Determinants of Tanzania and Kenya Trade in the East African Community: A Gravity Model Approach

Transcription

1 Determnants of Tanzana and Kenya Trade n the East Afrcan Communty: A Gravty Model Approach Benedct K. Mahona *,Godwn D. Mema Department of Economcs, Insttute of Fnance Management, P. O. Box 3918, Dar es Salaam Tanzana * E-mal of the correspondng author: mahona.ben@gmal.com Abstract The paper ams at studyng the determnants of trade n the East Afrcan Communty (EAC). The paper explans as to why despte havng fve members, the two countres Kenya and Tanzana domnates trade among EAC members.usng the aggregated gravty model, the study fnds that, Tanzana and Kenya Trade are much determned by the economc sze (GDP) of the EAC members rather than the Per capta GDPs of these countres. The coeffcent of a dstance varable has negatve mpact meanng the costs of tradng, tme related costs and costs related to market access are hgher. In a dsntegrated model, the economc sze of the respectve countres, Kenya and Tanzana,exerts a postve mpact but Tanzana s GDP have a hgher value of coeffcent than Kenya. The Tanzana s mporters GDPs coeffcent s postve but not sgnfcant for Tanzana, whle t s postve and sgnfcant for Kenya s export Trade. Importers populaton showed a remarkable contrbuton to blateral trade between Tanzana, Kenya and the rest of EAC. However, per capta ncome coeffcent explans that Kenya and Tanzana does not trade hgh ncome orented products for the coeffcent had a negatve sgn whch s sgnfcant. The dstance, representng the cost of tradng, s affects Tanzana export negatvely more than how t does for Kenya. The exchange rate coeffcent shows that prce competton s mportant because for Tanzana lowerng her currency does not half the export as f Kenya would do. Openness varable shows trade lberalzaton, perhaps the formaton of EAC bloc;measures have sgnfcantly mproved trade flows between EAC countres. Keywords: Gravty model, export orented economy; Regonal or Free Trade Agreements; producton s specalzaton and compettveness 1. Introducton Trade between EAC member states s arguably as old as the hstory of these states. However, trade between Kenya and Tanzana whch s the focus of ths paper s of partcular nterest for one maor reason; together the volume of trade between the two states consttutes over 45%of the entre EAC trade. Furthermore, there s currently more pronounced cross border nvestments between Kenya and Tanzana than n any other EAC member states. There are ndcatons that once formalzed; labor movements across the two countres are lkely to be more predomnant than n other member states. The East Afrcan Communty (EAC) s an ntergovernmental organzaton made of fve member states namely Tanzana, Kenya, Uganda, Rwanda and Burund. Hstorcally there has been two phases n the creaton of EAC. The frst phase (referred heren as EAC I) stretches from 1967 when the Treaty whch establshed EAC I was sgned up to 1977 when the scheme collapsed amd poltcal and economc frctons. Durng ths phase EAC I had only three member states namely Tanzana, Kenya and Uganda. The second phase (.e. EAC II) was embark upon by member states after realzng the loss of economes of scale and other benefts of the defunct EAC I and hence n 1999, a Treaty for establshng the second phase of EAC was sgned by the respectve Heads of State of the member countres. One of the sgnfcant developments of EAC II has been the ncluson of two new members; Rwanda and Burund thus makng the current membershp of EAC to fve. In terms of sze, EAC has a total area of 1,817,945 km2 and an estmated populaton of mllon people. The total Gross Domestc Product (GDP) of the regon was estmated (2007) to be 61US $ bn. whch culmnates nto an average per capta ncome of over US $ 450. Tradtonally the ratonale for the creaton of regonal ntegraton schemes lke EAC was n terms of trade promoton n member states. Whle trade promoton has remaned the cornerstone for the regonal economc groupngs there s an upsurge of lterature whch had advocated for a broader developmental focus of the schemes. The hstory of economc and socal cooperaton n East Afrca s as old as the hstory of the regon. There s suffcent evdence to support the clam that the peoples of the regon have enoyed close soco-economc tes among themselves. Long before the creaton of EAC (I) the colonal powers (manly Brtan whch had a mandate over Kenya, Uganda and Tanzana after the end of the Second World War) had devsed varous approaches amed at explotng the economc potentals of the regon. In 1917 for example, the Brtsh government establshed a customs unon between Kenya and Uganda amed at promotng trade between them. Tanzana (then Tanganyka) oned the unon n later n A further attempt amed at forgng closer economc tes n the (East Afrcan) regon was made n 1961 when the East Afrcan Common Servces 13

2 Organzaton (EACSO) was created to cater to common servces ( ncludng ralway, road, postal etc) that were offered n the regon. EASCO operatons were fnally stopped n 1967 when EAC (I) was created n The obectves of EAC (I) were, nter ala, to acheve a balanced economc growth wthn the EAC member states through the establshment of a common market, common customs tarff, and common publc servces ( see for example, Krkpatrck and Watanabe, 2005 and Ng ang a, 2006). The EAC (I) lasted for a perod of ten years but collapsed n 1977 amd the exstence of what seemed to be rreconclable poltcal and economc frctons of that era. Some analysts have observed for example that whle Tanzana and Uganda were at that tme expermentng wth dfferent brand of Afrcan socalsm Kenya was pursung the captalst path. Such poltcal dfferences were further complcated by a mltary coup n 1971 whch nstalled n Uganda Dctator Id. The collapse of EAC (I) had profound negatve mpacts on the economes of Kenya, Uganda and Tanzana. Besdes losng the economes of scale advantage the regon had pror to the collapse of EAC (I) each country had to re-start a costly programme of establshng servces and ndustres whch were somewhat effcently at the EAC level.realzng the lost advantages of EAC (I) the three member states began the process of re-establshng closer tes among them. Fnally n 1999 followng a seres of consultatons and negotatons a Treaty for establshng EAC (II) was sgned n Arusha, Tanzana. Apart from the expanson of the EAC membershp the other mlestones so far regstered n EAC nclude () the establshment of the customs unon () the establshment of the common market () convertblty of Kenya, Uganda and Tanzanan (v) harmonzaton of goods produced n the regon (v) reducton of tarff and non-tarff barrers and (v) free movement of stock (EAC, 2010). 2 Trade Flows between Tanzana and Kenya Although the EAC sub-regon has wtnessed a surge n trade flows among member states partcularly after the establshment n 1 st July, 2010 of the common market.the EAC common market has provded for four freedoms namely; free movement of goods, labor, servces and captal all of whch are expected to boost trade and nvestment wthn the regon. Interest n analyzng the trade flows between Tanzana and Kenya s based on economc as well as non-economc factors. The essence of focusng on trade flows between Tanzana and Kenya s that currently the two countres domnate the economy of the sub-regon, n many aspects. Kenya and Tanzana are currently the power houses of the EAC economy. In 2010 for example, the combned GDP of the sub-regon was US $ 79.3 bllon. Kenya s GDP durng the same year was US $ 32.2 bllon equvalent to 40.6%. Tanzana s GDP was US $ 22.9 bllon (equvalent 28.9%) durng the same year. Ths mples that the combned GDP of Kenya and Tanzana was equvalent to 69.5% of the total EAC GDP. The share of the other three countres (Uganda, Rwanda and Burund) was 30.5%. Accordng to data from the World Bank the two countres domnate the regon where GDP per capta s concern. In 2010 Kenya had the hghest GDP per capta (US $ 468.7) compared to Tanzana s per capta GDP of US $ The per capta GDP n the rest of EAC member states was lower than those for Kenya and Tanzana.Trade as a percentage of GDP was the hghest n Kenya (65.4%) but was closely followed up by Tanzana whose rato n the same year was 63.8%.Recent trade trends between Tanzana and Kenya (Table 1) suggest that apart from ther relatve large GDPs, they stll (Kenya and Tanzana) domnate the regon s trade flows(note 1: Table 1: Trade Flows n EAC: (US $ m) In 2009 Tanzana s exports to the rest of the World (ROW) was US $ 2,982,405mllon compared to the country s exports to Kenya whch were valued at US $ 192,904 mllon (equvalent to 6.5%). Ths rato ncreased to 8.0% n 2010 before levelng off to 4.7% n Tanzana s exports to Kenya durng the perod averaged US $ 246,368 mllon compared to the value of Tanzana s exports to ROW whch totaled US $ 3,922,637 mllon.kenya s an mportant destnaton for Tanzana s exports than all the other EAC countres combned(note 2: Table 2: Economc Growth n EAC Regon ) 3 Modelng the Determnants of Trade Flows between Kenya and Tanzana The gravty model used n nternatonal trade analyss draws ts theoretcal foundatons from gravty modelsused n natural scences. The model predcts blateral trade flowsbased on the economc szes of countres (usng GDP) and dstance between them. The model was frst used by Tnbergen n 1962 and pyhnen (1963) who postulated that trade between two countres ( and ) takes the form of: Where F s the trade flow, M are the economc masses of each country, D s the dstance between them and G s a constant. Besdes beng deployed to analyze trade flows between countres the model has also been used n nternatonal relatons ncludng evaluatng the mpact of treates and allances on trade. 14

3 Snce then (1963), the gravty model has become a popular nstrument n emprcal foregn trade analyss such that dfferent scholar has successfully appled the model to show flows of varyng types such mgraton, foregn drect nvestment and more specfcally to nternatonal trade flows. Accordng to ths model, exports from country to country are explaned by ther economc szes (GDP or GNP), ther populatons, drect geographcal dstances and a set of dummes ncorporatng some knd of nsttutonal characterstcs common to specfc flows. The followng are the theoretcal development whch has appeared n support of the gravty model snce the second half of 1970s. They show the theoretcal support of the research n ths feld whch was orgnally very poor. Anderson (1979) made the frst formal attempt to derve the gravty equaton from a model that assumed product dfferentaton. Bergstrand (1985, 1989) also explored the theoretcal determnaton of blateral trade n a seres of papers n whch gravty equatons were assocated wth smple monopolstc competton models. Helpman and Krugman (1985) used a dfferentated product framework wth ncreasng returns to scale to ustfy the gravty model. Deardorff (1995) has proven that the gravty equaton characterzes many models and can be ustfed from standard trade theores. Fnally, Anderson and Wncoop (2001) derved an operatonal gravty model based on the manpulaton of the CES expendture system that can be easly estmated and helps to solve the so-called border puzzle. The dfferences n these theores help to explan the varous specfcatons and some dversty n the results of the emprcal applcatons. There are a huge number of emprcal applcatons n the lterature of nternatonal trade, whch have contrbuted to the mprovement of performance of the gravty equaton. Some of them are closer related to our work. Frstly, Mátyás (1997) and (1998), Chen and Wall (1999), Breuss and Egger (1999) and Egger (2000) mproved the econometrc specfcaton of the gravty equaton. Second, Berstrand (1985), Helpman (1987), We, (1996), Soloaga and Wnters (1999), Lmao and Venables (1999), and Bougheas et al, (1999) among others, contrbuted to the refnement of the explanatory varables consdered n the analyss and to the addton of new varables. In other occasons the model has been used to test the effectveness of trade agreements and organzatons such as the North Amercan Free Trade Agreement (NAFTA) and the World Trade Organzaton (see for example Helman, 2003, Panagarya, 1999, and Krugman 2001 among others). In other occasons the model has been used to test the effectveness of trade agreements and organzatons such as the North Amercan Free Trade Agreement (NAFTA) and the World Trade Organzaton (Helman, 2003, Panagarya, 1999, and Krugman 2001 among others). Whle the model s basc form conssts of factors that have more to do wth geography andspatalty, the gravty model has been used to test hypotheses rooted n purer economc theores of trade as well. One such theory predcts that trade wll be based on relatve factor abundances. One of the common relatve factor abundance models s theheckscher-ohln model. Ths theory would predct that trade patterns would be based on relatve factor abundance. Those countres wth a relatve abundance of one factor would be expected to produce goods that requre a relatvely large amount of that factor n ther producton. Whle a generally accepted theory of trade, comparatve advantage has suffered emprcal problems. Investgatons nto real world tradng patterns have produced a number of results that do not match the expectatons of comparatve advantage theores. Notably, a study by Wassly Leontef who found that the Unted States, the most captal endowed country n the world, actually exports more n labor ntensve ndustres. Comparatve advantage n factor endowments would suggest the opposte would occur. Other theores of trade and explanatons for ths relatonshp were proposed n order to explan the dscrepancy between Leontef s emprcal fndngs and economc theory. The problem has become known as the Leontef paradox. The paper frst uses a smplfed verson of the gravty model to analyze the determnants of trade between Kenya, Tanzana wth the rest of EAC countres respectvely, and whether the creaton of the EAC has beneftted them. Ths smplfed verson of the gravty model examnes the determnants manly GDP as the economc sze, per capta ncome, and dstance. Essentally, the gravty model traces geographcal-spatal relatonshp of the foregn trade between Kenya and Tanzana n the EAC bloc. Used n the economcs sense, the gravty model stpulates that the economc sze of countres and geographcal dstance are the two basc factors determnng the blateral trade flows between Kenya and Tanzana. The stochastc form of the equaton to estmate the determnants of trade of trade flow between Kenya and Tanzana s gven as: o TD = Y 1 Y ( Y / P) ( Y / P ) D A µ Where: 5 Blateral trade volume (=export mport) between country and. and J=1,2...N GDPs of Kenya or Tanzana wth the rest of EAC countres (1) 15

4 ( Y / P) and ( Y / P) = product of country s and country s per capta GDPs where P means populaton = The dstance between the captal ctes of Kenya or Tanzana wth the rest of EAC Countres Other factors (ncludng cultural) nfluencng trade between block countres, n ths study we use factors lke terms of trade, openness, and nflaton Error term The basc, standardzed emprcal gravty equaton takes the followng form: LnTD = Ln Ln( P * P ) µ 0 1 ( Y * Y ) 2Ln[( Y / P*) *( Y / P*) ] 3LnD 4... (2) Prevous studes made a use of ths emprcal equaton to dentfy the blateral trade volume of country-pars n NxN countres settng. Usng the same equaton, ths paper tests how sgnfcantly the gravty model s applcable to explan Tanzana s or Kenya s blateral trade flows n the EAC and tres to extract mplcaton for Tanzana s and Kenya s trade polcy. The equaton s applcaton frst fxes Tanzana as, but leavng the rest of EAC as =1,2...N n a Nx1 settng, and then t does so for Kenya. In ths knd of equaton, all varables are n natural logarthm of real value terms expect dummy varables. Any varable wth relatvely small numbers are usually exempted from takng the logarthm (nflaton, terms of trade and openness for ths study). In partcular, snce the EAC countres borders to each other, use common language, have colonal relatonshp or hstorcal tes the dummy varable s not that much relevant n ths model. Hence, the model starts wth only three explanatory varables, namely the product of GDPs, the product of per capta GDPs and dstance. The explanatory varables, the product of GDPs serves as a proxy for the two countres economc sze, both n terms of producton capacty and sze of market. Larger countres, wth great producton capacty, are more lkely to acheve economes of scale and ncrease ther export based on comparatve advantage. They also possess large domestc market able to absorbed more mports. Therefore an ncrease n the product of the two Countres GDPs s expected to ncrease blateral trade volume. Thus we expect the estmated coeffcent of producton capacty to be postve. Populaton has smlar explanatons but when the coeffcent s postve, then the country has economes of scale whle a negatve sgn mples when t s bgger t exports less. Per capta GDP 1 s an explanatory varable that serves as a proxy for the ncome leve and/or purchasng power of the exportng and mportng countres. Snce, Tanzana or Kenya per capta s fxed, ths varable wll serve to predct whether Tanzana s or Kenya s trade flows are depend on ts tradng partner s ncome level. The dstance varable a trade resstance factor that represents trade barrer such transport costs, tme, cultural unfamlarty and market access barrers. Prevous lteratures nterpret the coeffcent of dstance varable as elastcty wth respect to an absolute level of geographcal dstance. In the gravty model, trade volume wll be larger between countres pars that are far from the rest of the world than between countres pars that are close to the rest of the world (Anderson and Van Wncoop 2003; Harrgan 2003) Analyss of the Home-Market Effect As explaned above, the explanatory varables, the product of GDP serves as a proxy for the two countres economy sze, both n terms of producton capacty and sze of the market. The per capta GDP, however, t serves as a proxy for the ncome level and/ or purchasng power of the exportng and mportng countres. When the ft of the gravty model for Tanzana was run, the results turned out that, Tanzana s trade depend much on GDP for ts blateral trade flow and t s negatvely related to the per capta patterns of trade. To capture the home-market effect, as Davd and Wensten (2003) showed the evdence of the broad homemarket effects, the model n equaton (2) s dsntegrated so as to make use of Y and Y as n separate form, wth T representng ether export of mport of the country. the model allow the coeffcents of the two terms possbly dfferent each other. Ths mples that there s possblty of sgnfcant home-market effect, meanng the advantage of a large home market as a foundaton of export of a good. However, tradtonal neoclasscal model of comparatve advantage suggest that, all else equal, a country wth extraordnarly strong demand for a good wll be an mporter of that good mplyng a reverse home market effect. Increasng returns leads to a home-market effect n dfferentated goods, whereas n homogeneous goods a gravty equaton wll apply, but 1 Bergstrand (1989), combnng economc geography and factor proportonal theory, derved the gravty equaton at the ndustry level whch predcts that the export of a good n blateral trade depend on ncome and per captal ncome as well, assumng a constant elastcty of transformaton of supples among dfferent markets. 2 A decrease of the dstance coeffcent ndcates that trade wth far away countres ncreases relatve to the trade wth closer countres, whereas an ncrease represents trade wth closer countres ncreases faster than that wth far way countres. The noton of relatve dstance remans sgnfcance at the NxN countres settng, whereas t s mportant drops sharply n our Nx1 because all dstances are measured absolutely from Tanzana or Kenya, we antcpate the coeffcent greater than one, but ts magntude matters. 16

5 wthout home market effects due to barrers to entry or natonal product dfferentaton. Lnearzng equaton (1) by takng natural log of all varables results nto equaton (3) as follows: LnTD = LnY LnY Ln Y loga µ ( / P) 4Ln( Y/ P) 5logD 6 (3) The current paper wll deploy a stochastc form of the model (equaton 3) to estmate the determnants of the trade flows between Tanzana and the rest of EAC Countres or Kenya and the rest of EAC Countres as follows:. LnEXP LnEXR µ = 0 1LnY 2LnY 3LnP ( ) 4LnP ( ) 5LnY ( / P) 6LnY ( / P) 7LnD 8 (4) Where: Y and Y = GDPs of Kenya or Tanzana wth the rest of EAC Countres, respectvely D= Geographcal dstance between the captal of Kenyaor Tanzana wth the rest of EAC Countres EXP = exports of Kenya or Tanzana to the rest of EAC countres EXR = blateral exchanger rate of ether Kenya or Tanzana wth rest of EAC countres, whch s defned as local currency value of 1 unt of country currency. (Y/P) = per capta ncome of Tanzana or Kenya and the rest of EAC countres. 4Emprcal Analyss 4.1. Estmatng the aggregated trade volume of Tanzana s and Kenya s blateral trade Table (3) below represents OLS results of Tanzana s and Kenya s trade wth the rest of EAC countres (Kenya s ncluded for Tanzana blateral trade and Tanzana vsa vsa); the overall performance of the model seems to be pretty good wth R-square value around for the basc equaton for Tanzana s trade and for Kenya s blateral trade. Most of the explanatory varables are hghly sgnfcant, meanng the gravty model s effectve n explanng Tanzana s blateral trade flows and that the gravty model s well applcable to a sngle country case. The log of the products of two countres GDPs, that s and, s hghly sgnfcant n determnng both Tanzana s and Kenya s blateral trade volume. The estmates coeffcent 1 s postve and about The result obtaned s consstent wth the basc hypothess of the gravty model that the trade volume wll ncrease wth an ncrease n economc szes. The varable also means that holdng constant for other varables; a 1 percent pont ncrease n GDP wll result n about percent pont ncrease n Tanzana blateral trade flows. Theoretcally t s a bt surprsng of fnd the coeffcent of the product of GDPs s greater than one, though the hgher value of the estmates mples a Heckscher-Ohln type 3 basng on nter-ndustry trade (Deardorff, 1998; Grossman 1998 p.30). Kenya s trade s a bt dfferent from that of Tanzana. The coeffcent of the Product of GDPs s postve but less to one, the value s estmated to be 0.71, meanng that 1 percentage ncrease of GDPs wll only lead to 0.7 percent ncrease n the blateral trade volume(note 3:Table 3: OLS results of Tanzana s and Kenya s blateral trade wth the rest of EAC countres There dfferent reasons as to why the ncrease n blateral trade volume s less proportonate to the ncrease n GDP. Three possble sources are stpulated out; the frst one s the exstence of relatvely large market homemarket effect (trfler 1995p. 1032; McCallum 4, 1995) as such there s possblty of a home market effect, meanng a smaller trade than the theoretcal predcton 5. Secondly, the presence of lower level of ntra-ndustry trade, where by the volume of trade s hgher n sectors characterzed by a monopolstc competton and/or scale economes (Harrgan, 2003). Thus a country enoyng a lesser scale economes wll trade a smaller volume. The thrd s the extent of trade barrers; the hgher and wder are the trade barrers the smaller wll be the trade volume. When the three reasons or source of the products of GDPs to be less than one are taken nto account, the hgher level of Tanzana home-market effect, the presence of nter-ndustry trade and lower trade barrers among EAC members account to the hgher value of the coeffcent of the product of GDPs 6. For the case of Kenya s trade, the presence of home- based market, lower level of ntra- ndustry trade, and the possblty of tradng away ther products as far as the EAC market s concerned. The per capta GDP varable s a sgnfcant factor n determnng Tanzana s and Kenya s blateral trade n the EAC bloc. However, t s negatvely related to the blateral trade flows. The estmated coeffcent ß 2 has a 3 Frankel (1998) showed the coeffcent lyng n the range of Our estmate more or less ft the range but n a hgher level and lower level for Kenya s trade. 4 Home-based market effect, such as localzed taste of local dstrbuton network, plays a greater role n trade. 5 Ths s true for a small market, theoretcally the home-market effect can be nterested as an elastcty of export wth respect to domestc ncome that exceeds the mportng country s ncome elastcty of export and mport for a country, thus engenderng the gravty analyss utlzng Y1and Y2 as a separate terms more approprate. 6 The results calls for a test of home market effect n the blateral trade of Tanzana s and Kenya strade 17

6 magntude of and predctng that a 1 percent ncrease n per capta GDP leads to about 2 and 0.5 percent decrease n blateral trade flow of Tanzana and Kenya, respectvely. The emprcal result s dfferent from the regresson analyses of Frankel (1997) that predcted that a 1% ncrease n per capta GDP leads to about 0.1% ncrease n blateral trade flows. The fndngs mples that Tanzana trade patterns follow a GDP pattern rather than a per capta pattern, relyng more on ts tradng partner s overall economc sze than ts ncome level 7. Therefore, Tanzana trade can be characterzed as the one whch depend more on exportng of quantty-based standardzed products that are senstve to the overall market sze, rather than exportng qualty-based hgh value-added products that a senstve to the tradng partner s ncome level. The estmaton of the dstance s sgnfcant wth the expected negatve sgn. The results show that geographcal dstance s an mportant factor for Tanzana s trade. The coeffcent ß 3 s estmated to be and for Tanzana s and Kenya s blateral trade, respectvely. The results are hgher compared to prevous studes (Frankel 1997; Wall 1999; Buch et al. 2003), but n the lne of Grossman (1998) who ponted out that most of emprcal gravty studes show a surprsng larger sze of the estmated coeffcent on the dstance varable. Furthermore, the coeffcent mples hgher cost of tradng n the EAC bloc because t reflects transport costs, tme and market access barrers 8, and that the data used are of nearby countres because a decrease n of the dstance coeffcent ndcates that trade wth far away countres ncrease relatve to the trade wth closer countres, whereas an ncrease represents trade wth closer countres ncrease faster than that wth far countres. However, Tanzana trade blateral shows are nearby concentrated trade of hgher cost than that of Kenya. 4.2 Estmatng the blateral export of Kenya and Tanzana wth the Rest of EAC countres The specfcaton used when the gravty model s appled to estmate blateral export for specfc product s the one explaned n equaton (3). A hgh level of ncome n the exportng country ndcates a hgh level of producton, whch ncreases the avalablty of goods for exports. Therefore, t s expected that 1 to be postve. Y The coeffcent of, 2, s also expected to be postve snce a hgh level of ncome n the mportng country suggests hgher mports. In case of populaton of the exporters, 3, may be ether negatve or postve sgned, dependng on whether the country exports less when t s bg (absorpton effect) or whether a bg country exports more than a small country (economes of scale). The coeffcent of the mporters populaton 4, has also ambguous sgn, for smlar reasons. As usual, the dstance coeffcent s expected to be negatve snce t s a proxy of all possble trade cost sources 9. The model also ncorporates dfferences n ncomes between exporters usng a varable of per capta ncome. And fnally, an exchange rate and openness varables are added to ths specfcaton.we estmate equaton 3 for export trade flows usng OLS method (Note 4: Table4: Random effect Regresson results of Tanzana s Export trade wth the rest of EAC (Kenya ncluded) countres and Note 5: Table 5: Random Effect Regresson results of Kenya s Export trade wth the rest of EAC (Tanzana ncluded) countres) The two tables above reports the nclusve of the ndvdual effects of varables. random or fxed effect can be used for analyss. We apply the random effect because t s approprate when estmatng typcal trade flows between randomly drawn samples of tradng partners from a larger populaton. On the other hand, the fxed effect model would be a better choce than random effect when one s nterested n estmatngtypcal trade flows between an ex ant predetermned selecton of natons (Egger, 2000). The only problem faced wth FEM s that we cannot drectly estmate varables that do not change over tme because the nherent transformaton wpes out such varables. We estmate equaton (4) nto categores startng wth the standard model to ncludng relatve prces and exchange rates. Table (4) and (5) carres the followng explanatons for the trade between Kenya and Tanzana respectvely. Colum 1 on both tables explans that export volume of Tanzana and Kenya has a postve relatonshp wth country s GDPs. The coeffcent 1 s postve for both countres, the value for Tanzana blateral trade s estmated to be 3.738, mples that a 1 percentage ncrease of Tanzana s GDP wll lead to about 3.74 percent ncrease n export volume. The coeffcent further, explan the home- market effect for Tanzana, and snce Tanzana s assumed to produce prmary products, the hgher coeffcent shows the lnk between a country s 7 The mpact s a bt less when the blateral trade for Kenya s consdered showng the lght of technology or dfferentated product advancement 8 Butch et al, argue that changes n dstance coeffcent do not carry much nformaton on changes n dstance costs over tme. Change n dstance costs are to a large extent pcked up solely n the constant term of the gravty models. 9 Recently Bougheas et al. (1999) showed that transport cost are a functon of not only of dstance but also of publc nfrastructure so some lterature augmented the gravty model by ntroducng addtonal nfrastructure varable ( stock of publc captal and length of motorway network. 18

7 market sze and ts export that does not exst n trade model that are based on solely comparatve advantage 10. However, Kenya s export quantty has a slghtly low value of 1, the estmate value of the coeffcent s 1.07, mples that a 1 percent ncrease of GDP leads to only 1 percent ncrease n the export quantty trade. When comes to trade partners GDPs, export quantty from Tanzana seems not to be favored by the economc development of her partner countres, 2 (0.153), the coeffcent of mporters GDP s postvely related to Tanzana s export quantty but not sgnfcant, whle the coeffcent for Kenya s trade, 2 (0.414), postve and hghly sgnfcant. Ths mples that for a 1 percent ncrease of economc development for Tanzana s trade partner countres leads to only 0.15 percent ncrease n her export trade, Kenya n the other hand enoys a 0.4 percent ncrease n her export trade. In addton, the proxy of market sze s further nvestgated by consderng the populaton of exporter and mporters. Exporter s populaton dd not show any remarkable output more than nurng the model and for that case were elmnated. The mporters populaton however, showed a sgnfcant and remarkable contrbuton to blateral trade of these two countres. In case of Tanzana s export quantty or trade, the varables showed a postve relatonshp, the coeffcent estmator was as compare to a negatve mpact to Kenya s export quantty. Ths means that a 1 percent ncrease n the populaton of Tanzana s trade partner would yeld a 1.88 percent ncrease n her export, whle t would yeld a 0.32 percent ncrease n Kenya s export. The output further, explan the nterdependency between the two countres (column 2); Tanzana s trade hghly depend on the larger sze (populaton s coeffcent s 1.88) of her tradng partners and not the economc development (the coeffcent of GDP for partner country s 0.87), but Kenya s trade enoy both the economc development (GDP s coeffcent s 0.23), and the populaton (coeffcent s estmate s 0.32). The per capta ncome (column 3) explans that both countres do not attract ncome level orented export of goods that they export. The coeffcent of the mporters per capta ncome s estmated to be for Tanzana s export and for Kenya s export. The regresson results mples that a 10 percent ncrease n the per capta ncome would lead to 18 percent decrease n Tanzana blateral export and only 3 percent decrease nto Kenya s blateral export. The polcy mplcaton of ths result s that snce Tanzana s n the era of competton t need to mprove more n producton sector though encouragng more value-added products and dfferentated f not sophstcated producton. Not only that but also the regresson results at ths column emphaszes the mportance of economc growth n lne wth the per capta ncrease. Comparng the mporters GDP s coeffcent s obvously that the per capta ncome ncrease facltates the sgnfcance of economc development. Tanzana s blateral export now s able to leap 1 percent ncrease when the GDP of her mporter ncrease by 1 percent and Kenya s able to leap 0.5 percent ncrease n her blateral export when the mporters GDP of her product ncreases by 1 percent. The mpact of dstance n blateral trade s hghly estmated to the model s expectatons, havng the negatve sgn but also t reflects the probty that trade s done wthn nearby countres. The only dfference between the two countres s that, Tanzana s exported quantty s negatvely affected hgher by the dstance than that how Kenya does. For nstance, a 1 percent decrease of trade costs, would lead to a 3 percent ncrease n Tanzana s export as compared to only 2.6 percent ncrease n Kenya s export 11. Our result s concomtant to the realty when the two countres are compared n terms of technology and nfrastructure development. In other words, dong busness n Tanzana s more expensve as compared to dong busness n Kenya (World Bank, 2012) 12. The coeffcents of dstance are hgher and at least constant n almost all the regresson columns expect when exchange rate varable s consdered. The dstance coeffcent s estmate to be for Tanzana s blateral export trade and only 2.23 percent for Kenya s blateral export. The man mplcaton of the dstance s coeffcent shows the value of money s vtal when consderng trade bloc. For Tanzana s export trade, the EAC bloc s more or less underutlzed snce currency dfference ncreases the trade barrers such that 1 percent ncrease n trade barrers has the mpact of 7 percent decrease n Tanzana s export trade and 2.23 percent n n Kenya s blateral export. Ths s a massage for polcymakers that havng a sngle currency would facltate trade flow ncrease n the bloc Column 4 of table (4) and (5) reports the regresson results when movement n exchange rate s consdered. The estmated coeffcent for exchange s postve for Tanzana s and negatve for Kenya s blateral export. The result sgnfes that, prce competton s mportant. Export quantty and value for Tanzana has beng favored by 10 In Krugman (1980), the demand for ndvdual goods vares across markets because of dfferences n consumer preferences (e.g., German consumers prefer beer, French consumers prefer wne), leadng producton of a good to concentrate n markets wth hgh levels of demand. 11 Recently Bougheas et al. (1999) showed that transport cost are a functon of not only of dstance but also of publc nfrastructure so some lterature augmented the gravty model by ntroducng addtonal nfrastructure varable ( stock of publc captal and length of motorway network. 12 Dong busness (World Bank, 2012), Tanzana s ranked 127 whle Kenya s ranked 109 among 187 countres 19

8 her lower exchange rate, but that of Kenya s beng half by the hgher exchange rate value of Kenya s currency. For Tanzana s trade, deprecaton (devaluaton) of exporter currency by 10 percent rases export by 7 percent, but 18 percent for Kenya s blateral export. Prce competton also adds the trade barrers to trade. For nstance a there s a remarkable estmated value of dstance s when exchange rate s regressed wth the basc varables.. The nterpretaton of the coeffcent of the openness varable s also relevant to our study. Ths s tested so as to explan the nfluence of Tanzana s and Kenya s trade and openness (defned as total exports plus total mports dvded by GDP) on the trade flows to ther partner countres. The results ndcated that the estmated coeffcent on the openness ndex s 0.92 for Tanzana s blateral trade and 1 for Kenya s trade. The results are robust wth the present specfcaton of the gravty equaton. Ths result mples that when Tanzana tarff and polcy lberalzaton change by 1% the country s export tend to ncrease by 0.9%, and by 1% for Tanzana and Kenya, respectvely. Ths suggests further that there s some correlaton between Tanzana tarff lberalzaton n EAC and her export performance. The fndng s n lne wth the standard nternatonal trade theores. In the presence of compensatory and complmentary polces, openness to nternatonal trade tends to accelerates development of poor countres (Dollar and Kraay, 2000).on the other hand, wth the same varables, column 5 t shows that openness over rules the mpact of dstance such that dstance coeffcent decreases to 1.18 for Tanzana and for Kenya, mples that wth the sandwch of openness dstance (trade costs) are reduced and the varable shows a postve mpact on trade flow. For Tanzana a 1 percent decrease of trade barrer would lead to 1.2 percent ncrease n blateral export whle t would lead to only 0.39 percent decrease of Kenya s blateral trade, other factors beng equal. Although hgh openness ndex s wdely accepted ndcator of the degree of trade lberalzaton, t s certanly not a suffcent ndcator of the welfare part of trade polcy reform. The World Development Indcator (2006), for example, show that Sub Saharan Afrca as a regon managed to ncrease ts openness ndex from 56% n 1975 to 66% n Chna on the other hand ncreased ts openness ndex from ust 9% n 1975 to 65% n Whle Sub Saharan Afrca hasn t managed to eradcate obect poverty, Chna has succeeded to do so. Ths fndng has mportant polcy messages for Tanzana and Kenya and EAC country members. It calls for polcy makers to focus more on the fundamentals of economc growth, nvestment, macroeconomc stablty, human resources and good governance as a strategy for addressng poverty. Flexblty s hghly needed whle pursung ths polcy. Snce complete openness mght expose a country to greater rsk from external shocks. Poor country lke Tanzana may fnd t hard to buffer these shocks and to bear the costs they ncur. 5.Conclusons and Recommendaton The purpose of ths paper s to contrbute to the on-gong debate on the determnants of trade between natons whch began durng the mercantlst era and was later shaped by the ntroducton of absolute and comparatve cost advantage theores. The Heckscher Ohln and later the Lnder hypotheses and other modern trade theores also contrbuted sgnfcantly to the understandng of trade flows between natons. Dfferent forms of the gravty model were tested n the current paper.the results obtaned generally support the basc hypothess of the gravty model that the economc sze of the respectve countresexerts a postve mpact on the trade between natons. Ths result s hardly surprsng. GDP s a reflecton of not only the producton capacty of a country but also the consumpton and export ablty of countres. Thus the results ndcate that the commodtes produced n both Kenya and Tanzana not only sustans consumpton levels n the two countres but also s exported to markets n the respectve EAC countres. The results obtaned wth respect to the combned per capta ncome n Kenya s and Tanzana s trade are unexpected. The mpact of ths varable was supposed to mrror that of combne GDPs. Instead, the result shows ths varable has a negatve mpact on the trade flows of both Tanzana and Kenya when tradng wth the rest of EAC countres. One possble explanaton for ths result s that Tanzana tends to depend more on exportng quantty products that satsfy the broad EAC market nstead of producng goods that are senstve to the EAC partners ncome levels.nonetheless, Kenya s products are senstve to EAC partners ncome level despte ts negatve sgn n the coeffcent of per capta ncome. The dstance varable was expected to have a negatve mpact on trade flows between Kenya s and Tanzana s Tradewth the rest of the EAC countres mples among other thngs the exstence of hgh costs of tradng between countres n the EAC. The costs of tradng refereed to here nclude transport costs (freght etc.), tme related costs and costs related to market access. Usng the dsntegrated trade volume by takng theblateral export trade of Tanzana and Kenyarevles that both countres depend has ther GDPs (exporters GDP meanng growth of ther Economy), and mporters GDPs too and also the populatons of the partner countres. The dstance shows that tradng n the EAC bloc s assocated wth hgh tradng costs. And for Tanzana more export s due to her low exchange rate whle hgh exchange rate for Kenya, strong currency, reduces t export to the rest of EAC countres. Openness varable shows trade lberalzaton measures have sgnfcantly mproved trade flows between EAC countres. However, the 20

9 dfferences seen among the due causes of Tanzana s and Kenya s trade flow n the EAC bloc concdes wth recent trade theory of factor dfferences. Due to ths factor abundance theory, or the Heckscher-Ohln (H-O) model, t s predcted that a country wll export commodtes that are relatvely ntensve n the factor wth whch the country s relatvely well endowed. The results obtaned heren have polcy mplcatons for both Kenya and Tanzana. They pont to the mportance of the two countres focusng more on polces amed at promotng economc growth wth a vew to promotng trade flows between ther countres and more sgnfcantly wthn the EAC flows. Furthermore there s need to encourage the ndustres whch cater specfcally for the needs of over 130 mllon people the regon s a ready market for products from such ndustres. Efforts need to be made especally on measures amed at reducng trade related costs n EACbloc. The countres have to come up wth concrete measures amed at reducng tarff and non-tarff restrctons to trade whle ncreasng nvestments. There s also, a need to undertake further trade lberaton of ther economes countres wth a vew to promotng trade actvtes wthn the regon.the nterest n analyzng the trade flows between the two countres has come as a result of the observaton that the two countres can loosely be referred as the economc powerhouses wthn the auspces of the EAC. The combned GDPs of Kenya and Tanzana consttute over 60% of the total EAC GDP. Lkewse there are sgnfcant recorded and unrecorded trade actvtes between Kenya and Tanzana than n other EAC countres. The mportance of the two countres n nfluencng not only trade actvtes but also economc development of the regon s lkely to ncrease wth the ntensfcaton of the EAC ntegraton especally towards the formaton of the common market and monetary unon. That s why a better understandng of the trade flows between Kenya and Tanzana s essental n enablng the EAC countres to formulate approprate macroeconomc polces for the development the regon. References Anderson, J.E..(1979), A theoretcal foundaton for the gravty equaton, Amercan Economc Revew, 69: Anderson J.E. and E. van Wncoop (2001), Gravty wth Gravtas: A Solutonto the Border Puzzle, NBER Workng Paper Anderson, J., van Wncoop, E. "Gravty wth Gravtas: A Soluton to the Border Puzzle." AmercanEconomc Revew (2003). Baer, JH Bergstrand (2009)."Bonus Vetus OLS:"A Smple Method for Approxmatng Trade-Cost Effects Usng the Gravty Equaton." Journal of Internatonal Economcs (2009) Bergstrand, J.H. (1985), TheGravty Equaton n Internatonal Trade: SomeMcroeconomc Foundatons and Emprcal Evdence. The Revew ofeconomcs and Statstcs 71: Bergstrand,J.H, (1989), The Generalzed Gravty Equaton, MonopolstcCompetton, and the Factor- Proportons Theory n Internatonal Trade, The Revew of Economcs and Statstcs 67: Bougheas S. et al. (1999), Infrastructure, Transport Costs and Trade, Journalof Internatonal Economcs 47: Breuss, F., and Egger, P. (1999), How Relable Are Estmatons of East-West Trade Potentals Based on Cross- Secton Gravty Analyses? Emprca 26 (2): Caruso, R. "The mpact of Internatonal Economc Sanctons on Trade, An Emprcal Analyss,"Peace Economcs, Peace Scence and Publc Polcy, vol.9, no.2 (2003). Chen, I-H., and H.J. Wall (1999), Controllng for Heterogenety n GravtyModels of Trade, Federal Reserve Bank of St. Lous Workng Paper A Deardorff, A.V. (1995), Determnants of Blateral Trade: Does Gravty Workn a Neo-Classc World? NBER Workng Paper 5377 Deardorff, A.V. (1998) Determnants of Blateral Trade: Does Gravty Work n a Neoclasscal World? In the Regonalzaton of the World Economy, edted by J.A. Frankel. Chcago: Unversty of Chcago Press. Egger, P. (2000), A Note on the Proper Econometrc Specfcaton of thegravty Equaton, Economcs Letters 66: Feenstra, R.C et al (2001). Usng the Gravty Equaton to Dfferentate among Alternatve Theores of Trade. The Canadan Journal of Economcs, Vol. 34, No. 2., pp. 431 Frankel, Jeffery A. (1997). Regonal Tradng Blocs: In the World Economc System. Washngton, DC: Insttute of Internatonal Economcs. October Hacker, R Scott, Enarsson, Henrk. The Pattern, Pull, and Potental of Baltc Sea Trade. The Annals of Regonal Scence, Vol. 37, No. 1, pp Helpman, E. and Krugman, P. (2003) Market Structure and Foregn Trade: Increasng Return, Imperfect Competton and Internatonal Economy,The MIT Press, Cambrdge, MA/London. Isard, W.,(1954) "Locaton Theory and Trade Theory: Short-Run Analyss". Quarterly Journal of Economcs, vol. 21

10 68, p Krugman, P and Obstfeld, E. (2003) Internatonal Economc Theory and Polcy, 6 th edton, Lmao, N., and A.J. Venables, (1999), Infrastructure, GeographcalDsadvantage and Transport Costs, Polcy Research Workng Paper 2257,World Bank. Mátyás, L. (1997) Proper Econometrc Specfcaton of the Gravty Model, The World Economy, Vol. 20, Issue No 3. McPherson, M.A et al (2000). A Re-examnaton of the Lnder Hypothess: a Random-Effects Tobt Approach. Internatonal Economcs Journal Vol.14 No. 3 Paas, T. (2000) Gravty Approach for Modelng Trade Flows between Estona and the Man Tradng Partners, Workng Paper, No. 721, Tartu Unversty Press Panagarya, A. (2000). Preferental Trade Lberalzaton: The Tradtonal Theory and New Developments, Journal of Economc Lterature 38, June, Poyhonen,P. (1963), A Tentatve Model for the Volume of Trade betweencountres, Weltwrtschaftlches Archv 90: Santos Slva, J.M.C. and Tenreyro, Slvana (2006), The Log of Gravty, The Revew of Economcs and Statstcs, 88(4), pp Santos Slva, J.M.C. and Tenreyro, Slvana (2011), Further smulaton evdence on the performance of the Posson pseudo-maxmum lkelhood estmator, Economcs Letters, 112(2), pp Soloaga, I., and Wnters, A. (1999) Regonalsm n the Nnetes: What Effectson Trade? Development Economc Group of the World Bank, mmeo. Rebecca M. (1989) A Poltcal-Economc Model of U.S. Blateral Trade. The Revew of Economcs and Statstcs, Vol. 71, No. 1. pp Tnbergen, J. (1962), Shapng the World Economy. Suggestons for aninternatonal Economc Polcy, New York. We, S.-J. (1996), Intra-natonal versus Internatonal Trade: How StubbornAre Natons n Global Integraton?, NBER Workng Paper Yeats, A. (1998), ÒDoes MercosurÕs Trade Performance Rase Concerns Aboutthe Effects of Regonal Trade Arrangements?Ó The Word Bank EconomcRevew 12 (1): Vner, J. (1950) The Customs Unon Issue. New York: Carnege Endowment for Internatonal Peace Notes: Table 1: Trade Flows n EAC: (US $ m) Year Tanzana s Exports to Tanzana s Exports to Rest of World (1) as % of (2) Kenya (1) (US $m) (2) (US $m) ,904 2,982, ,888 4,050, ,313 4,734, Avg ,368 3,922, Year Tanzana s Exports to Tanzana s Exports to Rest of World (2) (1) as % of (2) Uganda(1) (US $m) (US $m) ,651 2,982, ,205 4,050, ,634 4,734, Avg ,830 3,922, Year Tanzana s Exports to Tanzana s Exports to Rest of World (2) (1) as % of (2) Rwanda (1) (US $m) (US $m) ,805 2,982, ,802 4,050, ,160 4,734, Avg ,922 3,922, Year Tanzana s Exports to Tanzana s Exports to Rest of World (2) (1)as % of (2) Burund (1) (US $m) (US $m) ,632 2,982, ,132 4,050, ,848 4,734, Avg ,204 3,922, Source(s). World Bank 2010 EAC

11 Table 2: Economc Growth n EAC Regon Regon Burund Kenya Rwanda Tanzana Uganda Source: EAC Report (2011) Table 3: OLS results of Tanzana s and Kenya s blateral trade wth the rest of EAC countres Explanatory varables DependentVarable (Tanzana s Total trade (log)) Dependent Varable (Kenya s Total Volume of Trade (log)) OLS Coeffcent OLS Coeffcent Constant **(4.757) 6.269***(0.429) GDPs(Product) 1.115***(0.234) 0.702***(0.059) PCIs(Product) ***(0.461) ***(0.121) Dstance ***(0.532) ***(0.171) No. obsrv R Adusted R Hausman Test Note; 1) the numbers n parenthess are standard error 2) *** and ** and * means sgnfcant at 1%, 5% and 10% level, respectvely Table 4: Random effect Regresson results of Tanzana s Export trade to the rest of EAC (Kenya ncluded) countres Explanatory varables Model 1. Model 2. Model 3 Model 4 Model 5 RE Coeff RE Coeff. RE Coeff. RE Coeff. RE Coeff. Constant 6.652***(1.592) 3.806*(1.535) 8.141***(1.405) 16.88***(3.063) 3.412***(0.395) GDP (Tanzana) 3.738*(1.414) 3.842*(1.440) 3.842*(1.440) 4.278*(1.588) 1.448*(0.524) Importer GDP 0.153(0.134) **(0.284) 1.007***(0.244) (0.130) (0.120) Dstance (0.208) 3.029***(0.676) 2.605***(0.555) 2.606***(0.556) 7.433***(1.296) ImporterPop 1.883***(0.481) Importer PCI (0.482) Exc. Rate 0.708***(0.189) openness 0.959***(0.037) R Adusted R Hausman Test Note; 1) the numbers n parenthess are standard error 2) *** and ** and * means sgnfcant at 1%, 5% and 10% level, respectvely Table 5: Random effect Regresson results of Kenya s Export trade to the rest of EAC (Tanzana ncluded) countres Explanatory varables Model 1. Model 2. Model 3 Model 4 Model 5 OLS Coeff OLS Coeff. OLSCoeff. OLSCoeff. OLS Coeff. Constant 5.275***(0.789) 4.859***(0.389) 5.574***(0.374) 4.235***(0.484) 5.490***(0160) GDP(Kenya) 1.07*(0.504) 1.129*(0.483) 1.123*(0.459) 0.959*(0.521) 0.863***(0.079) GDP(mporter) 0.414***(0.078) 0.226**(0.071) 0.552***(0.055) ***(0.028) **(0.027) Dstance ***(0.432) ***(0.110) ***(0.160) ***(0.177) **(0.112) Pop.(Importer) 0.319**(0.158) PCI(Importer) (0.113) Exchange Rate ***(0.057) 0.090***(0.024) Openness 1.0***(0.050) R Adusted R Hausman Test Note; 1) the numbers n parenthess are standard error 2) *** and ** and * means sgnfcant at 1%, 5% and 10% level, respectvely 23