"Understanding the Welfare Implications of Preferential Trade Agreements" M. Ayhan Kose and Raymond Riezman

Transcription

1 "Understandng the Welfare Implcatons of Preferental Trade Agreements" M. Ayhan Kose and Raymond Rezman CGR Workng Paper No. 45/99 October 1999 Centre for the tudy of Globalsaton and Regonalsaton (CGR), Unversty of Warwck, Coventry CV4 7AL, Unted-Kngdom. URL:

2 Understandng the Welfare Implcatons of Preferental Trade Agreements M. Ayhan Kose a and Raymond Rezman 1 Graduate chool of Internatonal Economcs and Fnance, Brandes Unversty; Department of Economcs, Unversty of Iowa. CGR Workng Paper No. 45/99 October 1999 Abstract: Ths paper examnes varous mplcatons of Preferental Trade Agreements (PTAs), namely Customs Unons (CUs) and Free Trade Areas (FTAs), n the context of a mult-country general equlbrum model based on comparatve advantage consderatons. We calbrate the model to represent countres wth symmetrc endowments, and compare the mpact of those agreements wth free trade and a non-cooperatve Nash equlbra. Utlzng aggregate and dsaggregate welfare change measures, we quantfy the welfare effects of trade arrangements. In partcular, we develop a numercal approxmaton procedure to decompose the welfare changes nto two components assocated wth the varatons n terms of trade and volume of trade. The results of our analyss ndcate that FTAs are better than CUs on welfare grounds for the world as a whole snce both member and nonmember economes enoy welfare benefts n an FTA. Further, we show that, for certan endowment dstrbutons, upon formaton of an FTA, nonmember economes get larger welfare benefts than member economes do. Nonetheless, member economes have larger welfare gans n CUs than n FTA. Our welfare decompostons suggest that a sgnfcant fracton of the welfare changes n both member and nonmember countres s explaned by the volume of trade effect for both types of PTA. Ths mples that, havng free access to larger markets, along wth greater market power are both mportant aspects of PTA. Comparson across endowment dstrbutons ndcates that as countres become more dvergent n ther endowments, the volume of trade effect gets more pronounced for CUs as well as for FTA. The absence of polcy coordnaton between the members of FTAs decreases the market power of the member economes and nduces welfare losses that are assocated wth the terms of trade effect. However, the ten-ns of trade effect results n sgnfcant welfare gans for the members of CUs snce they ontly determne ther tarff rates. Keywords:., General Equlbrum, Tradng Blocs, Free Trade Agreements, Customs Unons, Tarffs, Welfare. Address for correspondence: Ray Rezman, Department of Economcs, Unversty of Iowa, Iowa Cty, IA 52242, UA e-mal: raymond-rezman@uowa.edu a Graduate chool of Internatonal Economcs and Fnance, Brandes Unversty, Waltham, MA 02254, e-mal: akose@lemberg.brandes.edu. 1 We benefted from the suggestons of Erc Bond. Carsten Kowalycyk, and teve Matusz. Earler versons of ths paper were presented at the 1997 North Amercan ummer Meetngs of the Econometrc ocety (Pasadena), 1997 Mdwest Internatonal Trade Meetngs (t. Lous), Unversty of Melbourne, Australan Natonal Unversty, and Unversty of Iowa. We would lke to thank the partcpants for ther feedback. The usual dsclamer apples.

3 Non-Techncal ummary: In ths study we construct a very smple model to study varous mplcatons of two types of preferental trade agreements, free trade assocatons and customs unons. In ths model countres have symmetrc endowments and dentcal preferences. Usng smulaton technques we calculate consumpton allocatons, tarffs, prces, and trade volumes that result from the varous trade arrangements. We compare the mplcatons of customs unons and free trade assocatons wth those of free trade and an ntal equlbrum for whch we use a Nash equlbrum n tarffs. Our study revels several regulartes and dfferences across preferental trade agreements: frst, n a free trade assocaton equlbrum, the nonmember country charges hgher tarffs on mports than the member countres of the agreement do. In contrast, the member countres charge hgher tarffs on mports than the nonunon country does n a customs unon equlbrum. econd, whle formaton of a free trade assocaton does not lead to an ncrease n the tarff rates of member countres, t nduces hgher tarff rates n the nonmember economy. trkngly, when countres have suffcently uneven endowments,.e. when they have seemngly more market power n ther export goods, the member countres reduce ther tarff rates n a customs unon equlbrum. Thrd, upon the formaton of a customs unon, the terms of trade of the members mproves at the expense of nonmembers. In contrast, a nonmember economy n a free trade assocaton equlbrum enoys a terms of trade mprovement, snce member economes do not coordnate ther tarff polces. Fourth, the formaton of free trade assocatons leads to more trade n both member and nonmember economes than the formaton of customs unons. After studyng the mpact of preferental trade agreements on model varables, we analyze the welfare mplcatons of dfferent types of preferental trade agreement. Our results suggest that free trade assocatons are better than customs unons on welfare grounds for the world as a whole. Also, member economes have larger welfare gans n customs unons than n free trade assocatons. Further, we show that, for certan endowment dstrbutons, upon formaton of a free trade assocaton, nonmember economes get larger welfare benefts than member economes do. Our welfare decompostons ndcate that a sgnfcant fracton of the welfare gans n both member and nonmember countres s explaned by the volume of trade effect for both types of preferental trade agreements. Ths result suggests that havng free market access s a maor beneft of partcpatng n a preferental trade agreement. The terms of trade effect generates relatvely large welfare gans n the member economes of customs unons snce the members ontly determne ther tarff rates. The absence of polcy coordnaton between the members of free trade assocatons decreases ther market power and ths produces welfare losses that are also assocated wth the terms of trade effect.

4 1. Introducton The world tradng system has been gong through many changes n recent years. Among the many one notably mportant change s that attenton has shfted away from the multlateralsm of the WTO (formerly GATT) towards preferental trade agreements. In partcular, the number of preferental trade agreements has nearly doubled n the last four years. Most of these agreements are Free Trade Areas rather than Customs Unons. 1 These developments as well as the ongong ntense debate about ther contrbuton to the multlateral tradng system have put varous mplcatons of preferental trade agreements n the forefront of research n nternatonal trade. 2 urprsngly, to date there has been no study provdng a comprehensve, general equlbrum examnaton of the smlartes and dfferences across dfferent types of preferental trade agreements (PTAs), namely Free Trade Areas (FTAs) and Customs Unons (CUs). 3 In ths paper, our obectve s to fll ths gap by systematcally analyzng the varous mplcatons of these agreements n a general equlbrum settng. In partcular, we examne the followng questons: frst, what are the effects of dfferent types of preferental trade agreements on welfare, tarffs, prces, and the volume of trade? In the context of ths broad queston, our man goal s to understand the welfare mplcatons of PTAs for member and nonmember economes as well as for the world, whle provdng further nsght nto the mpact of these agreements on the macroeconomc varables such as tarffs, prces, and the trade volume. econd, what are the contrbutons of the varatons n the terms of trade and volume of trade to the welfare changes assocated wth dfferent types of PTAs? To deal wth these questons we construct a hghly stylzed mult-country general equlbrum model n whch tarffs are endogenously determned. Our model s a pure-exchange economy n whch tradng patterns are determned by comparatve advantage consderatons. Ths feature of the model smplfes the analyss allowng us to employ numercal methods to fnd equlbrum allocatons. We calbrate ths model for artfcal economes n whch countres are symmetrc n ther endowments. Countres can sgn preferental trade agreements, such as CUs and FTAs, wth each other. In addton, they can establsh free trade (FT), or they can behave non-cooperatvely by chargng optmal tarffs on the mports from other countres,.e. Nash equlbrum. We smulate the model and fnd consumpton 1 ee ampson (1996) for more nformaton about the recent ncrease n the number of preferental trade arrangements. The WTO (1995) reports that most notfcatons made to GATT have nvolved free trade areas, and the number of customs unon agreements s small. 2 ee Rezman (1997) for an extensve dscusson of the queston whether or not the new spate of preferental trade agreements are a help or hndrance n the goal of attanng free nternatonal trade. ee also Perron and Whalley (1996), Baldwn and Venables (1995), and Ether (1996). 3 Perron and Whalley (1994, 1996) provde several results assocated wth CUs and FTAs usng a general equlbrum model calbrated to represent maor tradng regons n the world. Bagwell and tager (1993a, 1993b) study the nteractons between the formaton of FTAs and CUs and multlateral trade lberalzaton, but ther focus s on the behavor of tarff rates durng ths process. 1

5 allocatons, prces, terms of trade, and volume of trade for member and nonmember economes and for every tradng regme. We contrast the mplcatons of CUs and FTAs wth those of FT and Nash equlbrum. We utlze two complementary welfare change measures to examne welfare effects of PTAs. The frst calculates the aggregate consumpton change n member and nonmember economes whch occurs wth the formaton of trade agreements. We then examne a measure whch decomposes the welfare effects nto two components: the varaton n aggregate ncome nduced by the movements n terms of trade and the change n aggregate ncome caused by the change n volume of trade. The second measure provdes useful ntuton about dfferent sources of welfare effects by solatng the mportant forces affectng aggregate welfare changes: the terms of trade effect explans the changes n the market power and the volume of trade effect accounts for the mpact of free market access on the aggregate welfare. Our paper s part of a rapdly growng research program whch nvestgates a varety of ssues related to preferental trade agreements. Inspred by the semnal work of Vner (1950), ths lterature has mostly focused on trade creaton and trade dverson effects arsng from trade agreements. Krugman (1991a) examnes these two effects n the context of a monopolstcally compettve model and shows that formaton of CUs can potentally lead to hgher external tarffs and can consequently result n lower world welfare. Ths result s nterpreted as mplyng that recent PTAs consttute a potental threat to the multlateral tradng system, snce they ncrease the possblty of a global trade conflct. A number of studes, followng Krugman, analyze the mpact of the smultaneous formaton of CUs on tarffs and welfare by relaxng a varety of hs assumptons, such as absence of transportaton costs, sze of tradng blocs etc. 4 Although a sgnfcant fracton of recent PTAs have taken the form of FTAs, the massve body of the lterature sparked by Krugman s study has largely gnored FTAs, and exclusvely focused on the ssues pertanng to CUs. In a recent paper, Krueger (1995) rases ths ssue, and nvestgates the dfferences between FTAs and CUs. Utlzng the Vnearan termnology, she fnds that trade-creatng CUs are superor to the FTAs on welfare grounds. Her results suggest that an FTA results n more trade dverson than a CU does, snce sustanablty of FTAs requres a varety of rules of orgn requrements. Ths result s n sharp contrast wth the tradtonal argument advanced by hbata (1967) who clams that FTAs result n larger welfare gans than CUs. 5 4 ee Krugman (1991b), Deardorff and tern (1994), nclar and Vnes (1995), yropoulos (1995), Bond and yropoulos (1996), Frankel, ten, and We (1996), Haveman (1996), Bhagwat and Panagarya (1996), and ten and plmbergo (1996). 5 Pomfret (1996) also challenges Krueger s analyss and clams that the rules of orgn requrements, whle they could be detrmental to the welfare of member and nonmember countres, are unable to ustfy Krueger s concluson. Josh and hvakumar (1997) consder a model nvolvng a symmetrc Cournot olgopoly and fnd that FTAs result n hgher world welfare than CUs do. 2

6 Whle our analyss provdes some mportant nsghts nto these dscussons our approach to the problems posed above dffers n some key ways from others n ths lterature. Frst, we construct a fully specfed general equlbrum model whch requres only the specfcaton of fundamental endowment parameters. We are able to analyze the strategc nteractons between member and nonmember economes snce tarffs are endogenously determned n our model. econd, FTAs and CUs are easly examned n our framework and the dfferences and smlartes across dfferent types of PTAs are documented. Thrd, our smple model economy wth the symmetrc endowment structure solates the mpact of rules of orgn requrements assocated wth FTAs from the other fundamental consderatons whch shape the polces of member and nonmember economes upon the formaton of PTAs. Hence, our model asks the same questons as Krueger, namely are CUs preferred to FTAs, but focuses on how strategc nteractons, rather than rules of orgn requrements, affect welfare. In partcular, we consder terms of trade and volume of trade effects, and develop a numercal approxmaton method to decompose the aggregate welfare effects nto those two components. In ths regard, our analyss consttutes a maor departure from the lterature that was fostered wth the extensve use of the Vneran trade creaton trade dverson termnology. 6 Whle ths taxonomy has enlarged our understandng of dfferent effects of trade arrangements, and proven to be useful for smple descrptve arguments, a number of researchers have advanced concerns about the relatve merts of t. 7 For example, the trade creaton trade dverson termnology does not provde unambguous results about the welfare mplcatons of trade agreements. Further, t gnores the effects assocated wth the ntal tarff levels, and how the PTA s themselves nfluence tarff levels and terms of trade. Whle accountng for the changes n tarffs and terms of trade, our welfare decomposton also separates the market power and market access forces. The former force s assocated wth the terms of trade effect, and the latter one s descrbed by the volume of trade effect. We frst document the regulartes assocated wth the effects of PTAs on tarffs, prces, and the volume of trade: our fndngs suggest that formaton of CUs mght not result n hgher tarff rates and FTAs nduce lower protectve barrers. These two results challenge the conventonal noton that smultaneous 6 To the best of our knowledge, our study s the frst one whch numercally mplements ths type of welfare decomposton. 7 Hamlton and Whalley (1985) note that the usual focus n the customs unon lterature on trade creaton and trade dverson effects not only neglects the mportant ssue of the ntal level of partner protecton, but also neglects the potental for a ont terms of trade mprovement through common external protecton. Harrson, Rutherford, and Wooton (1993) dscuss the shortcomngs of ths termnology notng that ths method s, at best, unrelable and could gve lttle clue as to the overall welfare effect on the country. rnavasan, Whalley, and Wooton (1993) note that much of the lterature measurng the effects of regonal ntegraton agreements on trade and welfare uses a classcal Vnearan trade creaton/trade dverson framework whch s not well suted to the study and quantfcaton of more recent regonal ntegraton. Kowalyczyk (1996) provdes a crtcal analyss of the lterature and llustrates the potental problems whch mght arse wth the msuse of t. He concludes that the Vner s termnology has severed customs unon theory from the theory of multlateral tarff reform. 3

7 formaton of trade agreements can result n hgher protectve barrers whch consttute a potental threat to the multlateral tradng system. Regardng prces, our fndngs ndcate that n a CU equlbrum, the terms of trade of the members mproves at the expense of nonmembers. In contrast, a nonmember economy n an FTA equlbrum enoys a terms of trade mprovement, snce the member economes do not coordnate ther tarff polces. urprsngly, the consumers n the member countres pay lower prces for mports than those n the nonmember economy n an FTA because the member economes of FTAs charge lower tarff rates than the nonmember does. The study of trade volume reveals an nterestng regularty: the formaton of FTAs leads to more trade n both member and nonmember economes than the CUs. Unlke Krueger, we fnd that FTAs are better than CUs on welfare grounds for the world as a whole. Whle both member and nonmember economes enoy welfare gans n an FTA, only the member economes gan and the nonmembers lose n a CU. The total welfare gan of the unon members exceeds the loss of nonmember economy, and the formaton of a CU results n an ncrease n the world welfare over Nash equlbrum. Our results also suggest that member economes have larger welfare gans n CUs than n FTAs. Further, we show that, for certan endowment dstrbutons, upon formaton of an FTA, nonmember economes get larger welfare benefts than member economes do. Our welfare decompostons suggest that a sgnfcant fracton of the welfare gans n both member and nonmember countres s explaned by the volume of trade effect for both types of PTAs. Ths result mples that for welfare consderatons havng free access to larger markets s qute mportant relatve to havng more market power. The terms of trade effect accounts for much of the welfare gan of member economes of CUs snce the members ontly determne ther tarff rates. The absence of polcy coordnaton between the members of FTAs nduces welfare losses whch are also assocated wth the terms of trade effect. The organzaton of the paper s as follows: n secton 2, we present the model. Next, welfare gan calculatons are explaned n the followng secton. In secton 4, we examne the mplcatons of trade agreements and dscuss the results. We conclude wth a bref summary of our fndngs and suggestons for future research. 2. The Model We construct a general equlbrum model of a representatve world economy. 8 Our model s suffcently comprehensve to ncorporate an arbtrary number of countres. We assume that transacton 8 A number of researchers have recently used ths setup to analyze dfferent ssues: see Kennan and Rezman (1990), yropoulos (1995), Haveman (1996), Kose and Rezman (1997), and Rezman (1997). Our study extends the research program ntated by Kennan and Rezman n several dmensons: frst, unlke Kennan and Rezman who provde some examples only, we provde a comprehensve examnaton across dfferent types of PTAs. econd, we focus on the welfare mplcatons of trade 4

8 costs assocated wth nternatonal trade are equal to zero, and transfer payments between countres are not allowed. Countres set tarffs optmally and can choose to not be part of any trade agreement and charge the optmal tarff, or they could decde to on a coalton wth other countres. They could be part of an FTA, a CU or can mutually agree on establshng FT. In the FTA member countres agree to free trade between themselves, but are allowed to set ther external tarffs ndependently. A CU s an FTA wth the addtonal provson that the external tarff s set ontly by the members. A CU (or FTA) of all countres s, of course, Free Trade The Envronment Consder a world of n countres. In each country, the agents derve utlty by consumng m dfferent goods. Each country s endowed wth a fxed amount of each commodty. Let y be country s endowment of good. Assume that each country conssts of ndvduals wth dentcal Cobb-Douglas preferences. Then the utlty functon of a representatve agent s the same as the aggregate and gven by (1) U = β ln x m = 1 where U s the utlty of country, and β s the weght country puts on commodty m ( β = 1, = 1,..., n). x = 1 denotes the aggregate consumpton of good n country. Ths preference formulaton results n a lnear expendture system whch allows us to employ numercal methods to solve the model. Further, wth ths structure we do not have to specfy elastctes, and can state our results n terms of fundamental endowment parameters. The net mports n each good, z, s defned to be z = x - y. As we have already stated above, countres charge optmal tarffs (export taxes or subsdes) on mports (exports). We call these tarffs or subsdes tarffs n the succeedng dscusson. Denote the tarff charged by country on mports of good arrangements, and sources of welfare effects assocated wth dfferent types of PTAs. Thrd, we do not lmt our analyss wth only one trade pattern and consder the alternatves. 9 We do not consder the equlbra whch are consstent wth Artcle XXIV and Kemp-Wan adustment. ee yropoulos (1995) for an extensve dscusson of those equlbra. Whle the assumpton that the ex-post tarff rates should be consstent wth the Artcle XXIV s a seemngly plausble one, the World Trade Organzaton (1995) reports that only sx out of sxty-nne trade agreements are found to be n full conformty wth the Artcle. 5

9 by t. If the world prce for good s p, then the domestc prce of good n country s q = ( 1 + t ) p. Gven that each country conssts of dentcal ndvduals, aggregate demand s obtaned from maxmzng the utlty subect to the budget constrant m (2) p (1 + t ) x = I = p (1 + t ) y + p t z = 1,2,..., n, = 1,2,..., m. = 1 m = 1 where I s ncome of country and conssts of ncome from the endowment plus tarff revenue whch s rebated to consumers lump-sum The Equlbrum The aggregate expendture n each country must equal the value of ts endowment at the equlbrum. nce we do not allow trade defcts or surpluses, the balance of payments constrant of each country s gven by (3) W = p x = p y = 1, 2,..., n. m m = 1 = 1 W s the aggregate expendture of country. In addton to ths constrant, the world demand for each good, should be equal to world supply, Y : n (4) x = y = Y = 1, 2,..., m. n = 1 = The Numercal oluton Method. The logarthmc utlty results n a lnear expendture system, n whch agents allocates a fxed fracton of ther ncome on each good. The frst order condtons of the maxmzaton problem lead to the followng demand functons, (5) x β I = ( 1 + t ) p, = 1, 2,..., n, = 1, 2,..., m. 6

10 Denote expendture of country on good wth W. It s easy to show that expendture s allocated across goods n proporton to β and t. Hence, W = α W where α s a functon of β and t. The world supply of each good s normalzed to one unt. Ths mples that the world prce of each good s equal to aggregate world expendture on that good n (6) W = p, = 1, 2,..., m. = 1 Usng equatons (1), (5), and (6), we derve an analytcal expresson for utlty of each country as a functon of tarffs, for any gven preference and endowment dstrbuton. We employ a recursve numercal soluton method to fnd an approxmate soluton for equlbrum allocatons, prces, and tarffs snce we cannot obtan a closed form soluton for each equlbra. 10 The ntuton of our soluton method s straghtforward: for a gven endowment matrx, we can compute the equlbrum wth optmal tarffs. Thus, when a country consders changng ts tarff t has to make ths calculaton for any proposed tarff change. Once a country changes ts tarff we have to re-compute optmal tarffs for the other countres. Ths contnues untl no country wants to alter ts tarff rates, n other words, no country can ncrease ts utlty by changng ts tarffs. Operatonally, FTAs and CUs put constrants on the feasble tarff matrx. For an FTA, tarffs between the member countres are set to zero, and external tarffs are set ndependently. Alternatvely, one mght consder a more general case n whch we let the CUs or FTAs have tarffs between the member countres. We requre that free trade be practced between the members when a PTA s sgned. Ths means to compute the equlbrum, the approprate tarffs are constraned to be zero. Computng customs unons equlbra s a bt more complcated: one can thnk of a CU as an FTA wth the added feature that the external tarff s set ontly. There are two mportant ssues about the tarff determnaton problem. Frst, f the two FTA members exercse dfferent tarff rates, then ths results n trade deflecton and dfferent tarff rates cannot be sustaned. econd, n general, there s a conflct of nterest between the member countres as to what the external tarff should be n a CU. We calbrate our model n a way that so we can rule out both problems. pecfcally, we consder countres wth symmetrc endowments, and ths removes the possblty of emergence of those problems ee Kennan and Rezman (1990) for detals of our soluton method. 7

11 2.4. Calbraton In order to utlze the soluton algorthm, we should specfy the number of countres, the number of goods and endowment of each country. We assume that m=n=3 and β =1/3 for all,=1,2,3. Any varaton n cross country preferences or preferences across commodtes can be replcated by a sutable adustment of endowments. As we have already stated above we assume that the world supply of each good, Y, s one unt. Each country has an endowment of y = s unts of good and y = 05. ( 1 s) unts of good ( ). s, whch s between zero and one, denotes the degree of symmetry (or smlarty) between countres. Ths endowment structure mples that countres are dentcal up to a relabellng of the goods. As s ncreases, the countres become more dssmlar, and ther market power n ther export goods ncreases. Dependng on s, two symmetrc trade patterns can occur at any equlbrum: when s<1/3, each country exports two goods ( z < 0, ), mports the other ( z > 0 ), and charges taxes ( t < 0, ) on ts exports. When s>1/3, each country exports one good ( z < 0 ), mports the other two ( z > 0, ), and charges tarffs on ts mports ( t > 0, ). When s=a=1/3, there s no trade, snce countres have dentcal preferences and endowments. We let s vary from 0.1 to 0.9. Ths changes the dstrbuton of endowments across countres. For each s, we compute equlbrum allocatons n Nash, CU, FTA, and FT equlbra. In CU and FTA equlbra, two countres establsh a free trade agreement and leave the thrd one out. nce we lmt our analyss wth trade agreements between the symmetrc economes, we do not have to model the choce of coalton partners. 3. Measurng Welfare Changes In ths secton, we frst descrbe an aggregate welfare change measure whch s the compensatng varaton n consumpton. Then, we explan how we decompose the welfare effect for each country nto the changes assocated wth varatons n volume of trade and terms of trade The Aggregate Welfare Change Measure We use the measure of Compensatng Varaton n Consumpton to evaluate the aggregate welfare changes assocated wth trade agreements. 12 The compensatng varaton n consumpton s the fracton δ by 11 Rchardson (1993) dscusses the frst ssue, and shows that even f rules of orgn are strctly enforced t stll may not be possble to sustan dfferent tarff rates. Gatsos and Karp (1991) addresses the ssue of conflct of nterest between the members of the CUs. 12 Ths measure s slghtly dfferent than the Equvalent Income Varaton measure whch looks at the change n ncome at constant prces. The measure of Compensatng Varaton n Consumpton s wdely used n macroeconomcs and fnance 8

12 whch the consumpton allocatons should be decreased n a free trade equlbrum, whch can be an FTA, CU, or FT, to keep the representatve agent wth the same utlty as the one n the Nash equlbrum. o, the welfare change, δ, s calculated as N F U ( x, x, x ) = U (( 1 δ ) x,( 1 δ ) x,( 1 δ ) x ) (7) U U N F : utlty under a Nash equlbrum : utlty under a preferental trade equlbrum (FTA, CU, or FT) Unlke some other welfare measures, such as utlty level of the representatve consumer, our welfare change measure provdes a transparent nterpretaton of welfare effects of trade agreements, snce welfare changes are measured n consumpton unts. For example, t s possble to assess the percentage change n consumpton wth the establshment of a PTA. We also examne the mplcatons of preferental trade agreements on the world welfare. nce the countres are ex-ante dentcal up to a relabellng of the goods n ther endowments n our model, they wegh equally n our world welfare measure. The world welfare change, δ w, s calculated usng the same procedure above and gven by the followng formula (8) 3 N F w w w U ( x, x, x ) = U (( 1 δ ) x,( 1 δ ) x,( 1 δ ) x ) = 1 = 1 : country, = 1, 2, The Dsaggregate Welfare Change Measure In order to provde an ntutve bass for our results and to further examne the potental sources of the welfare effects of preferental trade agreements, we decompose the welfare changes nto two components whch are assocated wth the varatons n the terms of trade and volume of trade of each lterature to evaluate the costs of busness cycles and of the lack of nternatonal consumpton rsk sharng. ee Lucas (1987), Cole and Obstfeld (1991) for the use of ths measure n dfferent contexts. Kose and Rezman (1997) and Rezman (1997) utlze ths measure to evaluate the welfare mplcatons of dfferent types of PTAs. ee Perron and Whalley (1994) for the use of Equvalent Income Varaton, and Frankel, ten, and We (1996) for the use of utlty of the representatve agent as the welfare measure. 9

13 country. Our measure, whch we call t the dsaggregate welfare change measure, was frst carefully worked out by Kowalyzck (1996). 13 In the prevous secton, we focused on the aggregate welfare change whch s measured n terms of percentage changes n consumpton. Now, we measure the welfare effects by the varatons n aggregate ncome. Consder country whch nvolves n a preferental trade arrangement wth another country n our model. We denote the aggregate ncome of country by I. Let q be the m-element column vector of domestc prces,.e. tarff nclusve prces, of country. z s a row vector wth m elements lstng country s net mports,.e. the dfference between mports and exports. We can rewrte (2) n terms of country s net mports, and use the frst order condtons of the optmzaton problem to get (9) di = q dz It s possble to rewrte the equaton (3) to get the balanced trade condton, pz = 0, where p s an m- element column vector of world prces,.e. tarff exclusve prces. Now, totally dfferentatng ths equaton and subtractng (9) from t yelds the followng formula (10) di = 123 z dp + ( q p) dz tot vot In ths expresson, the frst term s the terms of trade (tot) effect whch s the nner product of the vectors of net mports and changes n world prces. The second component denotes the volume of trade (vot) effect whch s the nner product of the tarff wedge and the change n net mports. In our framework, the tarff rates are ad valorem, q = ( 1 + t ) p, where ( 1+ t ) s an nxn dagonal matrx wth 0 s on ts dagonal. The above welfare measure dsaggregates welfare changes for nfntesmal changes n tarffs. However, n our analyss tarffs change dscretely when countres move from one equlbrum to another. In 13 There were some earler studes suggestng ths type of dsaggregate welfare change calculaton, but Kowalczyk was the frst one who convncngly argued that ths type of welfare taxonomy s superor to the tradtonal trade creaton and trade dverson termnology. ee Bond (1990) for the dervaton of ths measure. ee Gunter (1989), Baldwn and Venables (1995), and Wnters (1997) for surveys on dfferent types of welfare decompostons. 10

14 order to adapt ths measure for our purposes, we develop a numercal approxmaton method whch allows us to use (10) to decompose the overall welfare effects assocated wth dscrete changes n tarffs. 14 To do ths we frst defne a transton path and dvde the total change n tarffs nto smaller dscrete steps. Accordngly, countres go through N steps n whch they reduce ther tarff rates proportonally. We then calculate (N-2) ntermedate equlbra on the transton path whch leads countres from the ntal equlbrum to the fnal one. Let k ( 1 k N ) ndex the equlbrum on the transton path. The ntal equlbrum, whch s the Nash equlbrum s denoted by k=1, and the fnal equlbrum, (FTA, CU, or FT), corresponds to k=n. In each step, the countres charge tarff rates whch are between the ntal and fnal tarff rates. Next, we solve the model for each ntermedate set of tarff rates, and fnd respectve vectors of prces, exports, and mports. Knowng these varables, we calculate the dsaggregate welfare gans for each equlbrum along the transton path. The terms of trade effect s approxmated wth z ( k ) p ( k ), and the volume of trade effect s found by [ q ( k ) p ( k )] z ( k ) for equlbrum k. Then, we sum the changes n welfare over each equlbrum on the transton path. The terms of trade effect for country s found by N 1 (11) tot = z ( k) p( k) k = 1 and smlarly the volume of trade effect s N 1 (12) vot = [ q ( k) p( k)] z ( k) k = 1 The summaton of (11) and (12) produces an approxmaton to the overall welfare change of country due to a PTA. We then compare ths estmated measure of welfare change wth the actual welfare change whch we have computed drectly. o each value of N gves us a welfare decomposton that can be compared to actual welfare. As N ncreases, the accuracy of our approxmaton gets better. By ncreasng N one can get the welfare decomposton as close as desred to the actual welfare level. Through smulatons we fnd that for our problem N=4 gets us suffcently accurate answers. 15 For example, consder the transton path from the Nash equlbrum to the FTA equlbrum n our model. 14 ee Harrson, Rutherford, and Wooton (1993) for a smlar welfare decomposton. They suggest the use of the same numercal approxmaton method, but ther welfare change measure s dfferent than ours. In partcular, ther measure decomposes the effects nto two parts assocated wth changes n domestc prces and changes n tarff revenue. 11

15 We know allocatons n the ntal and fnal equlbra. nce countres reach the fnal equlbrum n 4 steps, each member country (the nonmember country) reduces (ncreases) ts tarffs by 25 percent n the frst step. In the second step, the tarff rate charged by each member country (the nonmember country) s 50 percent lower (hgher) than that n the Nash equlbrum. In the thrd step, the tarff rate s decreased by 75 percent by the members, and s ncreased by 75 percent by the nonmember. Fnally, countres charge the tarff rates correspondng to the FTA equlbrum. We compute the terms of trade and volume of trade effect for each adacent ntermedate equlbra, and then sum each term across all of the tarff rates. Ths provdes the decomposton of the welfare gans due to a move from the Nash to the FTA equlbrum Understandng Trade Agreements In ths secton, we frst present the results of our smulatons whch provde several nterestng regulartes about the effects of PTAs on tarffs, prces, and the trade volume of member and nonmember countres. We, then, examne the welfare mplcatons of FTAs and CUs, and comment on the current debate about the relatve merts of preferental trade agreements. The last secton analyzes the welfare effects of trade agreements employng the terms of trade and volume of trade termnology. Whle presentng the results for both types of trade patterns, for brevty, we dscuss only those assocated wth the trade pattern n whch each country exports one good and mports the other two,.e. s s greater than a Tarffs, prces, and trade volume Fgure 1 presents the equlbrum tarff rates charged by member and nonmember countres n dfferent types of PTAs. As one would expect, the larger the dsparty n endowments, the larger s the equlbrum tarff rate. An ncrease n s can be nterpreted as an ncrease n the monopoly power of the countres n world trade, snce t ndcates that each country has a larger share of the world supply of ts export good. Ths nduces a rse n equlbrum tarff rates as s rses. Fgures 1a and 1b reveal a maor dfference between CUs and FTAs. In an FTA equlbrum, the nonmember country charges hgher tarffs on mports than the member countres of the agreement do. In contrast, the member countres charge hgher tarffs on mports than the nonunon country does n a CU equlbrum. Ths regularty can be explaned wth the externalty-nternalzng effect whch was frst 15 We check the robustness of our results to the number of ntermedate equlbra. For the values greater than 4, there were some small changes n the magntudes of gans (or losses), but these were not sgnfcant enough to have an mpact on our results. 16 We also compare the aggregate welfare gans wth the dsaggregated ones, and verfy the consstency of the two measures for each equlbra. There are an nfnte number of paths from the ntal to the fnal equlbrum. Harrson, Rutherford, and Wooton (1993) explan the path ndependence ssue for a smlar welfare decomposton. They show that the dsaggregated welfare change does not depend on the transton path. We use the same transton path for every equlbrum, so the dependence of welfare effects to transton path s not a problem here. 12

16 advanced by Kennan and Rezman (1990): n a CU there s nternal free trade (as n the FTA), and n addton, the unon members ontly set a common tarff on mports of goods from the nonunon country. Ths generates a tarff externalty whenever two countres mport the same good, because a tarff mposed by one country lowers the prce pad by both. In a CU, unlke n an FTA, the members nternalze the tarff externalty by settng ther tarffs ontly. In fgures 1c and 1d, we compare the pre-agreement tarff rates wth the post-agreement ones. Whle formaton of an FTA does not lead to an ncrease n the tarff rates of member countres, t nduces hgher tarff rates n the nonmember economy. trkngly, when countres have suffcently uneven endowments,.e. when they have seemngly more market power n ther export good, the member countres reduce ther tarff rates n a CU equlbrum. In partcular, when s s greater than 0.81, the members of the customs unon set a lower tarff rate than that they charge ndependently n a Nash equlbrum. In a recent paper, yropoulos (1995) convncngly argues that, snce the nternal free trade among CU members causes an ncrease n the prce elastctes of demand for the exports of all countres, the market power of unon members and non-unon members decrease, and they both reduce ther tarff rates. Combned wth our earler explanaton about externalty-nternalzng effect, whch ncreases the tarff rate, yropoulos argument on the trade-lberalzng effect explans the decrease n the tarff rates of member countres. When s s greater than 0.81, the trade lberalzng effect outweghs the externalty nternalzng force, and drves the tarff rates down n a CU. When s s less than 0.81, the externalty nternalzng force becomes more pronounced and nduces an ncrease n the common tarff rate practced by the unon members. We can explan the fall n the tarff rates charged by the members of an FTA wth the help of trade lberalzng effect. In an FTA, there s no externalty nternalzng force, and the only present effect s the trade lberalzng force whch provdes an ncentve to member countres to decrease ther tarff rates upon the formaton of the FTA. Unlke n an FTA, n a CU agreement, nonmember economes reduce ther tarff rates to explot the gans assocated wth the ncrease n trade volume. We further examne the change n tarff rates to assess the magntude of these two opposng forces. Fgures 1e and 1f present the percentage changes n equlbrum tarff rates movng from Nash equlbrum to CU and FTA equlbra. As s rses,.e. as each country gets a larger fracton of the world supply of ts export good, the trade lberalzng effect becomes more pronounced relatve to the externalty nternalzng effect. For example, when s s 0.4, formaton of a CU leads to a 42 percent ncrease n the tarff rates of the member countres. The ncrease n the tarff rates falls to 1.3 percent when s s 0.8. In an FTA equlbrum, the mpact of the trade lberalzng force becomes more evdent as s rses: when s ncreases from 0.4 to 0.8, the decrease n the equlbrum tarff rate of the member countres rses from 53 percent to 73 percent. 13

17 These results challenge the noton that smultaneous formaton of trade agreements must result n hgher protectve barrers whch consttute a potental threat to the multlateral tradng system. Krugman (1991a) shows that CUs can potentally ncrease external tarffs due to the non-cooperatve behavor of large economc unts. 17 He consders a model n whch each country produces a sngle good that s dfferentable from the other goods produced by other countres. He lmts hs analyss wth CUs whch are formed by symmetrcally endowed countres. Krugman s fndngs ncreased the fears that smultaneous formaton of PTAs can cause hgher protectve barrers, and lead to a global trade conflct. Our results pant a more optmstc pcture: frst, the formaton of a CU can, ndeed, lead to lower, not hgher, tarff rates even wthn the context of a model wth symmetrcally endowed countres that set tarffs optmally. The complete specalzaton assumpton, that drves Krugman s model, amplfes the externalty-nternalzng effect and, not surprsngly, results n hgher tarffs. On the other hand, our model, n whch trade s drven by the dfferences n endowment dstrbutons, llumnates the nteracton between the tarff ncreasng and tarff decreasng forces,.e. the externalty-nternalzng and trade-lberalzng effects, and shows that a CU arrangement between countres wth suffcently dverse endowments lowers protectve barrers. econd, and more mportantly, our fndngs suggest that understandng of preferental trade agreements and the new wave of regonalsm requres a much broader examnaton of dfferent types of tradng arrangements ncludng FTAs along wth CUs. For example, our results ndcate that the formaton of FTAs lead to lower tarff rates. In a recent study, the World Trade Organzaton (1995) reports that most of the regonal trade agreements take the form of FTAs, and the number of CUs s small. 18 Consderng that a number of recent trade agreements are FTAs, any analyss, whch solely concentrates on the CUs, unnecessarly exaggerates the magntude of the threat produced by these agreements and gnores the mportant dfferences, whch have maor mplcatons on several relevant varables, between dfferent forms of PTAs. 19 Fgure 2 provdes nformaton on the behavor of dfferent prce measures. It s evdent that as countres become more powerful n ther export markets,.e. as s rses, the prces of ther export goods 17 In partcular, he notes that ndeed, we may expect as a general presumpton that a customs unon, beng a larger unt wth more market power than any of ts consttuent members, wll have an optmal external tarff that s hgher than the preunon tarff rates of the member natons. 18 ee World Trade Organzaton (1995, p. 27). 19 A number of studes challenge the Krugman s result: Bond and yropoulos (1996) extend Krugman s work, and show that, for relatvely low elastcty of substtuton and low degree of comparatve advantage, formaton of CUs can lead to lower tarff rates. yropoulos (1995) studes the same setup here, and concludes that CUs can result n lower protectve barrers. nclar and Vnes (1995) fnd that when countres are suffcently smlar n endowments, formaton of CUs lowers tarffs. Moreover, they consder FTAs and fnd that when trade agreements take the form of FTAs, they do not result n hgher tarff rates. Perron and Whalley (1994) smulate a computable general equlbrum model calbrated to represent dfferent tradng regons n the world. They fnd that FTAs between large and small countres can result n lower tarff rates. Our study shows that both CUs and FTAs can result n lower protectve barrers. Our results are also dfferent from the ones of nclar and Vnes snce we fnd that CUs produce lower protectve barrers when the countres have suffcently dssmlar, not smlar, endowments. 14

18 ncrease. Fgures 2a and 2b show the mpact of polcy coordnaton on the terms of trade of member and nonmember economes. In fgure 2b, upon the formaton of the unon, the terms of trade of the members mproves at the expense of nonmembers. In contrast, a nonmember economy n an FTA equlbrum (fgure 2a) enoys a terms of trade mprovement, snce member economes do not coordnate ther tarff polces. As fgures 2c and 2d ndcate, consumers n the member countres of CUs and FTAs pay lower prces (tarff nclusve prces) for mported goods than those n nonmember economes. It s nterestng to observe that even though the terms of trade of FTA members get worse, the consumers n the member countres stll pay lower prces for mports than those n the nonmember economy (fgure 2c). Ths can be explaned wth a regularty we documented above: FTAs lead to hgher (lower) tarffs n nonmember (member) countres. Fgures 2e and 2f contrast domestc prce levels (tarff nclusve prces) n Nash equlbrum wth those n FTA and CU equlbra. As fgure 2e ndcates, for member countres the domestc prces of mported goods n Nash equlbrum are hgher than those n CU and FTA equlbra. Fgure 3 presents the effects of trade agreements on the trade volume of the member and nonmember countres. We measure a country s trade volume wth the rato of ts exports to ts GDP. From fgures 3a and 3b, t s clear that a member country of a PTA has a larger trade volume than the nonmember economy does. The gap between the tradng volumes of member and nonmember economes becomes more pronuounced as the dsparty n endowments ncreases. Fgures 3c and 3d show that the formaton of FTAs leads to more trade n both member and nonmember economes than the formaton of CUs. As t can be seen n Fgure 3e, FT (Nash) equlbrum results n the largest (smallest) trade volume. Inspecton of fgure 3f reveals that there s more trade between a member country and the nonmember country n an FTA equlbrum than a CU Aggregate Welfare Effects We frst analyze the effects of PTAs on the allocaton of resources usng the aggregate welfare measure descrbed n secton 3.1. We present the results of our smulatons n fgure 4. As endowments of countres become more dssmlar, they get larger welfare gans through establshng PTAs. Ths regularty s an expected one consderng that each country consumes all three goods, and as the endowments of countres get more dvergent, nternatonal trade becomes more mportant. Fgure 4a dscloses an nterestng feature of FTAs: when s s less than 0.68, member countres get larger welfare gans than the nonmember does. urprsngly, as the dsparty n endowments gets larger, n partcular when s s greater than 0.68, the nonmember country gets larger welfare gans than the members of the FTA do. In the prevous secton, we fnd that formaton of an FTA results n hgher tarff rates n nonmember countres than n member economes. In addton, we also document that the terms of trade of the nonmember country 15

19 mproves at the expense of the members. The trade lberalzng force nduces the FTA members to decrease ther tarff rates, and the absence of polcy coordnaton prevents them manpulatng the terms of trade n ther favor. Hence, the nonmember country s able to get large welfare benefts for certan endowment dstrbutons, and n some cases, these gans are larger than those of the members. Unlke FTAs, CUs are harmful to nonmember countres as t s shown n fgure 4b. In a CU, the nonmember country always loses, and the members always gan. The comparson of the sze of welfare gans assocated wth dfferent types of PTAs provdes further nsght about the effects of these agreements on members and nonmembers: for example, consder the world economy when s s equal to 0.8. Upon formaton of a CU, member economes ncrease ther consumpton by 8.4 percent, and the nonmember country suffers from an almost 12 percent fall n ts consumpton. When an FTA s establshed, member countres enoy more than 5 percent rse n ther consumpton whle the nonmember ncreases ts consumpton by almost 7 percent. Fgures 4c and 4d show that CUs (FTAs) are better than FTAs (CUs) for member (nonmember) countres on welfare grounds. Interestngly enough though, formaton of a CU does not necessarly mprove members welfare over FT. If each country s endowment of ts export good s greater than 0.67, then member economes are better off at an FT than at a CU. Ths result s seemngly contradctory to the noton that formaton of CUs helps to generate larger economc unts whch can, because of ther sze, manpulate the terms of trade n ther favor, and have larger welfare gans than the ones they can get at FT. However, s s beng small s equvalent to the CU s beng large n our framework. In other words, as countres become more dssmlar, the formaton of a CU does not necessarly ncrease the welfare of the members, snce the nonmember economy gets larger market power over ts export good, and the trade lberalzng force outweghs the externalty nternalzng force and leads to lower tarff rates n the member countres. Consderng the recent debate about the welfare mplcatons of PTAs, we can now answer the queston of that are CUs stumblng blocks or steppng stones to the attanment of global free trade? Our results ndcate that for certan endowment dstrbutons (when s<0.69), CUs pose a threat to the multlateral tradng system snce the member countres can get larger welfare gans at CU than at FT. Therefore, f we rule out transfer payments from the nonmember economes to the member ones, CUs can be consdered as stumblng blocks. 20 However, when countres are suffcently dssmlar, or when they have more market power n ther export markets, then CUs do not necessarly consttute a danger to the 20 Kowalczyk and ostrom (1994) analyze a model of customs unons n whch sde payments are permtted. Ther results ndcate that the use of transfer payments can facltate the attanment of free trade. 16

20 multlateral tradng system, and n fact they can serve as a steppng stone to the attanment of global free trade. 21 We establshed that whle CUs beneft member countres, they hurt nonmember economes. As far as FTAs are concerned both member and nonmember countres get welfare gans. We now turn our attenton to the effects of PTAs on world welfare. Fgure 4f presents the results of world welfare calculatons. As expected FT s the best outcome for the world as a whole. FTAs are always better than the CUs snce whle both the member and nonmember economes gan n an FTA, only members gan at the expense of the nonmember country n a CU. The total welfare gan of the members exceeds the loss of nonmember economy n a CU, and the world welfare ncreases over Nash equlbrum. Our results also shed lght on an mportant polcy debate whch has recently ganed momentum: are FTAs better than CUs? Krueger (1995), usng Vneran trade creaton trade dverson termnology, argues that CUs are better than FTAs from the vewpont of welfare economcs. Krueger assumes that after the formaton a CU, member economes levy a tarff rate whch s the average of ther pre-agreement tarffs, and the formaton of an FTA does not result n any change n tarff rates,.e. the member economes mantan ther pre-agreement tarffs. Krueger s analyss focuses on the rules of orgn requrements whch are exercsed by the members of FTAs to prevent trade deflecton from the low-tarff country to the hghtarff country when two FTA members charge dfferent tarff rates. These requrements nduce more trade dverson n FTAs than n CUs, and Krueger s results favor CUs over FTAs. Our analyss s dfferent than Krueger s: frst, we rule out rules of orgn problems by employng a symmetrc endowment dstrbuton. The rules of orgn consderatons bas the Krueger s concluson aganst FTAs. Whle abstractng from the rules of orgn requrements, our framework emphaszes the strategc nteractons between the members and nonmembers to dfferentate CUs from FTAs. econd, snce tarffs are endogenously determned n our model, we do not mpose any a pror assumptons about the pre- and post-agreement tarff rates. The tarffs comng out of the general equlbrum model used here do not ustfy the assumptons about the tarff rates made by Krueger. Thrd, whle Krueger s study consders only the world welfare, we examne the welfare mplcatons of PTAs for the member and nonmember economes as well as for the world. 21 To see ths, consder the followng experment: startng at Nash equlbrum, all the countres cooperatvely and unformly reduce ther tarff rates. If they reduce ther tarff rates by 75 percent, ths results n 8.7 percent consumpton gan n all countres. As we have already noted above, member economes ncrease ther consumpton by 8.4 percent upon the formaton of a CU. In other words, 75 percent multlateral unform tarff reducton results n larger welfare gans for all economes ncludng the potental members of a CU. Ths smple experment shows that through multlateral tarff reducton, countres can get larger welfare gans than they can get through establshng CUs. Ths result mples that formaton of CUs can be a frst step to the realzaton of multlateral free trade. We also examne the welfare gans assocated wth 25 percent and 50 percent multlateral tarff reductons. The results of our smulatons are presented n fgure 4f. 17