even though they mght be dfferentated, allowng both qualty and varety of goods to vary wth elastcty of substtuton gven by σ wll gve the g() functon (the denomnator) as g (2) 1 () = n ( / µ ) σ p j j 9

10 Agan, relatng the export prce to the prce n the orgn country gven transportaton costs wll gve p / = ) (3) j δ µ j ( p / µ Dj Where p s the prce of goods from country n country j, D s the dstance between the two countres and δ represents exportng country s varetes. Assumng, accordng to Dxt and Stgltz that all frms are the same sze (q), then n = M/q and defnng δ(σ 1) = θ 0, then θ σ 1 g() = M D /( qk ) (4) Implyng that market share for exporter s gven by s j = M D R (5) θ j j Rj, a term that can represent remoteness s gven by θ [ M D ] R = 1 (6) j e e ej However, t has been severally shown that n a frctonless world, Rj = 1/Mw (wth Mw representng world ncome). For estmaton purposes, we smply take the natural logs of ncome, trade flows and dstances to obtan ln F = a ln M + bln M c ln D + d ln R + ε (7) j j j j j Whle the standard gravty model does well n predctng and/or explanng trade based on just ncome and dstances of two countres, t leaves out a sgnfcant amount of unexplaned varaton n trade (Head, 2003). As a consequence, many works (ncludng Glck and Rose, 2002; Carrere, 2004; Rose, 2001; Frankel and Rose, 2002; Rose and Engel, 2002; Wlson, 2003, 2004; Njnkeu et al, 2008) augment the tradtonal gravty model. Most of such augmentaton are not exactly very theory-based, but are rather rooted n understandng of the underlyng propertes of the economes beng modelled. Tradtonal varables for augmentaton generally nclude ncome per capta, adjacency, common language or colonal tes, border effects and membershp of regonal ntegraton arrangements (Head, 2000 among several others). The ratonale behnd ncome per capta as an addton to ncome s that hgher ncome countres generally trade more than lower ncome ones. In the same ven, countres that are contguous mght trade more (partcularly there mght be preponderance of along-the-border tradng that are not captured by dstance). Agan, t has been shown that countres wth common hstorcal, colonal and language lnks tend to trade more wth each other than those that do not share these lnks. A classc example gven n the lterature on ths s Inda s trade wth the Unted Kngdom and Inda s trade wth Pakstan. Whle a standard gravty model wll predct same volume of trade for Inda- Pakstan and Inda-UK (or even more for the former), actual trade s much hgher for the latter owng to hstorcal and language tes. In the case of the SADC whch s the man focus of ths work, border and regonal ntegraton effects offset. The border effect smply denotes the frcton that occurs n trade on account of geographcal demarcaton of countres whch hnders 10

11 unfettered networkng among frms. Regonal ntegraton effects show the mproved mpact on trade among countres that may arse on account of membershp to the same regonal ntegraton arrangement. In partcular, wth the attenton of ths work beng on trade facltaton, we need such augmentaton as descrbed above. Indeed, several works lke Lmao and Venables (2001); Clark et al (2004) and Njnkeu et al (2008) use augmented gravty models to explan the mpact of trade facltaton. b. Augmented Model Trade Facltaton Representaton A key challenge of emprcal analyses of trade facltaton s that of defnton and data. Maur (2008) ctng Wlson (2002) noted that there s no unversal understandng of what trade facltaton s, reflectng dfferences, as well as some evoluton, n vews of what should be the reforms undertaken to reduce the cost of tradng. He then proceeded to defne trade facltaton as the smplfcaton of the trade nterface between partners. Such trade nterface s composed n a broad sense of complance to government rules by traders, enforcement by authortes of these rules (ncludng taxes), exchange of nformaton, fnancng, nsurance, ICT and legal servces, transport, handlng, measurement and storage. There are also tangble and ntangble aspects of such facltaton measures. The vast coverage of ssues n trade facltaton also reflects n dfferences n perceptons and understandng of trade agreements. But even beyond the conceptual, capturng the specfc measures of emprcal data that should feed nto measures of trade facltaton s even more dffcult. Maur (2008) underscores trade facltaton reform as the sum of efforts undertaken at the natonal, regonal and multlateral level desgned to reduce trade transacton costs. The mplcaton of the above s that there are dverstes n the polcy nstruments and measures that could be adopted as trade facltaton and even more dverstes n the emprcal proxyng of nstruments for t. Thus, the need of any emprcal work n the area s to clearly justfy the use of any nstrument relatve to ther relevance to the specfc trade ssues of the geographcal area under dscusson and of the objectves of the specfc work beng undertaken. To help muddle through the facltaton maze, Wlson (2003) adopted four broad measures that accordng to them generally should meet polcymakers needs. These are ports effcency, customs envronment, regulatory envronment and servces sector nfrastructure. Whle there seems to be some consensus that these are relevant measures, there s sgnfcant dversty n approaches to constructng the ndcators that should feed nto each of the measures. For example, n a subsequent work (Wlson et al, 2004), they use a mx of ndcators from three surveys; the frst by Kaufmann et al (2002), the second from World Economc Forum Global Compettveness Report ( ) and the thrd from IMD Lausanne, World Compettveness Yearbook (2002). Adoptng the same measures, Njnkeu et al (2008) use dfferent ndcators drawng manly from World Economc Forum Global Compettveness Report ( and ). The general understandng then s that wthn the broad range of acceptable measures for trade facltaton, ndvdual works ncorporate ndcators that are most relevant to the ssues and the group beng dealt wth. Ths work wll augment the four broad measures proposed by Wlson (2003) n two ways. Frst, the composton of the ndcators wll be slghtly dfferent. Ths mght lead to adjustment n the nomenclature for dentfcaton of the measures so constructed. Secondly, we drop one of the measures (customs envronment) and replace t wth two measures, number of documents 11

12 requred for trade and tme taken for trade (both drawn from the World Bank Dong Busness Scoreboard). In our vew, these are broader measures of polces and programmes to facltate trade among countres. So we shall construct ndcators along fve measures, three of whch are taken from the Global Compettveness Report (GCR) of the World Economc Forum. The last two wll come from the Dong Busness Report of the World Bank. The measures are entry and ext nfrastructure (whch shall ncorporate ports effcency, but nclude an addtonal varable llegal payments for exports and mports), regulatory envronment and domestc nfrastructure (ncorporatng physcal and socal nfrastructure at a broader level than that used n any prevous work). These three measures are all drawn from the Global Compettveness Report. The fourth measure s documents for trade (comprsng number of documents requred for exports and mports) whle the ffth s tme taken for trade (aggregatng tme taken for mports and exports), the last two beng drawn from the Dong Busness Report. Entry and ext nfrastructure wll be the average of the three ndcators of the ratng of qualty of facltes n both ar and sea ports as well as llegal payments for mports and exports. The ratonale s that the three show the state of support physcal and logstc nfrastructure for entry ponts for most countres and the qualty of nfrastructure n these wll defne the ease of both human and commodty traffc n and out of the country. In addton, the last ndcator captures both tarff and non-tarff barrers and complements the tme and documents for trade n the Dong Busness Report whch shall also be used n ths work. The regulatory envronment measure s the average of three ndces ncludng wastefulness of government spendng, favourtsm n decsons of government offcals and effcency of legal framework. Waste n government spendng reduces avalablty of fundng for trade-facltatng programmes whle favourtsm n government decsons mpact on effcency of resource allocaton that s necessary for effectve productvty and trade. The legal framework on the other hand largely crcumscrbes the ease of dong busness and provdes ncentves for approprate behavour by economc agents whle protectng the vulnerable. Domestc nfrastructure measure comprses the average of three ndcators overall nfrastructure qualty, qualty of electrcty supply and qualty of the educatonal system. It measures the broad ndces of support physcal nfrastructure avalable for busness and trade n a country and largely mpacts on trade qualty and volume. Electrcty supply and qualty s consdered especally mportant for Afrcan countres where ths s stll a major mpedment to productvty and trade. The qualty of the educatonal system complements the ndcators on physcal nfrastructure by ncorporatng a human development angle to growth and trade mperatves. The two measures from the Dong Busness Report are Tme for Trade and Documents for Trade. Tme for trade wll take the average of tme for mports and exports whle Documents for Trade wll take the average of documentaton for both mports and exports. The tme for mports and exports n the Dong Busness report s reported n number of days t takes for mport and export respectvely. Documents for mports and exports are recorded n terms of the number of documents requred to be able to mport and export respectvely. Both are absolute numbers taken from hard data unlke n the Global compettveness Report where they are ratngs. Incorporatng these measures n the model wll yeld. 12

13 ln(1 + F + α D t j ADJ ) = α t t t 1 lny + α2 lnyj + α3 ln(100 + TARj) + α4dj + α5 ln + α RE + α INF + α TT + α DT + α PE + α RE α D 16 SADC 8 + α D 17 9 ECOWAS 10 + α D 18 j CEMAC 11 j + α D 19 + α INF + α TT 12 COMESA j EEI + α D ENG j + α DT + α D FR j + ε j (8) Note that n the place of M, we have chosen to use standard notaton for ncome Y; TAR s tarff for countres and j; D s dstance and D ADJ stands for dummy for adjacency; D SADC, D ECOWAS, D CEMAC, D COMESA, are dummes for the dfferent regonal ntegraton groupngs represented by countres n the sample; D ENG and D FR represent the two major languages n Afrca, Englsh and French wth the former beng the offcal language among SADC member countres. Both theoretcally and based on prevous emprcal works, ncome mpacts on trade postvely (Soderstrom 2008; Chasrsawatsuk and Chasrsawatsuk, 2007; Zak, 2008; etc). So the expected sgn of the coeffcent of Y s postve. Tarff should have a negatve mpact on trade. The GCR ndces are based on ratngs that range from 1 to 7; countres performng well on the ndces are rated hgh whle those not performng well are rated low. As such, hgher scores ndcate relatve strength n a partcular ndcator. As such, mprovements on ndcators of trade facltaton ncludng entry and ext nfrastructure (EEI), regulatory envronment (RE) and nfrastructure (INF) should have a postve mpact on trade flows. On the other hand, the Dong Busness report for the ndces we are usng are based on hard data on the number and amount of documents and tme taken to transact busness. As such, countres wth lower numbers of documents and tme are consdered to perform better than others wth hgher numbers. As such, reducton n number of documents and tme for tradng should mprove trade flows. So the ndcators TT and DT are expected to have negatve sgns. Adjacency, membershp of SADC regonal groupng and common language should mpact trade postvely. But membershp of the other regonal trade arrangements should mpact on South Afrca s exports negatvely gven that t s not a member of these. c. Country Representaton and Data The data for the study shall nclude 23countres from Afrca (10 of whch are SADC member countres and the rest spread among ECOWAS, COMESA and CEMAC). For a complete lst, see Appendx Table 2. Data wll manly be drawn from the Global Compettveness Report (GCR), whch over the last one decade, has metamorphosed nto the most comprehensve rankng of compettveness ndces across countres. As at 2007, t covered about 25 Afrcan countres wth 92 ndcators spread over 23 measures and therefore facltates access to relevant measures of compettveness facltaton n data-poor and weakly covered Afrca n useful dmensons. Equally, the Dong Busness Report (DBR) of the World Bank s a very measured assessment of selected ndces of support to busness actvtes across countres. Indcators n the DBR are fewer than those n the GCR wth the former havng 31 ndcators spread across 10 measures. However, t covers more countres havng about 181 countres as at Consequently, wthn the range of ndcators that t covers, attenton to Afrca s consderably broad enough to help trade facltaton research. The two publcatons shall form the man data sources for trade facltaton ndcators n ths work. 13

14 The rest of the data are collated from a varety of sources ncludng the Easy Data database consstng of more than annual, quarterly and monthly tme seres from as early as 1988 to the present. Over 170 countres and 20 regons (Afrca, SADC, EC, NAFTA, etc.) and 15,000 commodtes wthn 145 ndustres at all HS and ISIC levels are avalable. Databases accessble through the World Integrated Trade Soluton (WITS) software desgned to ntegrate several trade related databases and provde easy access nclude TRAINS (UNCTAD), COMTRADE (UN Statstcs Dvson) and Integrated Database of the WTO (IDB-WTO) and Consoldated Tarff Schedule (CTS-WTO) are used. Data on dstances were obtaned from Unversty of Essex Project webste; calculated usng the great dstance formula that apples the longtude and lattude of the captal cty of each country n the sample. Indvdual countres GDP and mports from South Afrca were deflated usng each country s GDP deflators obtaned from USDA (and calculated from a combnaton of sources ncludng the World Development Indcators, ERS estmates and baselne regonal aggregatons). There were some gaps n the data that we had to ether fll usng a varety of approaches. For example, tarff profles for Ghana, Gamba and Zmbabwe for 2006 were not avalable. So the tarff rates for 2006 were obtaned by a lnear combnaton of the values. For a few other countres, the unavalable profles were for We agan obtaned such through a lnear trend of the 2006 and 2007 values. No data exsts for trade facltaton measures for Ghana n 2007; so we had to use the same approach to obtan ts 2007 values based on scores n 2005 and Data on trade facltaton for Zamba for 2006 and Benn and Cameroon for 2005 were mssng and we calculated these usng ther relatve scores n 2005 and 2007 for the frst and scores n 2006 and 2007 for the last two. No data exsts for all sample countres n the Dong Busness entry for We had to leave ths as zero for all countres. Ths s lkely to affect the results, but we thnk that mght be better than to construct data for all countres across two measures for two years whch wll no doubt ntroduce seral bas nto the results. In the Appendx Table 1, we try to show the correlaton among the trade facltaton varables. Expectedly, the correlaton among the varables from each source of trade facltaton s postve and qute hgh (0.77 for entry and ext nfrastructure and regulatory envronment; 0.94 for entry and ext nfrastructure and domestc nfrastructure and 0.75 for regulatory envronment and domestc nfrastructure n the GCR and 0.83 for documents for trade and tme for trade n the DBR). However, across sources, the correlaton s weak and mostly negatve. V. Emprcal Results The results from estmatng equaton 8 wth dfferent sets and subsets of the sample are shown n Table 2. The estmaton s a two step procedure that frst obtans estmates usng both the tradtonal and augmented varables of equaton 8 wthout country fxed effects and afterwards, ncludes only the trade facltaton varables alongsde ether GDP or populaton, wth country fxed effects. The essence of the two stage process s to be able to adequately seve the mpact of the trade facltaton varables ndependent of nose from other varables n the model. Table 2: Estmaton Results 14

15 A Wthout Fxed Effects B Wth Fxed Effects Varable Coeffcent t-statstc Coeffcent t-statstc C a/ b/ c/ d/ e/ GDP a/ b/ c/ d/ e/ Populaton Tarff a/ b/ d/ d/ Exchange Rate a/ c/ d/ e/ Dstance Trade Documents Trade Tme Domestc Infrastructure Entry and Ext Infrastructure Regulatory Envronment Englsh Language SADC Membershp ECOWAS Membershp CEMAC Notes: a/ represents estmate from models nvolvng only tme for trade b/ represents estmate from models nvolvng only domestc nfrastructure c/ represents estmate from models nvolvng only number of documents requred for trade d/ represents estmate from models nvolvng only domestc regulatory framework e/ represents estmate from models nvolvng only entry and ext nfrastructure 15

16 Besdes, the tradtonal gravty model has just two major explanatory varables output and dstance. Snce dstance s not ncorporated here, t then makes sense to evaluate our facltaton varables wth only output at least as a frst step n understandng ther mpacts. The two results are shown n panels A and B of each table. Appendx Table 2 on the other hand summarzes the ndvdual country fxed effects obtaned from panels B of Table 2. All varables wth the excepton of tme for trade and documents for trade are log-lnearzed. Consstent wth our hypothess, we could not confrm postve mpact of output for most of the modellng wthout country fxed effects. In fact, n many of the estmates, the coeffcent of output was negatve sgnallng that the weak output base of Afrcan countres mpacted negatvely on trade. Ths seems to ndcate that on account of relatvely low GDP, trade n ths regon s drven more by some other varables. For example, the populaton varable s one of the most potent. For many versons of the estmaton, populaton was consstently more sgnfcant (and postvely sgned) than output. Ths may mply recognton of the potental mpact of populaton n market exploraton even n the face of low ncome of some Afrcan countres. Ths mght be the case n trade wth a country lke Ngera where ncome level s low but hgh populaton may reflect potentally large market that can be explored for the future. The exchange rate varable also has a very sgnfcant postve coeffcent n the model wthout country fxed effects mplyng that deprecaton helps mprove trade expectedly so! However, ntroducng country fxed effects reduces the overall mpact of exchange rate. Whle the coeffcent retans a postve sgn, t s no longer sgnfcant. The mpact of tarff n the model wthout country fxed effects was so nsgnfcant and sometmes wrongly sgned that t had to be dropped. However, turnng to the model wth country fxed effects, the coeffcent assumed the rght sgn. But t stll remaned largely nsgnfcant relatve to other broad macroeconomc varables n the model. In our vew, ths s partly a reflecton of the decreasng role of tarffs n overall trade an ssue that formed the bass of assessment of extra-tarff constrants to trade as n ths study. As rghtly noted by Baldwn, the WTO has succeeded n drastcally reducng tarffs across natons over the last half century. For most of the modelng, t was not possble to confrm a coeffcent for dstance algned to theory our estmate has a postve sgn. The postve sgn remaned consstent throughout the dfferent sub sample and cross secton selectons and wth dfferent combnatons of the explanatory varables. Ths, n our vew, s not exactly as counterntutve as t may appear at frst. In fact, t seems to be consstent wth our hypothess that theoretcal postulaton of negatve mpact of dstance for trade may not hold n all cases. Prmarly, gven ts geographcal poston, elte status n Afrca and unque hstory, South Afrca s prmary trade and economc relatons seem to be more wth Europe and the rest of the World than wth neghbourng countres. The country s one of very few Afrcan countres that export manufactures and so trade ntensvely wth faraway countres. Whle many Afrcan countres trade wth South Afrca; the relatve share of such trade s mnscule compared to that from the rest of the world as we earler showed n Fgure 1. In effect, the estmaton results reflect the fact that most of South Afrca s bg tradng partners are dstant natons. Ths result calls nto queston the standard proposton of the gravty model wth a possble excepton to the rule worth further nvestgatng. Interestngly other studes n the past have also found smlar trends (see Ptgala, 2005 for example). More as f to 16

17 further buttress the dstance ndcator, adjacency equally showed up negatve ndcatng that relatve to the rest of the world, South Afrca trades less wth ts mmedate neghbours. Lke the dstance ndcator, the coeffcent of adjacency n the estmaton results s qute sgnfcant. Of course, as has been severally admtted, there mght be qute a szable proporton of across-theborder trade that mght not be captured by offcal data. Estmates from the avalable offcal data though seem to suggest that sharng borders does not necessarly place any of SA s mmedate neghbours at an advantage relatve to other tradng partners. Ths may be partly on account of technologcal advancements n both transport and communcaton that have greatly shrank the mportance of proxmty or t may be on account of the relatve sze of output of the neghbourng countres. Agan, as n the dstance varable, we had to drop adjacency. We could not confrm the sgnfcance of some other augmented varables n the model lke Englsh language. An nterestng result arses wth the coeffcent of the SADC membershp dummy. Taken alongsde other standard determnants of trade, t consstently showed up negatve. Even though the coeffcent s not very sgnfcant, the natural tendency would be to nfer that SADC as a group has negatve mpact on SA s trade wth ts neghbours. However, gven that the sample s shorter than would have allowed an explct nference of ths sort, t may be too early to come to ths concluson. A better approach s to agan pont t out as an ssue worth further nvestgatng. Interestngly, far-away ECOWAS has a postve mpact on exports from South Afrca, probably because of Ngera, but t was not sgnfcant enough to mert attenton. The coeffcent of CEMAC membershp was negatve but largely nsgnfcant. Among the trade facltaton measures, entry and ext nfrastructure seems to be the most sgnfcant factor that affects trade. Ths result s consstent for both estmatons wth and wthout country fxed effects. The data for constructon of the ndex nclude arport and seaport nfrastructure and rregular payments for mports and exports. To a large extent, these summarze the logstc challenges that drectly face trade (Clark et al, 2004). The results seem to ndcate that whle such factors as domestc regulatory envronment, whch deal wth regulaton at home may be mportant for overall productvty and possbly for exports, ther mpact on mports at the global level may be margnal. The same goes for domestc nfrastructure whch does not seem to have very drect nterface wth goods from other countres at the tme of exchange. The results show both varables are margnal for determnng mports from South Afrca by other countres. Beng qute encompassng and reflexve of trends n the other varables, the results on entry and ext nfrastructure s not surprsng. About 14 of the 22 countres n the sample (excludng South Afrca) have postve nfluence on South Afrca s exports (data for Namba and Botswana were a bt problematc though and ths could have accounted for the outcome for those countres). These postve mpacts reflect (ostensbly unexploted) potental for ncreasng trade. Thereafter, we brng n the ndvdual trade facltaton varables one at a tme to evaluate ther mpacts gven country specfc effects. Ths set of equatons was estmated wth only the GDP and/or populaton as prncpal accompanyng varable 2. The results are shown n panel B of Table 2. Whle GDP entered most of the varables (as opposed to populaton used n the model 2 Lmted degrees of freedom make t dffcult to estmate wth the whole array of basc and augmented gravty varables alongsde the trade facltaton varables when fxed effects of the many countres ncluded n the sample are taken nto account. 17

18 wthout fxed effects), t s sgnfcant at 5 percent probablty levels n 3 out of the fve equatons retanng only margnal (10 percent and above) sgnfcance n the others. Tarff entered four out of the fve equatons, but whle t was correctly sgned, t was not sgnfcant n any. The same goes for exchange rate. Among the trade facltaton measures, domestc nfrastructure, the number of documents requred for mport and export and tme (n days) taken for export and mport have clearly sgnfcant effects when country fxed effects are accounted for. The frst postvely and the last two negatvely; all accordng to specfcaton. The ndcaton then s that wthn Afrca, domestc nfrastructure, documentary requrements for tradng and tme taken for berthng and cargo clearance assume added sgnfcance. Ths result s consstent wth fndngs n Lmao and Venables (2001), Kurz et al (2008) among others that show that Afrcan countres sgnfcantly lag behnd n customs reforms that reduce documentaton and enhance turn around tme for cargo at ports. Dtto for domestc nfrastructure; a lot has been sad on ths and much more can stll be sad wth respect to trade wthn Afrca. Entry and ext nfrastructure was agan moderately sgnfcant (ths tme at 10 percent) whle the sgnfcance of the regulatory envronment could not be drectly confrmed. All varables reman rghtly sgned though. VI. Some Implcatons for Polcy and Research Ths work set out to evaluate the mpact of trade facltaton measures on trade wthn Afrca, usng exports from South Afrca. The study s undertaken wthn the context of conclusons by Aldaz-Carroll (2006) that developng countres face an ncreasng need to upgrade the standards of ther domestc markets and of ther exports. To do so, the work examned exports from South Afrca to 23 countres n Afrca, 10 of whch are n the SADC. We constructed fve measures from a total of 13 ndcators from the Global Compettveness Report of the World Economc Forum and the Dong Busness Index of the World Bank. Usng an augmented gravty model, we obtaned estmates from data coverng the perod 2005 through The estmaton results seem to support the kck-off hypotheses that there are exceptons to the provsons of theory on the mport of dstance, output and tarff on trade among contguous countres. For exports from South Afrca, populaton seems to be more mportant than output, ndcatng attenton to market potentals rather than current purchasng power when country fxed effects are not taken nto consderaton. When t s, output becomes mportant, and s regularly more mportant than populaton. Wth the dsmantlng of tarff barrers, exchange rate nfluences have become crtcal n determnng trends n trade. The coeffcent of the dstance ndcator s partcularly atheoretcal. Whle theory provdes for a negatve mpact, estmates ndcate postve mpact showng that South Afrca exports more to dstant natons than to ones that are nearer t. In ths same drecton, the coeffcent of adjacency was negatve mplyng that beng close endows no greater advantage to any of the countres n the sample relatve to others. A most compellng result, though not as sgnfcant as other relevant varables, s that of the coeffcent of SADC, whch showed up to mpact negatvely on exports. Whle the outcome of ncludng other regonal groupngs n the model could be gnored for the tme beng, the mplcatons of a negatve sgn for SADC regonal groupng for South Afrca s exports are not mnscule. Importantly, the sgnal beng sent s that the hgh nvestment nto the formaton of the regonal groupng currently have had net negatve mpacts on exports from South Afrca to the rest of the countres n the regon, consstent wth ndcatons from Fgure 1. Certanly, there are dynamc gans from regonal ntegraton (Neary, 2001; Coe et al, 2005, Velde and Meyn 2008, 18