ANALYSING THE TRADE EFFECTS OF THE EU-SA & SADC TRADING AGREEMENTS: A PANEL DATA APPROACH

Transcription

1 SAJEMS NS 14 (2011) No ANALYSING THE TRADE EFFECTS OF THE EU-SA & SADC TRADING AGREEMENTS: A PANEL DATA APPROACH André Jordaan and Patrck Kanda Department of Economcs, Unversty of Pretora Accepted February 2011 Abstract Ths study nvestgates the trade effects of the EU-SA and SADC preferental trade agreements of whch South Afrca s a member. Usng a panel data estmaton of the gravty model of blateral trade and based on data from 1994 to 2008, the study fnds the EU-SA preferental trade agreement to have a sgnfcant trade expanson effect. The study further reveals that an nformatve concluson on trade effects of the SADC preferental trade agreement can only be reached once the agreement has been fully operatonal. The study also recommends that trade polcy n South Afrca should ncreasngly be geared towards broad-based multlateral lberalsaton. In addton, South Afrca should promote regonal economc stablty and development through supportng regonal trade agreement ntatves. Key words: trade creaton, trade dverson, preferental trade agreement, panel data estmaton, gravty model of blateral trade JEL: F14; F15 1 Introducton Preferental Trade Agreements (PTAs) have ncreased n number durng the last decade n both developng and developed countres (Musla, 2005:117). PTAs have been sad to enhance trade by approxmately 50 per cent whereas a recent report from the Natonal Bureau of Economc Research (NBER) suggests that PTAs n fact trple the volume of trade between members n the agreement (Head, 2003:11). South Afrca s currently a sgnatory to two sgnfcant Preferental Trade Agreements (PTAs): the European Unon- South Afrca (EU-SA) agreement sgned n 1999 and the Southern Afrcan Development Communty (SADC) agreement ratfed n Prevous research on South Afrca s nternatonal trade such as Chauvn and Gauler (2002), Rangasamy and Brck (2007), Smet (2007) and Kalaba (2007) has focussed on analysng ssues such as trade lberalsaton, export performance, comparatve advantage, trade wth developng countres vs trade wth developed countres, and the determnant of South Afrca s nternatonal trade. Some other studes such as Rangasamy and Blgnaut (2005) and Cassm, Onyango and Van Seventer (2004) have nvestgated the openness and compettveness of the South Afrcan economy. However, few studes (Kwentua, 2006 and Holden & McMllan, 2006) have nvestgated the mpacts of the EU- SA and SADC preferental trade agreements to whch South Afrca s a promnent sgnatory. Followng varous trade reforms n the 1990s, the South Afrcan economy has been charactersed by an mprovement n nternatonal trade (Rangasamy & Blgnaut, 2005:375). Prevous research also ndcates that the EU and the USA are some of the most mportant destnatons of South Afrcan exports n contrast wth the SADC whch, compared to the other groupngs, s a relatvely smaller tradng partner (Smet, 2007:14). On the other hand, evdence based on data suggests that durng the perod from 1994 to 2008, there has been a sgnfcant declne and n some nstances stagnaton n trade between South Afrca and some of ts maor tradng partners n the EU, despte the fact that from 2000 South Afrca and the EU have been part

2 230 SAJEMS NS 14 (2011) No 2 of the same preferental trade agreement. However, trade wth countres such as Chna and Japan, wth whom South Afrca does not have a trade agreement, soared durng the same perod of tme. Kwentua (2006) fnds evdence of trade creaton n the EU-SA agreement and ncreased trade between EU-SA members and the rest of the world. Holden and McMllan (2006) also suggest that the EU-SA agreement enhanced both exports and mports whle the SADC agreement only stmulated exports for ts members. In the lght of emprcal evdence of reduced trade between South Afrca and some of ts maor tradng partners n the EU, as well as ncreased trade between South Afrca and countres that are non-members of the EU-SA or SADC agreements, the purpose of ths study s to nvestgate trade effects of preferental trade agreements n whch South Afrca s a member, namely the EU-SA and the SADC agreements. Ths study uses an augmented verson of the gravty model of blateral trade specfed n Ghosh and Yamark (2004) and a panel data estmaton of the gravty model of blateral trade as outlned n Martnez-Zarzoso and Nowak-Lehmann (2003) and Egger (2000). The rest of the paper s organsed as follows: after a revew of South Afrca s nternatonal trade envronment and related lterature n Secton 2, the methodology based on the gravty model of blateral trade framework s dscussed n Secton 3. Thereafter, an emprcal estmaton of the gravty equaton s conducted n Secton 4. Results are presented n Secton 5 and Secton 6 concludes the paper. 2 Lterature revew 2.1 South Afrca n global trade Openness of the South Afrcan economy Rangasamy and Blgnaut (2005: ) and Kalaba (2007:1) argue that the engagement of South Afrca n the global communty and the mplementaton of a seres of new economc polces n the 1990s caused a move from hghly protected and dstorted markets to open ones. Rangasamy and Blgnaut (2005: ) also report that externally-orented ndustres were found to have hgher rates of growth compared to other ndustres n the South Afrcan economy n the post 1990s perod. The drve of ths process of openng up the economy was enhanced by the endorsement n 1994 of the General Agreement on Tarffs and Trade (GATT) and the executon of PTAs wth the EU and the SADC n 1999 and 2000 respectvely. The salent facet of these agreements s the reducton of mport protecton. The EU-SA agreement has an asymmetrc nature whereby dutes on 95 per cent of the EU s mports from South Afrca wll totally be phased out by the end of the agreement s 12-year duraton; whle on the other hand, only 86 per cent of South Afrca s mports from the EU wll become duty-free durng the same perod. The SADC agreement, though created n 1996, was only endorsed by 11 of ts current members n Accordng to ths agreement, 98 per cent of ntra SADC trade wll have to become free of duty by the year Owng to the fact that South Afrca s the largest economy n the regon, the agreement requres t to undergo faster lberalsaton reforms than other countres n the communty (Mabugu & Chtga, 2007:5) Trade lberalsaton South Afrca s economy has experenced a steady course towards trade restructurng durng the past few decades due to domestc and nternatonal pressure. The obectve of ths reform was to enhance the allocaton of resources through more compettve and export-orented polces. A partcular emphass was drected to the export of non-gold commodtes. Consequently, some export promoton measures were adopted to counteract the ant-export bas of protecton and mprove the compettveness of some ndustral sub-sectors. However, these ncentves to restructure trade were later on negatvely affected by a sgnfcant apprecaton of the rand caused by an export boom of gold commodtes. Up untl the late 1980s, South Afrcan exports were domnated by prmary commodtes, especally gold

3 SAJEMS NS 14 (2011) No (Petersson, 2005:1). Accordng to Smet (2007:17), South Afrca s trade actvtes expanded after t oned the WTO n Chauvn and Gauler (2002:14) report that South Afrca decreased ts tarffs by approxmately 4250 tarff lnes between 1990 and 1996 wth the am of reducng tarff rates from around 210 to 6 n the same perod. Edwards (2005:774) argues that the lberalsaton of tarffs n South Afrca durng the 1990s caused an mportant declne n the level of effectve and nomnal protecton. There was a decrease of 8.7 per cent and 35.3 per cent n the manufacturng s mean nomnal and effectve protecton rates respectvely from 1994 to All traded ndustres reported a declne n protecton durng the same perod. As a consequence, fewer commodtes were affected by tarff dstortons n 2004 than n Nonetheless, the tarff structure stll remans ntrcate wth the tobacco, textles, clothng and footwear ndustres charactersed by hgh levels of effectve and nomnal protecton. Edwards and Lawrence (2008:585) argue that more trade lberalsaton should form part of polces amed at mprovng export dversfcaton n South Afrca. They also argue that trade polcy n South Afrca has exercsed a consderable mpact on the nature and growth of trade. Before the accesson of South Afrca to ndependence, exports and mports were sgnfcantly obstructed by trade protecton. Jonsson and Subramanan (2001:219) fnd evdence that trade lberalsaton also had an mportant contrbuton to the ncreased longrun growth potental of South Afrca through ts effect on total factor productvty growth Trade, compettveness and growth Accordng to Mtonga (2006:50-51), there was an ncrease n the volatlty of the real exchange rate of the rand possbly due to a msalgned currency 1 after the accesson of South Afrca to democracy n Followng the apprecaton of the rand, the compettveness of South Afrcan exports on the global market eroded because South Afrcan exports became more expensve. Ths n turn caused a reducton n producton, profts and employment. 2 Ths fact hghlghts the mportance of compettveness n nternatonal trade and ts relatonshp wth economc growth as the South Afrcan government s move to tackle the ssues of hgh levels of unemployment and prevalent poverty has been centred on realsng sustanable economc growth. Matthee and Naudé (2007:16) report that n addton to the quantty of exports, the nature and qualty of commodtes that are exported s very mportant n the analyss of export-led growth n the case of South Afrca. Peet and Koch (2005:1) argue that a foundaton of South Afrca s economc development polcy s export growth. Consequently, crcumstances surroundng foregn markets also affect the performance of the South Afrcan economy Features of South Afrca s nternatonal trade South Afrca has made mportant efforts amed at supportng blateral relatons wth ts man tradng partners. After the democratc electon of 1994, the EU and the USA allowed dutyfree access for South Afrca s exports to ther respectve markets. For nstance, the Afrcan Growth and Opportunty Act (AGOA) sgned n 2000 enhanced the accessblty of South Afrca s exports nto the US market (Petersson, 2005:1). Aspects nfluencng trade development n South Afrca nclude (Kalaba, 2007:4): Colonal and poltcal hstory: There s a strong trade relatonshp between South Afrca and the UK, Germany and the Netherlands; The nature of the commodtes: Afrca consttutes an mportant destnaton of exports of manufactured and fnshed consumer goods from South Afrca whereas mports of machnery, hgh-tech goods and electronc equpment nto South Afrca orgnate from developed countres; The development of nfrastructure: Maor constructon proects such as the Gautran ral lnk, constructon and upgrade of stada, road constructons, and upgrade of hghways have caused hgher mports of constructon machnery; The ncome level of the tradng partner: South Afrcan exports have the tendency to be destned to countres wth hgher per

4 232 SAJEMS NS 14 (2011) No 2 capta GDP; Exchange rates: Imports nto South Afrca are enhanced by a strong rand to the detrment of exports. Edwards and Lawrence (2008:606) also argue that total mports and non-commodty exports were receptve to fluctuatons n the real exchange rate. Rangasamy and Brck (2007: ) argue that South Afrca s export performance s dctated by the Organsaton for Economc Cooperaton and Development (OECD) growth n the long run. They also argue that South Afrca s exstng tradng relatonshps should be exhaustvely exploted n addton to mprovng the compettveness of South Afrcan exports as well as broadenng export destnatons. Ther study also dentfes three aspects charactersng trade between South Afrca and ts man tradng partners. Frst, the OECD represents a sgnfcant tradng partner of South Afrca wth 65 per cent and 59 per cent of South Afrca s exports and mports respectvely beng accounted for by trade between South Afrca and the OECD. Second, there s a huge concentraton of South Afrca s tradng lnkages,.e. two-thrds of mports nto South Afrca orgnated from a group of twenty countres n the perod from 1992 to Moreover, the very same countres absorbed approxmately three quarters of exports from South Afrca durng that same perod. Lastly, the destnaton of exports from South Afrca became more condensed after the 1990s whle on the contrary the sources of mports nto South Afrca were dversfed Trend analyss and composton of trade Table 1 shows that South Afrca ranked 26th and 22nd for merchandse exports and mports respectvely, not takng nto consderaton ntra EU trade n The stuaton s dfferent for commercal servces trade, wth South Afrca rankng 28th and 23rd for exports and mports respectvely, excludng ntra EU trade. Therefore, t can be concluded that though South Afrca has a relatvely better stance n global trade than most developng countres, there stll s room for mprovng ths current stance. Source: WTO (2009a) Table 1 South Afrca s rank n world trade (2008) Rank Exports Imports Merchandse Excludng ntra-eu trade Commercal servces Excludng ntra-eu trade Between 2003 and 2008, South Afrca has been a net mporter, consstently recordng a trade balance defct durng ths perod. Specfcally, South Afrca s a net exporter to Afrca and a net mporter as far as trade wth Europe, the Amercas and Asa s concerned (Quantec, 2009). Developed countres and the emergng economes mport natural resources from South Afrca because South Afrca has a sgnfcant endowment of natural resources and hence has a comparatve advantage for these goods. On the other hand, emergng economes and the ndustralsed economes manly export low-tech goods and hgh-tech manufactures to South Afrca respectvely. South Afrca s tradng partners n the SADC regon have a comparatve advantage n natural resources as well and therefore the composton of trade between South Afrca and ts neghbours dffers sgnfcantly from that between South Afrca and developed countres (Smet, 2007:14). Large trade volumes both n terms of exports and mports are accounted for by the trade between South Afrca and Asa, Europe and the Amercas. South Afrca mports more from Asa, Europe and the Amercas than t exports to them (Quantec, 2009). The most mportant exports from and mports nto South

5 SAJEMS NS 14 (2011) No Afrca are manufactured goods and mnerals (DTI, 2009). Ths goes n lne wth the fndngs of Smet (2007:14-21) that South Afrca s an exporter of mneral products and metals to the world. Imports nto South Afrca are manly consttuted by machnery, transport equpment and crude ol. The surge n the demand for transport equpment and machnery s manly responsble for the negatve trade balance n South Afrca. 2.2 Trade creaton and trade dverson Trade creaton refers to the net ncrease n trade resultng from a shft n hgh cost domestc goods to lower cost mported goods from a PTA member country. On the other hand, trade dverson refers to the shft of exstng trade from lower cost non-pta producers to a hgher cost PTA member producer. Flynn (2008:1) argues that trade creaton and trade dverson affect the economy dfferently. In general, trade creaton, whch results n a net economc gan, s the motve for countres to engage n PTAs gven the fact that the prce of an mported commodty s less than the domestc prce of the same commodty. On the other hand, trade dverson generates a net economc loss whereby domestc consumers pay hgher prces for mports from a hgh cost PTA member than they would have pad f the mports were sourced from a low cost non-pta member. The latest surge n the number of PTAs has revved the debate around the ssue of the effects of trade creaton and trade dverson and the relevancy of these agreements (Sarker & Jayasnghe, 2007:102). Accordng to Amposah (2002:2), the maor ssue wth a preferental trade agreement s whether the gans from trade creaton surpass the loss from trade dverson. Therefore a preferental trade agreement wll be consdered favourable f t produces greater trade creaton than dverson. Trade creaton and trade dverson form a maor component of economc ntegraton,.e. the amalgamaton of economc actvtes among countres. The ntensty of trade creaton and trade dverson dmnshes as the countres economes become hghly ntegrated (Flynn, 2008:3). Some authors such as Ghosh and Yamark (2004), Musla (2005), Cernat (2003), Vcard (2009) and Clausng (2001) suggest that generally, PTAs have net trade creaton effects whereas Rahman, Shadat and Das (2006) and Coulbaly (2004) fnd evdence that some other PTAs (e.g. SADC, MERCOSUR) are assocated wth net trade dverson effects. Sarker and Jayasnghe (2007) fnd evdence that the EU has sgnfcantly ncreased agr-food trade among the members at the expense of trade wth non-members, whch reduced sgnfcantly. Usng a cross-sectonal estmaton of the gravty model of blateral trade based on 1998 data from a sample of 39 countres, Kwentua (2006) nvestgates the trade creaton and trade dverson effects of the EU-SA agreement and fnds evdence that both trade between members of the EU-SA agreement and trade between members and non-members of the EU-SA agreement ncreased, and therefore concludes that the EU-SA agreement s trade creatng. Kwentua (2006) ponts out that the ncrease n trade between EU-SA members and the rest of the world could be attrbuted to an ncome effect. Based on 1994 to 2004 data coverng 136 countres and usng both cross-sectonal and panel estmatons, Holden and McMllan (2006) nvestgate whether the EU-SA and SADC agreements have had any effects on South Afrca s trade. Ther analyss also extends to the Afrcan Growth and Opportunty Act (AGOA) sgned n 2000 between the USA and a host of Afrcan countres. The cross-sectonal results fnd an nsgnfcant mpact whle the panel results fnd evdence of a postve mpact. Specfcally, the panel results show that the EU-SA agreement stmulated both exports and mports durng the perod 1994 to 2004 whereas the SADC agreement only stmulated exports. The AGOA results were not sgnfcant, ndcatng that durng that perod, South Afrcan exports had not benefcated from preferental access nto the USA market. 3 Methodology 3.1 The gravty model of blateral trade In emprcal research, the estmaton of trade creaton and trade dverson effects s manly carred out n two ways: the use of Computable

6 234 SAJEMS NS 14 (2011) No 2 General Equlbrum (CGE) modellng or the use of the gravty model of blateral trade. CGE modellng s relevant for ex-ante analyss,.e. analyss done before trade between two countres actually takes place. The gravty model s approprate for ex-post analyss,.e. analyss done after trade has taken place between two countres (Cernat, 2003:7). Tnbergen and Pöyhönen frst appled the gravty model to the study of nternatonal trade flows n the early 1960s. Snce then, a sgnfcant amount of research n nternatonal trade has used the gravty model of blateral trade as ther emprcal tool because the model gves a better ft to the maorty of regonal as well as nternatonal trade flows data sets (Mátyás, 1998:397; Mátyás, 1997:363; Cheng & Wall, 2005:49; Egger & Pfaffermayr, 2003:571). Many authors, ncludng Clausng (2001), Ghosh and Yamark (2004), Cernat (2003), Musla (2005), Rahman, Shadat and Das (2006), Sarker and Jayasnghe (2007), Vcard (2009), Carrère (2006) and Coulbaly (2004) have used the gravty model of blateral trade to assess trade creaton and trade dverson usng dummy varables that capture a country s membershp n a preferental trade agreement. The orgnal form of the gravty model of blateral trade assumes that trade flows between two countres can be lkened to Newton s gravtatonal force between two obects. Trade flows are drectly proportonal to the countres ncome (GDP) and nversely proportonal to the dstance separatng them. A set of dummes can also be added n the specfcaton of the model to account for factors enhancng or restranng the trade flow. After the md 1970s, there was a development n theores that supported the gravty model of blateral trade. Anderson (1979) makes the ntal formal endeavour to derve a gravty model of blateral trade based on product dfferentaton. Anderson and Wncoop (2003) further argue that the maor feature of the gravty model of blateral trade s the dependence of trade flows on a trade resstance factor. In a quest to prove the strength of the theoretcal foundaton of the gravty model of blateral trade, Oguledo and MacPhee (1994) derve a gravty model from a lnear expendture system. The basc functonal form of the gravty model of blateral trade s as follows: Y Y X D X where represents blateral trade flows (usually exports), Y s the GDP (economc mass of country (reporter), Y s the GDP of country (partner), D s the dstance between countres and. The stochastc log-lnearsed verson of the basc gravty model of blateral trade s as follows: log X where represents the whte-nose error term. Accordng to Martnez-Zarzoso and Nowak- Lehmann (2003:296), the generalsed form of the gravty model of blateral trade assumes that exports from country to country s a functon of the countres ncomes,.e. GDPs, ther populatons and the dstance between them, as well as a vector of dummy varables: X 0Y Y N N D A log logy logy log D u X where represents blateral trade flows (usually exports), Y (Y ) s the GDP of the exportng (mportng) country, N ( N ) s the populaton of the exportng (mportng) country, D s the dstance between captal ctes or economc centres of the two countres, A captures any factor enhancng or restrctng blateral trade and u s the whtenose error term. Expressed n logarthmc form for estmaton purposes, the generalsed gravty model of blateral trade s as follows: x 0 1 y 2 y 3n 4n 5d h Ah u (h 5) h

7 SAJEMS NS 14 (2011) No where A h h represents the sum of preferental h trade dummy varables and Ah takes the value of one when a gven crteron s fulflled (for nstance beng a member of a preferental trade agreement), zero otherwse. Lower-case letters represent the logarthm of respectve varables,.e. exports, GDPs, populatons and dstance. 3.2 Model specfcaton Varables: The specfcaton of varables capturng trade creaton and trade dverson effects n the gravty model of blateral trade n ths study follows the model specfcaton n Ghosh and Yamark (2004): X 0Y Y N N ER D PTA PTA u X where s the exports from South Afrca to ts tradng partner, Y s South Afrca s GDP, Y s the tradng partner s GDP, N s South Afrca s populaton, N s the tradng partner s populaton, ER s the average real effectve exchange rate of the rand, D s the dstance between the captal ctes of South Afrca and the tradng partner, PTA s a dummy varable capturng whether South Afrca and ts tradng partner both belong to the same preferental trade agreement (SADC or EU-SA),.e. ntra bloc trade, PTA s a dummy varable capturng current membershp status of South Afrca n the preferental trade agreement. The correspondng stochastc log-lnear specfcaton s: x 0 1y 2 y 3n 4n 5er 6d 1PTA 2 PTA u PTA 1 f SA and PTA 1 f only SA ts partner both belong to the same PTA, 0 s a member of the PTA, 0 otherwse otherwse Contrary to the approach of defnng PTAs whereby the value of PTA dummes s constant over tme n a database and followng Coulbaly (2004), the dummy varables PTA and PTA n ths study are tme-varant. Therefore, these dummy varables wll take the value of one only from the effectve date of endorsement of the PTA. However, because these varables wll have a constant value over tme for some of the cross-sectons, they wll be estmated together wth tme-nvarant varables as n Coulbaly (2004:4-5). Expected sgns: The relatonshp between exports and both GDP measures s expected to be postve. A hgher GDP n South Afrca means a hgher producton capacty whch n turn translates nto the ablty of the South Afrcan economy to export more (supply sde). On the other hand, a hgher GDP for a tradng partner country means a hgher absorpton capacty,.e. the tradng partner country s able to mport more (demand sde). Accordng to Martnez-Zarzoso and Nowak- Lehmann (2003:296) and Armstrong (2007:5), there s no clear a pror relatonshp between exports and the populatons of both the exportng and mportng countres. The estmated coeffcent of the exporter s populaton could ether be postve or negatve dependng on whether the exporter has a large populaton and exports more (economes of scale) or the fact that the exporter has a large populaton but exports less (absorpton effect). In the same ven, the estmated coeffcent of the tradng partner country s populaton could ether be postve or negatve. Exports and the rand s exchange rate are expected to be postvely related as hgher rates of exchange (deprecaton of the rand) would mean that t s cheaper for the tradng partner country to source the requred amount of rands to effect payments for mports, resultng n hgher demand for South Afrcan exports. Beng a proxy for transportaton costs, dstance s normally expected to be negatvely related to the flow of exports,.e. the hgher the dstance, the hgher the costs nvolved n tradng and therefore a negatve effect on trade flows. However, as shown by Marmoutou, Pegun and Pegun-Fessolle (2009) and Brun, Carrère, Gullaumont and de Melo (2005), dstance can bear a dfferent role n a gravty

8 236 SAJEMS NS 14 (2011) No 2 model of blateral trade. Marmoutou et al. (2009) partcularly show that the larger the tradng partner country s GDP, the less the effect of dstance on trade flows. Trade creaton and trade dverson effects n nternatonal trade studes are usually measured by usng a par of dummy varables PTA and PTA (Ghosh & Yamark, 2004:215; Cernat, 2003:9). If 1 (the coeffcent of PTA ) s postve, ths s an ndcaton that there s more ntra PTA trade between South Afrca and ts tradng partner than predcted by the other varables of the gravty model of blateral trade. Ths s evdence of trade creaton. On the other hand, f 2 (the coeffcent of PTA ) s postve, then extra PTA trade,.e. trade between a PTA member and a non-pta country s more than a random par of countres. Ths ndcates the openness of the PTA to mports from the rest of the world. Smlarly, f 2 s negatve, then there s less trade wth non-pta countres. Ths s evdence of trade dverson. Alternatvely, Cernat (2003:9) gves the followng comprehensve nterpretaton of the sgns of the PTA dummy varables coeffcents: Trade creaton & trade expanson PTA Trade dverson PTA Trade expanson PTA Trade contracton PTA 0, PTA 0 0, PTA 0 0, PTA 0 0, PTA 0 4 Estmaton Accordng to Egger (2000:26) and Egger and Pfaffermayr (2003: ), many studes n the past have estmated the gravty model of blateral trade usng a cross-secton methodology. However, as dscussed n Egger (2000:26), Egger and Pfaffermayr (2003: ) and Martnez-Zarzoso and Nowak- Lehmann (2003:298), panel data estmaton of the gravty model of blateral trade has many advantages over cross-secton analyss. For nstance, the role of the busness cycle and the nteractons between varables over a long perod of tme can be captured by usng panel data analyss. In addton, country-specfc effects that do not change over tme can be unravelled. Another beneft of usng panel data analyss s that the rsk of gettng based estmates s lowered. The F-test s used to determne whether countres n the sample are homogeneous (poolablty) or heterogeneous,.e. the exstence of country-specfc ndvdual effects. The test result reects the null hypothess of poolablty n favour of the alternatve hypothess of country-specfc ndvdual effects by usng fxed effects. 3 As ponted out n Egger (2000:26), Martnez- Zarzoso and Nowak-Lehmann (2003: ), usng a panel data framework results n the ssue of choosng the correct specfcaton of ndvdual effects as ether random or fxed effects. Mátyás (1997: ), Mátyás (1998: ), Egger (2000:26), Egger and Pfaffermayr (2003:572), Cheng and Wall (2005:54) and Martnez-Zarzoso and Nowak- Lehmann (2003:299) opt for a fxed effects specfcaton of the gravty model of blateral trade. Egger (2000:26) and Martnez-Zarzoso and Nowak-Lehmann (2003:299) specfcally argue that a fxed effects model s approprate for studes that analyse trade flows between an ex-ante pre-selected group of countres. Snce ths study analyses the trade creaton and trade dverson effects of the EU-SA and the SADC agreements wth partcular focus on trade between South Afrca and ts maor tradng partners over the perod 1994 to 2008, the fxed effects specfcaton of the gravty model of blateral trade s used. Followng Martnez-Zarzoso and Nowak- Lehmann (2003: ), tme-nvarant varables n the gravty model (such as dstance) cannot be drectly estmated wth a fxed effects model as these varables would be elmnated n the estmaton process. Such tme-nvarant varables can be estmated by runnng a separate regresson wth the tme-

9 SAJEMS NS 14 (2011) No nvarant varables as the ndependent varables and the ndvdual fxed effects as the dependent varables. The gravty model of blateral trade n ths study s therefore estmated n two stages as n Martnez- Zarzoso and Nowak-Lehmann (2003: ). Stage one: Country-par fxed effects model x 1 y 2 y 3n 4n 5 er u where represents the ndvdual fxed effects. Stage two: Pooled model 0 1D 2PTA 3 PTA where represents the ndvdual fxed effects, D s the dstance between countres and, PTA s a dummy varable capturng ntra bloc trade, PTA s a dummy varable capturng extra bloc trade and s the whtenose error. Data: The data set n ths study comprses 585 observatons whch nclude 15 annual observatons ( ) for 39 countres representng South Afrca s maor tradng partners. An exhaustve lst of countres formng part of the sample s found n the Appendx. Table 4 (see Appendx) gves a descrpton of varables used n ths study as well as ther sources. Panel unt root tests are conducted for each varable n the fxed effects model to ascertan that all varables are statonary before the actual estmaton can be carred out. Results of panel unt root tests n Table 5 (see Appendx) report that all varables are statonary. 5 Results Results for the frst-stage estmaton are reported n Table 2. Both GDP measures (for South Afrca and the tradng partner) are postve and statstcally sgnfcant as per a pror expectaton. South Afrca s populaton has a sgnfcant negatve effect on exports. Martnez-Zarzoso and Nowak-Lehmann (2003:296) and Armstrong (2007:5) argue that a negatve relatonshp between exports and populaton s an ndcaton of an absorpton effect. Ths means that a country wth a bg populaton would ndcate that the domestc market s large enough to absorb a consderable share of domestcally produced goods and thereby reducng the amount of domestcally produced goods that could be exported. In ths case, ncreases n the populaton sze would result n lower exports. However, ths argument would only seem to be relevant for countres where poverty s not a bg ssue. In the case of South Afrca, a country wth one of the hghest Gn coeffcents n the world, soco-economc factors such as crme and llteracy could be an ndrect cause of lower exports. For nstance, hgher populaton would relatvely translate n a hgher proporton of poverty. As a matter of fact, crme and poverty are closely related (Ludwg, Duncan & Hrshfeld, 2001). In turn, a hgher ncdence n the crme rate could negatvely affect nvestment and productvty n an economy and thereby lower the export capacty. As per a pror expectaton, Table 2 also shows that a weaker rand (hgher exchange rate) enhances exports from South Afrca to ts tradng partners. Country-par fxed effects are reported n Table 6 (see Appendx). These effects ndcate the exstence of specfc factors that enhance or restrct trade between South Afrca and a partcular country. Countres wth postve fxed effects (hghlghted rows) such as the maorty of SADC countres n the sample have ndvdual specfc factors (for nstance geographc and cultural proxmty n the case of SADC countres) that enhance ther respectve trade wth South Afrca. In the same ven, countres wth negatve fxed effects such as the USA, France and Ngera have ndvdual specfc factors that constran ther respectve trade wth South Afrca (for nstance hgher per capta ncome n the case of developed countres).

10 238 SAJEMS NS 14 (2011) No 2 Table 2 Frst-stage (fxed effects model) regresson results Varable Coeffcent Std. Error t-statstc Probablty Constant term Tradng partner s GDP *** South Afrca s GDP *** Tradng partner s populaton ** South Afrca s populaton *** Rand s average real exchange rate *** Cross-secton fxed (dummy varables) Effects Specfcaton R-squared Mean dependent var Adusted R-squared S.D. dependent var S.E. of regresson Akake nfo crteron Sum squared resd Schwarz crteron Log lkelhood Hannan-Qunn crter F-statstc Durbn-Watson stat Prob(F-statstc) *** *, **, ***: statstcally sgnfcant at the 10%, 5% & 1% levels respectvely. All varables are n logs and estmatons done n E- vews. Table 3 presents the results of the second-stage estmaton. In the case of South Afrca, dstance s reported to be playng ts tradtonal role (proxy of trade costs) n a gravty model framework,.e. dstance has a small negatve bearng on exports from South Afrca to ts tradng partners. Most mportantly, the sgns of PTA dummes coeffcents are also reported n Table 3. Table 3 Second-stage (pooled model) results Varable Coeffcent Std. Error t-statstc Probablty Dstance E *** Intra SADC Extra SADC *** Intra EU-SA Extra EU-SA *** Constant term *** R-squared Mean dependent var -1.28E-10 Adusted R-squared S.D. dependent var S.E. of regresson Akake nfo crteron Sum squared resd Schwarz crteron Log lkelhood Hannan-Qunn crter F-statstc Durbn-Watson stat Prob(F-statstc) *** *, **, ***: statstcally sgnfcant at the 10%, 5% & 1% levels respectvely. Estmatons done n E-vews. Followng the nterpretaton à la Cernat (2003:9), there s sgnfcant evdence that the SADC preferental trade agreement has had a trade contracton effect. However, these results have to be taken wth cauton as the SADC preferental trade agreement had been scheduled to ntate a comprehensve mplementaton n 2008 only. In fact, countres

11 SAJEMS NS 14 (2011) No lke South Afrca, Zmbabwe and Maurtus have lberalsed ther tarffs between 2000 and 2008 whle on the other hand, Zamba, Malaw and Mozambque only effectvely oned the trade agreement n Angola and the Democratc Republc of the Congo are set to on the trade agreement at a later stage. It s mportant to note that though the SADC as a development body has been n exstence snce 1996, the SADC preferental trade agreement started beng mplemented only n The SADC attaned the status of a free trade area n January 2008 wth mports tarffs on 85 per cent of all goods beng elmnated n the ntal 12 countres. There s also an ndcaton that the SADC preferental trade agreement s yet n the process of facltatng trade (SADC, 2009). However, these results suggest that thus far, the agreement has not yet stmulated trade (both wthn and wthout the regon) for the few countres that have been sgnatores pror to Therefore, a further encompassng and nformatve analyss of the SADC preferental trade agreement trade effects can only be carred out n the future. The EU-SA preferental trade agreement s reported to have both trade creaton and trade expanson effects. However, due to the fact that the coeffcent of ntra EU-SA trade s not statstcally sgnfcant and that only the coeffcent of extra EU-SA trade s found to be statstcally sgnfcant, ths translates nto the fact that countres n the EU-SA agreement are open and trade more wth the rest of the world than a random par of countres (Ghosh & Yamark, 2004:215). Ths goes n lne wth the emprcal fndng based on export data analyss that reveals a reducton of trade volumes between South Afrca and some European countres and an ncrease n trade between South Afrca and countres n the rest of the world such as Chna and Japan between 1994 and Concluson Ths study nvestgated trade effects of the two most sgnfcant preferental trade agreements to whch South Afrca s a sgnatory (.e. the SADC and EU-SA preferental trade agreements) followng evdence of reduced trade between South Afrca and some of ts maor tradng partners n the EU-SA preferental trade agreement and ncreased trade between South Afrca and countres n the rest of the world such as Chna and Japan. Usng a country-specfc fxed effects panel data estmaton of the gravty model of blateral trade, the study reported that the SADC preferental trade agreement has had a trade contracton effect. However, ths result s to be taken wth prudence consderng the fact that the SADC preferental trade agreement s not yet fully operatonal. On the other hand, the EU-SA preferental trade agreement has been reported to be trade expandng through ncreased trade between ts sgnatores and the rest of the world. Another mportant fndng of ths study was on the effect of the South Afrcan populaton on exports. The study reported that contrary to the absorpton effect argument of Martnez- Zarzoso and Nowak-Lehmann (2003:296) and Armstrong (2007:5), soco-economc factors such as crme and llteracy could be the catalysts of the negatve relatonshp between exports and populaton sze n South Afrca. Ths argument of the soco-economc effect of populaton sze on the productvty and export capacty of an economy extends to most developng countres as they are trapped n poverty. Based on the fact that South Afrca s an appealng trade partner due to ts regonal economc mportance, and followng Kowalsk, Lattmore and Bottn (2009:8), trade polcy n South Afrca should focus on an extended multlateral lberalsaton approach as supported by the fndngs of trade expanson effects of the EU-SA preferental trade agreement. The EU-SA dd not sgnfcantly create more trade between ts members; rather, there s evdence of more trade takng place between South Afrca and non-eu members n the rest of the world. In ths partcular nstance a preferental trade agreement should be consdered as a second-best alternatve as t conveys a rsk of unnecessary dscrmnaton. As far as the SADC preferental trade agreement s concerned, ths ntatve s proftable to other SADC countres (Kowalsk, Lattmore and Bottn, 2009:48-49). Hence, n the obectve of promotng regonal economc stablty and development, South Afrca, as the

12 240 SAJEMS NS 14 (2011) No 2 regonal economc hub, has an oblgaton to support the SADC preferental trade agreement ntatve to stmulate trade and growth n the regon. Nonetheless, also confrmng the fndngs that t would be premature to assess the actual trade effects of the SADC preferental trade agreement, Kowalsk, Acknowledgement Fundng from Trade and Industral Polcy Strateges s hereby acknowledged Endnotes Lattmore and Bottn (2009:48-49) pont to the fact that the analyss of the mportance of the SADC preferental trade agreement to South Afrca s hampered because some trade flows among SADC countres are usually not reported. 1 Currency msalgnment refers n ths case to the fact that a currency becomes overvalued as a result of the level of domestc consumer prce exceedng the level of consumer prce abroad when expressed n the same currency (Engel, 2009). 2 As a result of lower demand for South Afrcan exports due to hgher prces, the levels of producton declned, causng n turn a reducton n frms profts and consequently ob losses. 3 F (1.43 ) F ( ) Reect H (Homogene ty). crt Computed Country-specfc fxed effects are vald. 0 References AMPONSAH, W.A Analytcal and emprcal evdence of trade polcy effects of regonal ntegraton: mplcatons for Afrca. [Onlne] Avalable at: [Accessed ]. ANDERSON, J.E A theoretcal foundaton for the gravty equaton. Amercan Economc Revew, 69(1): ANDERSON, J.E. & VAN WINCOOP, E Gravty wth gravtas: a soluton to the border puzzle. Amercan Economc Revew, 93(1): ARMSTRONG, S Measurng trade and trade potental: a survey. Asa Pacfc Economc Papers, No BRUN, J., CARRÈRE, C., GUILLAUMONT, P. & DE MELO, J Has dstance ded? Evdence from a panel gravty model. The World Bank Economc Revew, 19(1): CASSIM, R., ONYANGO, D. & VAN SEVENTER, D.E The state of trade polcy n South Afrca. Trade and Industral Polcy Strateges. [Onlne] Avalable at: (Accessed ]. CARRÈRE, C Revstng the effects of regonal trade agreements on trade flows wth proper specfcaton of the gravty model. European Economc Revew, 50(2006): CERNAT, L Assessng South-South regonal ntegraton: same ssues, same metrcs. Polcy Issues n Internatonal Trade and Commodtes Study Seres, No. 21. CHAUVIN, S. & GAULIER, G Regonal trade ntegraton n Southern Afrca. Centre d Etudes Prospectves et d Informatons Internatonales (CEPII), workng paper No CHENG, I. & WALL, H.J Controllng for heterogenety n gravty models of trade and ntegraton. Federal Reserve Bank of St. Lous Revew, 87(1): CLAUSING, K. A Trade creaton and trade dverson n the Canada-Unted States free trade agreement. Canadan Journal of Economcs, 34(3): COULIBALY, S On the assessment of trade creaton and trade dverson effects of developng RTAs. Unversty of Lausanne. [Onlne] Avalable at: [Accessed ]. EDWARDS, L. & LAWRENCE, R South Afrcan trade polcy matters: trade performance and trade polcy. Economcs of Transton, 16(4): EDWARDS, L Has South Afrca lberalsed ts trade? South Afrcan Journal of Economcs, 73(4): EGGER, P A note on the proper econometrc specfcaton of the gravty equaton. Economcs Letters, 66(2000): EGGER, P. & PFAFFERMAYR, M The proper panel econometrc specfcaton of the gravty equaton: a three-way model wth blateral nteracton effects. Emprcal Economcs, 28:

13 SAJEMS NS 14 (2011) No FLYNN, S.I Trade creaton and dverson. Ebsco research starters. [Onlne] Avalable at: [Accessed ]. GHOSH, S. & YAMARIK, S Does trade creaton measure up? A re-examnaton of the effects of regonal tradng arrangements. Economcs letters, 82(2004): HEAD, K Gravty for begnners. Unversty of Brtsh Columba. [Onlne] Avalable at: [Accessed ]. HOLDEN, M. & MCMILLAN, L Do free trade agreements create trade for South Afrca? Trade and Industral Polcy Strateges (TIPS). Trade and Industral Montor, No. 37. JONSSON, G. & SUBRAMANIAN, A Dynamc gans from trade: evdence from South Afrca. IMF Staff Papers, 48(1):197. KALABA, M Prncples of trade: SA and ts developed and developng partners. TradeInvestSA. [Onlne] Avalable at: [Accessed ]. KWENTUA, G.E Trade creaton and trade dverson effects n the EU-South Afrca free trade agreement. Unpublshed master s thess. Lousana State Unversty. [Onlne] Avalable at: [Accessed ]. KOWALSKI, P., LATTIMORE, R., & BOTTINI, N South Afrca s trade and growth. OECD Trade Polcy workng papers, No. 91, OECD Publshng, OECD. LUDWIG, J., DUNCAN, G. J. & HIRSHFIELD, P Urban poverty and uvenle crme: evdence from a randomzed housng moblty experment. The Quarterly Journal of Economcs, 116(2): MABUGU, R. & CHITIGA, M South Afrca trade lberalzaton and poverty n a dynamc mcrosmulaton CGE model. Unversty of Pretora, Department of Economcs workng paper, No MAPCROW Travel dstance calculator and map between world ctes, vewed 16 September 2009, MARIMOUTOU, V., PEGUIN, D. & PEGUIN-FESSOILLE, A The dstance-varyng gravty model n nternatonal economcs: Is the dstance an obstacle to trade? Economcs Bulletn, 2(2): MARTINEZ-ZARZOSO, I. & NOWAK-LEHMANN, F Augmented gravty model: an emprcal applcaton to MERCOSUR-European Unon trade flows. Journal of Appled Economcs, 6(2): MATTHEE, M. & NAUDÉ, W Export dversty and regonal growth: emprcal evdence from South Afrca. Workng papers UNU-WIDER research paper, World Insttute for Development Economc Research (UNU-WIDER). MÁTYÁS, L Proper econometrc specfcaton of the gravty model. The World Economy, 20(3): MÁTYÁS, L The gravty model: some econometrc consderatons. The World Economy, 21(3): MTONGA, E The real exchange rate of the rand and compettveness of South Afrca s trade. MPRA paper, No [Onlne] Avalable at: [Accessed ]. MUSILA, J. W The ntensty of trade creaton and trade dverson n COMESA, ECCAS, and ECOWAS: a comparatve analyss. Journal of Afrcan Economes, 14(1): OGULEDO, V.I., & MACPHEE, C.R Gravty models: a reformulaton and an applcaton to dscrmnatory trade arrangements. Appled Economcs, 26(2):107. PEET, M.A. & KOCH, S.F Techncal barrers to trade faced by South Afrcan SMME S. Unversty of Pretora, Department of Economcs workng paper. [Onlne] Avalable at: [Accessed ]. PETERSSON, L Export dversfcaton and ntra-ndustry trade n South Afrca. South Afrcan Journal of Economcs, 73(4): QUANTEC RSA nternatonal trade database. [Onlne] Avalable at: [Accessed ]. RAHMAN, M., SHADAT, W. & DAS, N.C Trade potental n SAFTA: an applcaton of augmented gravty model. Centre for Polcy Dalogue (CPD) occasonal paper seres, No. 61. RANGASAMY, R. & BLIGNAUT, B How exposed s the South Afrcan economy to nternatonal trade? South Afrcan Journal of Economcs, 73(3):

14 242 SAJEMS NS 14 (2011) No 2 RANGASAMY, L. & BRICK, K The mplcatons of OECD growth for South Afrcan exports. South Afrcan Journal of Economcs, 75(4): SARKER, R., JAYASINGHE, S Regonal trade agreements and trade n agr-food products: evdence for the European Unon from gravty modellng usng dsaggregated data. Agrcultural Economcs, (37): SMET, K Stuck n the mddle? The structure of trade between South Afrca and ts maor tradng partners. Venna Unversty of Economcs & B.A., workng paper, No SOUTH AFRICA. Department of trade and ndustry (DTI) Trade statstcs. [Onlne] Avalable at: [Accessed ]. SOUTH AFRICAN RESERVE BANK Onlne download faclty. Avalable at: [Accessed ]. SOUTHERN AFRICAN DEVELOPMENT COMMUNITY (SADC) SADC free trade area. [Onlne] Avalable at: [Accessed ]. UNITED NATIONS STATISTICS DIVISION. 2009a. Commodty trade statstcs database (COMTRADE). [Onlne] Avalable at: [24 September 2009]. UNITED NATIONS STATISTICS DIVISION. 2009b. Mllennum Development Goals. [Onlne] Avalable at: [Accessed ]. VICARD, V On trade creaton and regonal trade agreements: Does depth matter? Revew of World Economcs, 145(2): WORLD TRADE ORGANISATION. 2009a. Trade profles 2009: SA rank n nternatonal trade (2008). [Onlne] Avalable at: [Accessed ]. WORLD TRADE ORGANISATION. 2009b. Regonal trade agreements gateway. [Onlne] Avalable at: [Accessed ]. WORLD TRADE ORGANISATION. 2009c. Regonal trade agreements by WTO member. [Onlne] Avalable at: [Accessed ]. Appendx Lst of countres [2009 exports rank as n DTI (2009)] Chna - (CHINAS) Unted States - (NAFTA) Japan - (North-East Asa) Germany - (EU) Unted Kngdom - (EU) Swtzerland - (EFTA & OTHER) Netherlands - (EU) Inda - (SAARC) Mozambque - (SADC) Zmbabwe - (SADC) Zamba - (SADC) Belgum - (EU) Span - (EU) Italy - (EU) Hong Kong - (CHINAS) Kenya - (North-East Afrca) Korea Rep. South - (North-East Asa) Angola - (SADC) France - (EU) Australa - (Pacfc-Contnent) Ngera - (West Afrca) Unted Arab Emrates - (Mddle East) Dem. Rep. of Congo - (SADC) Israel - (Mddle East) Malaysa - (ASEAN) Tanzana - (SADC) Canada - (NAFTA) Malaw - (SADC) Poland - (Central Europe) Sngapore - (ASEAN) Ghana - (West Afrca) Indonesa - (ASEAN) Brazl - (MERCOSUR) Maurtus - (SADC) Saud Araba - (Mddle East) Thaland - (ASEAN) Afghanstan - (South Mddle Asa) Turkey - (Mddle East) Sweden - (EU)

15 SAJEMS NS 14 (2011) No Table 4 Data Sources Varable Defnton Source X Exports from country to country (constant 1990 USD prces) UN Statstcs (2009a) Y GDP of country (constant 1990 USD prces) UN Statstcs (2009b) Y GDP of country (constant 1990 USD prces) UN Statstcs (2009b) N Populaton of country UN Statstcs (2009b) N Populaton of country UN Statstcs (2009b) ER Real effectve exchange rate of country s currency: Average SA Reserve Bank D Dstance between captal ctes of country and country Mapcrow PTA PTA dummy (ntra PTA) WTO (2009c) PTA PTA dummy (extra PTA) WTO (2009c) = South Afrca = Tradng partner Table 5 Panel unt root tests (LLC) (IPS) ADF Fsher PP - Fsher Export (0.000)*** (0.000)*** (0.000)*** (0.000)*** South Afrca s GDP (0.000)*** (0.000)*** (0.000)*** (0.000)*** Tradng partner s GDP (0.001)*** (0.000)*** (0.000)*** (0.000)*** South Afrca s populaton (0.000)*** (0.000)*** (0.000)*** (0.011)** Tradng partner s populaton (0.000)*** (0.000)*** (0.000)*** (0.027)** *, **, ***: statstcally sgnfcant at the 10%, 5% & 1% levels respectvely. Probabltes are gven n brackets Table 6 Country-specfc fxed effects Country Chna USA Japan Germany UK Swtzerland Netherlands Inda Mozambque Zmbabwe Zamba Belgum Span