PhD dissertation. A Dynamic General Equilibrium Analysis of Jordan s Trade Liberalisation

Transcription

1 PhD dssertaton A Dynamc General Equlbrum Analyss of Jordan s Trade Lberalsaton Omar Ferabol Eultzstraße Chemntz emal: omar.ferabol@wrtschaft.tu-chemntz.de PhD supervsor: Prof. Dr. Bernd Lucke Referee: Prof. Dr. Rchard Tol 1

2 1 Introducton Ths dssertaton ams at assessng the effects on the Jordanan economy of the preferental trade lberalsaton process undertaken by Jordan wth the European Unon (EU). The Assocaton Agreement (AA) between Jordan and the EU was sgned n 1997 and entered nto force n It elmnates progressvely tarffs on most ndustral goods mported by Jordan from the EU. Custom dutes on agrcultural products and processed agrcultural goods are gradually and only partally elmnated. After the 12-year transton perod n whch mport dutes are reduced, the Agreement ams eventually at creatng a free-trade area for most ndustral products between the EU and Jordan. The reducton of tarff rates on EU mports nto Jordan s expected to result n postve effects for the Jordanan economy. Lower mport dutes leads to lower mport prces of nvestment and consumpton goods, that n turn brngs about a postve mpact on consumer welfare. On the other hand, trade lberalsaton reduces government revenue. The magntude of the adverse effects wll be nfluenced by the measures taken by the Jordanan government to counteract the effects of revenue loss. Ideally, mport duty reducton ought to be accompaned by an approprate and parallel process of complementary economc reforms, such as reducton n government spendng, modernsaton of the tax system and broadenng of the tax base n order to offset the loss n custom dutes. Therefore, together wth the economc effects of trade lberalsaton on Jordan, ths work ams also at drawng mplcatons for domestc polcy responses accompanyng the trade lberalsaton process. In order to assess the mpacts of the Assocaton Agreement wth the EU on the Jordanan economy, a dynamc computable general equlbrum (CGE) model s specfed and then calbrated to the Jordanan economy. Ths methodology allows to capture fully the chan of events n the domestc economy, ther nteractons and ther dynamc effects when a polcy opton s mplemented. Partcular emphass s placed on the effects on consumer welfare. Usng a dynamc CGE model, the mpacts of gradually decreasng and eventually elmnatng tarff barrersnjordanformosteu ndustral goods are assessed. However, gven the need for domestc reforms parallel to the trade lberalsaton process, the mpacts of preferental trade lberalsaton are assessed along wth polcy choces amng at counterbalancng the negatve effects of trade lberalsaton on government revenue. 2

3 Computable general equlbrum models rely on socal accountng matrces (SAMs) to capture natonal ncome, producton and nput-output nformaton, and am at smulatng and evaluatng economc polces. The use of CGE models for polcy analyss has become wdespread for a wde range of applcatons for both developed and developng economes (de Melo, 1988). An appled CGE model should have the followng essental characterstcs: () consumers endowments of producton factors, () consumers preferences and demand functons for commodtes, () producton technology avalable to frms, and (v) set of equlbrum condtons (Shoven, 1983). Equlbrum n the model s charactersed by a set of prces and output levels n each ndustry such that, for all commodtes, market demand and supply are equal. Demand functons are homogeneous of degree zero and profts are lnearly homogeneous n prces. Therefore the absolute prce level has no mpact on the equlbrum outcome and only relatve prces are of any sgnfcance n the model. Market demands are the sum of ndvdual household demands, and they satsfy the Walras law (Shoven and Whalley, 1984). In dynamc models, household behavour s determned by the maxmsaton of the dscounted lfetme utlty. The nstantaneous utlty functon s defned over the doman of the consumpton goods n the economy and n some models t ncludes also lesure (Perera and Shoven, 1988). A complete equlbrum dataset for a sngle year must then be assembled. On the assumpton that the data represent an equlbrum of the economy, functonal parameters, such as share and shft parameters, are calbrated,.e. they are estmated n such a way that the model soluton reproduces the ntal dataset, called benchmark equlbrum. However, some parameters, namely the elastctes, are taken exogenously from the exstng lterature. Calbraton n a dynamc context requres addtonally the model to be parametersed to yeld an ntertemporal balanced growth path when the base polcy s mantaned. Exogenous shocks are then mplemented n the model, n order to compute a counterfactual equlbrum determned by the new polcy regme. The mpact of the polcy change s then assessed by comparson between counterfactual and benchmark equlbra (Shoven and Whalley, 1992). In analysng a wde range of polcy ssues, the general equlbrum approach has a man advantage over the partal equlbrum one, namely the possblty of capturng fully the chan of events and ther nteractons. In order to analyse the detaled effects of mport tarff reducton, the chan of events takng place when tarffs are cut should be examned (Bandara, 1991). A tarff rate reducton affects demand 3

4 patterns. The relatve prces of mports and domestc goods change and mports ncrease. Ths has an effect on the allocaton of resources wthn the tarff-reducng country. Consequently, changes n mport tarffs can not be consdered separately, snce ther repercussons are spread throughout the economy, through channels that affect producton, consumpton and nvestment decsons. Moreover, gven that trade lberalsaton s not mplemented n solaton, but t requres combnaton wth other approprate polces, ts economc effects should be computed together wth those brought about by the assocated polces. To my knowlege, there are two studes on Jordan s trade lberalsaton usng CGE models. D. Lucke (2001) mplemented a statc model to assess the fscal effects on Jordan of the Assocaton Agreement wth the EU, and to address the ssue of fscal responses amng at counteractng the loss n government revenue. Hosoe (2001) used a statc model to analyse the mpacts of the mplementaton of the Uruguay Round and the free trade arrangement wth the EU on Jordanan welfare. He fnds postve welfare effects brought about by the Uruguay Round and an addtonal welfare gan due to the EU-Jordan prefental trade agreement. The model mplemented n the frst part of the analyss s a neoclasscal dynamc computable general equlbrum (CGE) model, n whch one representatve household maxmses her future dscounted utlty by choosng optmal consumpton and nvestment paths. In the domestc economy full employment and perfect competton are assumed. Imperfect substtuton between domestc and foregn goods characterses nternatonal trade flows. Jordan s assumed to be a small economy,.e. t s a prce-taker n the nternatonal markets. The model s calbrated to 1998 dataset. Smulaton results of the process of preferental trade lberalsaton undertaken by Jordan show that the Assocaton Agreement wth the EU rases consumers welfare n Jordan and has postve mpacts on all macroeconomc varables n the long-run. However, n the short-run prvate consumpton s negatvely affected by trade lberalsaton, and ths may rase concerns about poltcal feasblty of the process of openng up domestc trade. Trade lberalsaton processes undertaken by many developng countres over the past years have been accompaned by wdespread concerns that openng up domestc trade n developng countres wll affect negatvely the poor and t wll deterorate the dstrbuton of ncome. Whereas most economsts agree on the fact that open economes perform better than closed ones, and open polces provde a sgnfcant 4

5 contrbuton to economc development and growth, many commentators fear that, both n the short and n the long-run, trade lberalsaton mght be harmful for poorer agents n the economy (Oxfam Internatonal, 2003 and 2005). In fact, t mght well be, as argued by Asbett (2005), that people s nterpretaton of the avalable evdence of the mpacts of trade lberalsaton on poverty s strongly nfluenced by ther values and by ther belefs about the process of globalsaton. Wnters et al. (2004) survey the emprcal work on trade lberalsaton and poverty. They pont out that there s plenty of evdence that trade lberalsaton affects each household groups, and that the ablty of households to respond to trade lberalsaton mpacts dffers across householdsgroups. Thetheory suggests that tradelberalsaton mght allevate poverty n the long-run and on average, and the emprcal evdence supports ths vew. However, they also warn that ths vew does not assert that trade polcy s always among the most mportant determnants of poverty reducton or that the effects of trade lberalsaton are always benefcal to the poor. Instead trade lberalsaton mples necessarly some dstrbutonal changes and, at least n the shortrun, t may reduce the welfare of some ndvduals and some of these may be poor. Wnters et al. (2004) also pont out that, gven the varety of factors that have to be taken nto account, t wll hardly be surprsng that there are no general comparatve statc results about the mpact of trade lberalsaton on poverty. However, n a WTO specal study, Wnters (1999) concludes that trade lberalsaton generally contrbutes strongly to poverty allevaton. He also recognses that most reforms mght create some losers, even n the long-run, and that some reforms could have temporarly a negatve mpact on poverty. The model wth one representatve household, descrbed above, s then extended to nclude heterogeneous consumers. Indvdual households tax rates, wage rates, ntal endowments of assets, transfers from government and abroad and ndvdual preferences are calbrated from data from a 2002 household survey. Introducng heterogeneous households nto a standard neoclasscal dynamc CGE model allows to address the ssue of how trade lberalsaton affects dfferent households. In the context of general equlbrum modellng several studes have been conducted to assess aspects of ncome dstrbuton (see Remer, 2002 for a survey). However, the approach used n ths dssertaton s the frst one analysng ncome dstrbuton n a dynamc general equlbrum framework wth utlty maxmsng agents as used by Ramsey (1928), Cass (1965) and Koopmans (1965). Theoretcal contrbu- 5

6 tons analyse the effects of mplementng heterogeneous consumers nto a neoclasscal framework (Chatterjee, 1994 and Casell and Ventura, 2000). However, the restrctons on the utlty maxmsng agents mposed by ths strand of lterature are not fulfllednthsmodelandwouldbeneglectedbytheavalablesurveydataforjordan. Specfcally, they assume the same rate of dscount for all household groups, whereas n the mult-household model mplemented n ths dssertaton the categores of households are charactersed by dfferent rates of tme preference, whch are calbrated from the dataset. Therefore, ths approach can be regarded as novel. As one would expect, effects of trade lberalsaton on Jordan are dfferent across ndvdual households, and n some smulatons one household group even experences a welfare loss. Therefore trade lberalsaton s not always Pareto mprovng for Jordan. In addton effects on welfare and ncome dstrbuton are opposte. Whle on the one hand welfare gans are slghtly larger for low-ncome households, on the other hand the gap n ncome between rch and poor ncreases, especally n the long run. The results are drven by the fact that captal stock of hgh-ncome households ncreases much more n the long run due to explotaton of nvestment ncentves. Moreover, poor households use ther amount of captal assets to smooth consumpton. The remanng fndngs confrms the analyss suggested by the model wth one representatve household on the aggregate level. Both models are programmed n the mathematcal software Gauss and are solved wth the relaxaton algorthm proposed by Trmborn et al. (2006). The dssertaton s structured as follows. Chapter 2 descrbes the Assocaton Agreement between Jordan and the EU and deals wth the update of the nputoutput table for Jordan. In chapter 3, the effects of preferental trade lberalsaton on the Jordanan economy are analysed by means of a standard trade CGE model, n whch one representatve consumer chooses optmal consumpton and nvestment path so as to maxmse future dscounted utlty. The model s calbrated to 1998 data. In chapter 4, the model s extended to nclude sx representatve households, n order to assess the welfare mpact of trade lberalsaton on each household class. As mentoned above, households represent dfferent ncome groups wth dfferent consumpton and tme preferences, levels of wealth, ncome, tax rates, and government transfers. The dataset s based on the 2002 socal accountng matrx (SAM) for Jordan, n whch households data are taken from the 2002 Jordanan Household Survey. For convenence, n the dssertaton the one representatve consumer model s denoted 6

7 as standard trade model, whle the model wth sx household classes s called poverty model. Chapter 5 draws the man conclusons. The appendces provde equatons and glossares of both the standard trade and poverty models, and tables and detals about the I-O table update. 7

8 2 Insttutonal framework and dataset 2.1 The EU-Jordan Assocaton Agreement The economc relatons between Jordan and the European Unon (EU) are governed by the Euro-Medterranean Partnershp, whch s mplemented through the EU- Jordan Assocaton Agreement (AA) and the regonal dmenson of the Barcelona Process. The EU-Jordan Assocaton Agreement s part of the blateral track of the Euro-Medterranean Partnershp. The ams of the Partnershp are to provde a framework for the poltcal dalogue, to establsh progressve lberalsaton of trade n goods, servces and captal, to mprove lvng and employment condtons, to promote regonal cooperaton and economc and poltcal stablty, and to foster the development of economc and socal relatons between the partes. The fnal am of the Assocaton Agreement s the creaton of a free trade area for most ndustral products between the EU and Jordan over a perod of 12 years, n conformty wth the provsons of the General Agreement on Tarffs andtrade(gatt). The Euro-Medterranean Partnershp was launched at the Euro-Medterranean Conference between the European Unon and ts orgnally 12 Medterranean Partners 1, and governs the polcy of the EU towards the Medterranean regon. The Euro- Medterranean Conference was held n Barcelona n 1995, and marked the startng pont of the Euro-Medterranean Partnershp, a wde framework of poltcal, economc and socal relatons between the Member States of the European Unon and Partners of the Mddle East and North Afrca (MENA) regon. The Euro-Medterranean Partnershp comprses two complementary tracks, the blateral and the regonal agenda. The framework for the blateral agenda s the Assocaton Agreement. The regonal agenda s mplemented through a number of regonal workng groups on a range of polcy ssues ncludng trade, customs cooperaton, and ndustral cooperaton. The latest EU enlargement, on 1st May 2004, has brought two Medterranean Partners (Cyprus and Malta) nto the European Unon, whle addng a total of 10 to the number of Member States. The Euro-Medterranean Partnershp thus comprses 35 members, 25 EU Member States and 10 Medterranean Partners (Algera, Egypt, Israel, Jordan, Lebanon, Morocco, Palestnan Authorty, Syra, Tunsa and Turkey). 1 The 12 orgnal partners are: Israel, Morocco, Algera, Tunsa, Egypt, Jordan, the Palestnan Authorty, Lebanon, Syra, Turkey, Cyprus and Malta. 8

9 Lbya has observer status snce Before the start of the Euro-Medterranean Partnershp, relatons between the EU and the countres n the MENA regon were ruled by the Cooperaton Agreements datng from the 1970s. Under the 1977 Cooperaton Agreement Jordan were granted duty-free access to the EU markets for most ndustral products and preferental access for agrcultural commodtes. The Cooperaton Agreement was unlmted n duraton, and t was not recprocal. In 1979 the Agreement allowed Jordan exports to enter the EU market free of quanttatve restrctons. The Euro-Medterranean Assocaton Agreement (AA) between Jordan and the European Unon was sgned n November It entered nto force on May 1 st, 2002, and replaced the 1977 Cooperaton Agreement. The Assocaton Agreement allows mports nto the EU of Jordanan products free of custom dutes and free of quanttatve restrctons, wth the excluson of agrcultural goods and processed agrcultural products. Custom dutes and charges on mports nto Jordan of EU products are progressvely abolshed, and dutes on agrcultural products are gradually and partally elmnated. The Agreement ams eventually at creatng a free-trade area for most ndustral goods between the EU and Jordan wthn 12 years by ts entry nto force. Table 2.1 shows the tme schedule of reducton of custom duty rates on EU mports to Jordan, provded by the Assocaton Agreement (Chapters 1 and 2 of Ttle II, Annex II and Lsts A and B of Annex III). Chapter 1 and Lsts A and B of Annex III of the Agreement apply to most ndustral goods, whle Chapter 2 and Annex II deal wth agrcultural goods and processed agrcultural products. The left column n table 2.1 shows the tme perod, n each other column the percentage of the baseyear mport tarff rates charged n the relevant perod are shown for four dfferent groups of goods lsted n the Assocaton Agreement. The group of commodtes n the second column of the table,.e. products lsted n Annex II, ncludes agrcultural products and processed agrcultural products. For these goods reducton of mport tarff rates starts four years after the entry nto force of the AA, and s only partal. The other groups of goods comprse the remanng ndustral products, for whch trade lberalsaton s complete. The establshment and the promoton of cross-border cooperaton wth the Medterranean Partners wll also be an mportant element of future regonal ntegraton. Jordan s already at the core of the man ntegraton process n the regon. It s a member of the Medterranean Arab Free Trade Area, the so-called Agadr agreement, that 9

10 was sgned n May 2001 wth Egypt, Morocco and Tunsa. Jordan has also sgned blateral FTAs wth several countres n the MENA regon, and s a member of the Great Arab Free Trade Area (GAFTA), wth other 13 countres who are members of the Arab League. After jonng the World Trade Organzaton (WTO) n Aprl 2000, as a step towards even broader trade lberalsaton Jordan sgned free trade agreements wth the Unted States n October 2000, and wth the European Free Trade Assocaton (EFTA) n June perod Annex II Lst A Annex III Lst B Annex III remanng entry nto force of the AA 100% 80% 100% 0% one year after 100% 60% 100% 0% two years after 100% 40% 100% 0% three years after 100% 20% 100% 0% four years after 90% 0% 90% 0% fve years after 80% 0% 80% 0% sx years after 70% 0% 70% 0% seven years after 60% 0% 60% 0% eght years after 50% 0% 50% 0% nne years after 50% 0% 40% 0% ten years after 50% 0% 30% 0% 11 years after 50% 0% 20% 0% 12 years after 50% 0% 0% 0% Table 2.1. Tarff reducton schedule of the AA. Trade lberalsaton n the form of a preferental trade agreement wth the EU s expected to provde benefts to Jordan n terms of lower mport prces of nvestment and consumpton goods that brng about hgher consumer welfare. The economc mpact of trade lberalsaton can be separated nto two types, statc and dynamc. The statc mpact s due to the nduced reallocaton of exstng resources, the dynamc mpact takes nto account the effect of openng up trade on the rate of captal accumulaton (Hoekman and Djankov, 1997). Therefore a key role n such a process s played by nvestment demand, that s potentally mportant to the dynamc behavour of output over the long-run (Francos et al., 1997 and Baldwn, 1993). On the other hand, trade lberalsaton reduces government revenue, due to decreasng mport tarff dutes. Such an mpact s lkely to be partcularly strong for Jordan, where government revenue reles heavly on custom dutes. 2 The magntude of the 2 Import dutes from EU trade n Jordan n the perod averaged 12% of total tax revenue and 2% of GDP, total mport dutes averaged more than one-thrd of total tax revenue and about 6% of GDP (Abed, 1998). 10

11 adverse effects on government revenue wll be nfluenced by the measures taken by the Jordanan government to counteract the effects of revenue loss. As ponted out n chapter 1, trade lberalsaton should be accompaned by an approprate and parallel process of economc reforms, such as reducton n government spendng, modernsaton of the tax system and broadenng of the tax base n order to offset the loss n custom dutes. As measures of fscal reform, the Jordanan government has harmonsed the General Sales Tax (GST) rates on domestc and mported goods, has replaced the GST, ntroduced n 1994, by a Value Added Tax (VAT) n 2000, and has undertaken an ncome tax reform n Update of the nput-output table Jordan s economy s currently undergong a rapd process of trade lberalsaton and market-orented economc reform. As mentoned above, the general sales tax (GST) has been replaced by a value-added tax (VAT), prvatsaton of state enterprses ganed momentum and Qualfyng Industral Zones establshed n economc cooperaton wth Israel have proved very successful. In the past few years, Jordan accessed the WTO and sgned free trade agreements, among others, wth the European Unon and the USA, whch provde for a stepwse reducton of mport tarff rates. Scentfc analyss amed at assessng the mpact of varous polcy reforms has largely reled on the use of computable general equlbrum (CGE) models, gven that suffcently long and relable tme seres for econometrc analyss are not avalable. Unfortunately, even for CGE analyss major mpedments exst. One of the major obstacles s gven by the fact that no recent nput-output (I-O) table for the Jordanan economy s avalable. Such a table s essental n organsng the avalable data for a partcular base year n the socal accountng matrx (SAM) whch s of basc mportance for CGE modellng. The most recent nput-output table for Jordan dates back to The matrx s therefore rather old and mght not adequately reflect the structural changes whch took place n the Jordanan economy snce the begnnng of the reform perod n the md-1990s. And even worse, the classfcaton used n the 1987 I-O table s ncompatble wth the system of natonal accounts (NA) currently used, as the NA system was substantally revsed n Whle the sectoral nomenclature of the data before and after the revson s smlar, an uncrtcal dentfcaton of sectors 11

12 wth smlar labels s, n fact, napproprate snce the dfferences n the defntons are non-neglgble. Updatng the 1987 I-O table s a task wth huge data requrements. Many of the data necessary for the update are n realty not avalable, and therefore estmates must be used. In order to update the 1987 I-O table the bproportonate RAS method (Bacharach, 1970, Bulmer-Thomas, 1982) s mplemented. Ths method can be used to update an old nput-output table f at least the row sums and the column sums of the I-O table are known. The RAS method The vectors and matrces of the model are ntally defned. The column vector y s the sectoral supply n the domestc economy,.e. domestc sales plus mports, where y s supply of sector =1,.., n y = y 1.. y n (1) x s the column vector of sectoral output, whch s a composte of domestc sales and exports, where x s output of sector x = x 1.. x n (2) The square matrx Q s the nput-output table of ntermedate consumpton goods q 1,1.. q 1,n Q = (3) q n,1.. q n,n where q,j s the spendng of sector j for ntermedate nput good, for, j = 1, 2,.., n. 12

13 ThesquarematrxA s the table of nput-output Leontef coeffcents (Leontef, 1966): a 1,1.. a 1,n A = a n,1.. a n,n where each coeffcent a,j s the spendng of sector j for the ntermedate good produced by sector dvded through by output of sector j, for, j =1, 2,.., n (4) a,j = q,j x j (5) The equlbrum between sectoral supply and demand n the domestc economy s therefore gven by the dentty y = Ax + z (6) where z s the column vector of sectoral spendng for fnal goods,.e. the sum of prvate consumpton, government consumpton and nvestment. Then r s defned as the column vector of total ntermedates produced by each sector,.e. the row sums of the matrx Q r = Qι (7) where ι s a vector of ones. Ths vector may be thought of and may be defned as total ntermedate supply. Smlarly, let c be the column vector of total ntermedate consumpton of each sector,.e. the column sums of Q: c = ι 0 Q (8) Ths vector may be thought of as total ntermedate demand. The matrx Q s known for one base year only, 1987 n ths partcular applcaton. Denote ths matrx by Q. For all subsequent years matrx Q s unknown, but t s assumed that the vectors r and c are known. The RAS or bproportonal method (Bacharach, 1970, Bulmer-Thomas, 1982) conssts n adjustng the rows and the columns of the exstng matrx Q, such that the entres n the adjusted matrx wll 13

14 add up to the row and column totals relatve of the update year. For ths purpose, the RAS method assumes that each entry n the updated matrx Q s bproportonal to the known ntal matrx Q,.e. that there are weghts w and z j for the typcal element q,j such that q,j = w q,j z j (9) Clearly, the weghts must be chosen subject to the restrcton that the row and column sums of the adjusted matrx equal the known margnal totals P j q,j = r, =1,..,n P q (10),j = c j, j =1,.., n The problem can be solved usng an teratve algorthm. The algorthm s wrtten n GAUSS by Bernd Lucke. Ths however, assumes data avalablty to whch the next secton now turns. Data For the base year 1987 data provded by the Department of Statstcs (DOS) of Jordan nclude the nput-output table along wth data on the sectoral spendng for ntermedate consumpton goods, sectoral data on output, exports and mports. The nput-output table Q s a square matrx -.e. the number of actvtes s the same as the number of commodtes - and ncludes 51 economc actvtes. For subsequent years ( ) data on ntermedate consumpton, gross output, mports and exports are avalable. These data are publshed wthn the revsed system of natonal accounts and are thus nconsstent wth the sectoral classfcaton used n the 1987 I-O table. Revsed natonal accounts data for 1987,. e. data conformng wth the new classfcaton are also avalable. Unfortunately, ths s not the case for the I-O table. In order to mnmse errors ncurred by the change n the classfcaton the actvtes are aggregated to just nne sectors producng goods. Further, the lack of dsaggregated data for mports and exports of servces mpled that all eght servce sectors had also to be aggregated. Ths s certanly a drawback, partcularly for a country such as Jordan, where the servce sector plays a very mportant role n the economy and a more detaled dsaggregaton for servces would be approprate. Therefore, the updated I-O matrces end up wth 10 sectors, as shown n Table

15 No. Economc actvty 1 Agrculture, huntng, forestry, and fshng 2 Mnng and quarryng 3.1 Manufactures of food, beverage and tobacco 3.2 Manufactures of textles, apparels, and leather product 3.3 Manufactures of wood, paper, prntng 3.4 Manufactures of petrolum and chemcals 3.5 Manufactures of rubber and other non metallc mneral 3.6 Manufactures of basc metals and fabrcated metal execpt machnery and equpment 3.7 Other manufactures 4-9 Servces Table 2.2. DOS classfcaton wth aggregated servces. Unfortunately, both for 1987 and all subsequent years, the revsed data are not sutable to easy applcaton of the RAS method. Three man problems were encountered. The frst problem s that varables are often evaluated at dfferent prces, e.g. ntermedate consumpton at producer prces and output at basc prces. To adjust the data, all varables are thus evaluated at producer prces. Basc prce s the prce receved by the producer from the purchaser for a unt of good or servce, mnus any taxes payable and plus any subsdes recevable on that unt. Producer prce s the value receved by the producer for a unt of product, mnus any deductble tax (such as VAT) charged on the purchaser, but t ncludes non-deductble taxes and subsdes. Ths requres the transformaton of sectoral output evaluated at basc prces nto sectoral output at producer prces by applyng the relevant net tax rate on the basc-prce output level. A second problem faced wth the data concerns the dfferent classfcatons of nternatonally traded goods that make fgures for mports and exports of goods ncompatble wth the rest of the data. The classfcaton used by the DOS n the orgnal nput-output table s smlar but not dentcal to the Harmonzed System (H.S.), whch s a classfcaton ncludng only goods. Avalable external trade statstcs for goods are provded under the Brussels Tarff Nomenclature (B.T.N.) classfcaton for the perod and under the H.S. from 1994 onwards. Thus correspondences must be used to convert external trade data from B.T.N. and H.S. nto the approprate DOS classfcaton. Some of these correspondences had to be constructed partcularly for ths purpose. The appendx provdes detals about the concordances used. The thrd problem s certanly the major one. As explaned above, applcaton of the RAS method requres the use of data across producton sectors on total output 15

16 (x), domestc supply (y), and supply and demand of ntermedate goods (r and c). Whereas vectors x, y and c are known for all years, the vector of ntermedate goods supply r s known only for the base-year,.e The vector r needs therefore to be derved for the remanng years. Note that data on fnalusesacrossproducton sectors are not avalable ether, so that t s mpossble to compute r as the resdual from output mnus fnal uses. Instead, the strategy used here conssts n estmatng r by usng sectoral data on supply and by adjustng t to make total demand for ntermedate goods n the economy equal to ntermedate goods total supply. Scalars, vectors and matrces referrng to the orgnal 1987 data are denoted wth and tme ndces are omtted n order to keep notaton smple. Defne s as the rato of r, total ntermedate nput supply of sector n 1987, to ỹ, supply of sector n 1987: s = r ỹ (11) Collect these ntermedate producton shares n the vector s: s = s 1.. (12) The vector s s used to obtan estmates of r for each year, on the assumpton that each r s proportonal to the respectve y.thevarabley s referred to as domestc supply, not output. The reason for ths s gven by the fact that entres q,j s n the matrx Q are domestc sales - ncludng mports and excludng exports. Moreover, the elements of r must be adjusted to make total supply of ntermedates equal to total demand n the ntermedate goods market,.e. P r = P c. Theestmateofr for each year s therefore gven by: P j r = s y c j Pj s (13) jy j s n Snce the revsed 1987 data dffer qute substantally from the orgnal fgures, the frst step conssts n updatng the 1987 I-O table to the new classfcaton. Ths enables to check how strongly the change n the accountng system affects the Leontef coeffcents. In order to dstngush between the Leontef coeffcents of the orgnal 16

17 1987 nput-output table and Leontef coeffcents of the updated 1987 table (based on revsed data), the former are denoted as 1987o and the latter as 1987r. Table 2.3 below shows the orgnal 1987 Leontef coeffcents,.e. the nput-output coeffcents computed by makng use of the orgnal 1987 dataset on ntermedate consumpton and gross output. By comparson, Table 2.4 shows the 1987r Leontef coeffcents obtaned from applyng the RAS method ,067 0,000 0,215 0,000 0,000 0,000 0,000 0,002 0,001 0, ,000 0,118 0,000 0,000 0,000 0,048 0,041 0,002 0,004 0, ,167 0,000 0,122 0,007 0,003 0,001 0,001 0,001 0,000 0, ,000 0,000 0,000 0,499 0,003 0,000 0,000 0,001 0,001 0, ,000 0,002 0,013 0,006 0,222 0,011 0,018 0,008 0,004 0, ,031 0,112 0,016 0,016 0,050 0,144 0,192 0,035 0,035 0, ,010 0,002 0,012 0,010 0,002 0,007 0,056 0,014 0,206 0, ,002 0,031 0,024 0,014 0,023 0,014 0,022 0,177 0,084 0, ,011 0,050 0,004 0,016 0,014 0,003 0,006 0,021 0,294 0, ,148 0,160 0,200 0,102 0,101 0,040 0,150 0,087 0,250 0,178 Table 2.3. Orgnal 1987 Leontef coeffcents ,082 0,000 0,212 0,000 0,000 0,000 0,000 0,003 0,001 0, ,000 0,100 0,000 0,000 0,000 0,068 0,037 0,004 0,002 0, ,196 0,000 0,115 0,004 0,004 0,002 0,001 0,001 0,000 0, ,000 0,000 0,001 0,487 0,006 0,001 0,000 0,002 0,001 0, ,000 0,002 0,012 0,003 0,310 0,015 0,016 0,013 0,002 0, ,040 0,101 0,017 0,011 0,077 0,216 0,185 0,062 0,024 0, ,015 0,003 0,014 0,007 0,003 0,013 0,063 0,029 0,159 0, ,003 0,029 0,025 0,009 0,037 0,021 0,022 0,323 0,057 0, ,016 0,053 0,005 0,012 0,025 0,005 0,007 0,044 0,229 0, ,204 0,158 0,223 0,074 0,170 0,065 0,158 0,169 0,182 0,170 Table 2.4. Estmated 1987 Leontef coeffcents. Results and Conclusons Usng the RAS algorthm for all subsequent years from 1988 to 2001 allows to analyse how the nput-output coeffcents change between the base-year 1987o and over the perod 1987r In the analyss of the Leontef coeffcents, t s sensble to focus on those coeffcents that are n some sense mportant. Whle many dfferent approaches - all of them somehow arbtrary - to choose the level of mportance are avalable, two reasonable crtera seem to be: 17

18 ()toselectthosecoeffcents a,j, whose value s large n at least one perod, where large values are those equal to or larger than 0.1; () to take those coeffcents a,j, whose assocated spendng for ntermedate nputs q,j s large n at least one perod, where now large values are defned those equal to or larger than 10% of total spendng of sector j for ntermedate consumpton goods c j n ths perod. Clearly, the crteron defned n () dentfes a subset of the coeffcents dentfed by crteron (), snce () postulates that the value of a certan ntermedate be more than 10% of total output value, whle () merely postulates that t be more than 10% of total ntermedate consumpton expenses of the partcular sector. Table 2.5 shows the mean values and the standard devatons (n brackets) of all Leontef coeffcents, computed over the perod 1987r The coeffcents whose value s larger than 0.1 for at least one observaton are shown n bold. Accordng to such crteron, there are 23 large coeffcents. Fgures n talcs show the coeffcents whose ntermedate consumpton entry s larger than 10% of total sectoral spendng for ntermedates for at least one observaton. Under crteron (), the group of large coeffcents ncludes the same 23 coeffcents selected under crteron (), together wth addtonal 9 coeffcents (0.015) (0.000) (0.039) (0.000) (0.000) (0.000) (0.000) (0.001) (0.000) (0.001) (0.000) (0.013) (0.000) (0.000) (0.000) (0.011) (0.005) (0.001) (0.000) (0.001) (0.054) (0.000) (0.017) (0.003) (0.001) (0.000) (0.000) (0.000) (0.000) (0.001) (0.000) (0.000) (0.000) (0.059) (0.002) (0.001) (0.000) (0.001) (0.000) (0.001) (0.000) (0.000) (0.001) (0.002) (0.022) (0.005) (0.002) (0.002) (0.000) (0.001) (0.011) (0.012) (0.003) (0.007) (0.009) (0.050) (0.019) (0.004) (0.002) (0.005) (0.006) (0.001) (0.004) (0.007) (0.001) (0.005) (0.014) (0.005) (0.022) (0.008) (0.001) (0.004) (0.004) (0.006) (0.005) (0.006) (0.003) (0.030) (0.005) (0.006) (0.007) (0.012) (0.001) (0.011) (0.006) (0.002) (0.002) (0.008) (0.030) (0.007) (0.028) (0.020) (0.027) (0.034) (0.016) (0.012) (0.012) (0.020) (0.018) (0.018) Table 2.5. Means and standard devatons of the coeffcents As can be seen, most of the large coeffcents le along the man dagonal and on the bottom row. Ths means that most of ntermedate trade nvolves several 18

19 actvtes that buy ntermedate nputs from themselves -.e. ntra-sectoral trade between the same sector plays an mportant role - and sectors that buy ntermedate goods from the servces sectors. More mportantly, the standard devatons of all large coeffcents are small, suggestng that the RAS procedure computed farly smlar coeffcents for all the years. Ths may be nterpreted as an ndcaton that the approxmatons used to assemble the approprate data and the update method n general may have worked qute well, snce, despte large swngs n partcular mportexport data, smlar estmates have been obtaned for all of the years. In order to fnd out f and how much Leontef coeffcents have changed over tme, the large coeffcents are regressed on a constant and tme trend: a,j = α + βt (14) Table 2.6 shows the sgn of tme trends of large coeffcents, whose estmate of β s sgnfcant. By lookng at selected graphs depcted n appendx 1, the general mpresson s that trend-nduced changes n Leontef coeffcents are slow and far from dramatc. Wth few exceptons, tme trends of Leontef coeffcents are postve for ntermedates produced by non-servce sectors and negatve for ntermedates produced by servce sectors neg pos pos 3.2 neg pos pos pos pos 3.5 pos pos pos pos pos pos 4-9 pos pos neg neg neg neg neg Table 2.6. Tme trends of the Leontef coeffcent estmates. Trendng Leontef coeffcents can, n prncple, ether reflect technologcal change or changes n market structure. Technologcal change s dfferent to measure, hence one could try to explore the hypothess that changes n Leontef coeffcents are mostly due to changes n market structure. For ths purpose, data on the number of frms n each sector are checked snce an ncrease n a Leontef coeffcent mght be due to a decrease n vertcal ntegraton and thus an ncrease n the number of frms n 19

20 aspecfc sector. Tme-seres data for the perod are avalable for sectors 2 to 3.7, whereas they are only partally avalable for sector 4-9, and not avalable at all for sector 1. The complete dataset s shown n the appendx. After excludng sector 1 because of lack of data and ncludng dummes for the mssng fgures n sector 4-9, the number of frms n each sector s regressed aganst a constant and tme. The number of frms shows sgnfcantly postve tme trend over the perod for all sectors. Ths can be taken as an evdence supportng the vew that vertcal ntegraton has decreased, and that competton has ncreased, partcularly n the manufacturng sectors. Regressng the Leontef coeffcents aganst a constant and the number of frms s supposed to yeld some nformatve and suggestve result. However, only 11 Leontef coeffcents depend sgnfcantly on the number of enterprses, as shown n table neg 3.1 neg pos pos pos 3.5 pos pos pos pos pos pos Table 2.7. Effects of number of frms on Leontef coeffcents. 20

21 3 The Standard Trade Model General equlbrum modellng approach for polcy analyss has become wdespread for both developed and developng economes. In developng countres, CGE models are commonly used for a wde range of polcy ssues. The polcy applcatons range from long-run development strateges on growth and resource allocaton, to tax and trade polcy reforms. As ponted out by de Melo (1988), the ssue of foregn trade polcy has occuped a center place n most of the applcatons. Even n the applcatons that do not focus on foregn trade, the way foregn trade s modelled plays a fundamental role n determnng the outcome of polcy smulatons. Over the past decades, the nterest generated by computable general equlbrum (CGE) modellng n applcatons to developng countres can be explaned by many factors. Frstly, the CGE modellng approach s approprate when analysng polcy changes and external shocks that affect the whole economy. Secondly, constructon of CGE models has been facltated by the development n many developng countres of relevant and statstcal data bases, such as socal accountng matrces (SAMs). Fnally, the computatonal constrants on the mplementaton of CGE models have been removed by advances n numercal soluton technques (Bandara, 1991). Many general equlbrum studes have assessed the economc mpacts of tarff reform and domestc complementary polces n developng countres. Harrson et al. (1996) assess the mpacts on Turkey of a custom unon arrangement wth the EU. Regonal ntegraton wth the EU s found to rase welfare n Turkey between 1% and 1.5%, dependng on the complementary polces adopted by the Turksh government. By usng a standard statc general equlbrum, Hoekman and Konan (1999) nvestgate the effects of the free trade agreement between Egypt and the EU on Egypt s welfare. They fnd large gans n welfare condtonal on elmnatng regulatory barrers and red tape. In a statc general equlbrum model for Syra, B. Lucke (2001) studes dfferent scenaros of preferental trade lberalsaton wth the EU, and focuses on the effects of tarff reform on government budget. The study fnds that government revenue losses caused by reducton n the EU mport dutes are farly large, but stll manageable. Go (1994) uses a model n a parsmonous and dynamc framework to examne ntertemporal effects of external shocks and adjustment polces n the Phlppnes, and concludes that complementary measures, consstng of domestc tax reform, are needed. Devarajan and Go (1998) present a smlar model, and analyse the 21

22 response of the Phlppnan economy to a terms-of-trade shock, tarff lberalsaton and fscal polcy changes. Harrson et al. (1997), usng a multregonal model, fnd that the mplementaton of the Uruguay Round has a negatve mpact on welfare n countres of the MENA regon. Prevous studes by Hosoe (2001) and D. Lucke (2001) on Jordan s trade lberalsaton mplemented statc models wth homogenous agents and focused on aggregate welfare and fscal effects. Hosoe (2001) nvestgates the mpacts of two trade polcy scenaros for Jordan, the Uruguay Round mplementaton and the establshment of a free trade area wth the EU, by usng a statc model based on Devarajan et al. (1990). Smulaton of the Uruguay Round shows that ts mplementaton would ncrease Jordan s welfare by 0.28%. The EU-Jordan FTA scenaro would further ncrease Jordan s welfare by 0.16%. The work by D. Lucke (2001) focuses on fscal effects of the EU-Jordanan Assocaton Agreement, and dscusses fscal responses amng at overcomng the loss n government revenue, such as smplfyng and harmonsng tax rates, and broadenng the tax base. However, these models do not account for ntertemporal effects due to captal accumulaton. The model mplemented n ths chapter s a neo-classcal open-economy snglecountry ntertemporal model, t bulds on prevous work done by Ferabol et al. (2003), whch s based on the dynamc framework developed by Devarajan and Go (1998). Dscounted lfetme utlty of the representatve consumer s maxmsed by choosng optmal consumpton and nvestment paths. In the domestc economy there are ten producton sectors, nne of whch producng goods and one producng servces. Producton sectors wll be denoted by the subscrpt. Perfect competton and full employment are assumed n all sectors. Frms use ntermedate nputs and value added output to produce fnal output wth a Leontef producton technology. Value added output s n turn a constant elastcty of substtuon (CES) composte of prmary nputs, captal and labour. Producton factors are assumed to be perfectly moble across sectors. Internatonal trade flows are charactersed by mperfect substtuton between domestc and foregn goods. Fnal sectoral output Q s allocated across domestc sales D and exports E through a constant elastcty of transformaton (CET) functon. Total sectoral absorpton X s an Armngton (1969) composte of domestc good D andmportedgoodm. Itsdfferentated among four uses: prvate consumpton C, government consumpton G, ntermedate nput q, nvestment I. The parameters n the Armngton functons are the same for all uses, as well as prces. 22

23 The domestc country s assumed to be a prce-taker n the nternatonal markets, that s world prces of mports and exports are exogenously determned. The model s mplemented by means of the mathematcal software Gauss and by employng the relaxaton algorthm proposed by Trmborn (2006). 3.1 Consumers The representatve consumer chooses consumpton and new captal so as to maxmse her dscounted lfetme utlty, subject to the budget constrant, the moton equaton of captal, the equalty between savngs and nvestment, and the gven ntal captal stock. The optmsaton problem s gven by: subject to max Z 0 u (C t ) e ρt dt (15) K = I δk = YD P C C P I δk (16) K (0) = K 0 (17) where C, YD, K are aggregate consumpton, dsposable ncome and captal of the representatve household, respectvely, I s aggregate nvestment, P C s the composte consumpton prce, P I s the composte prce of nvestment. The household dscounts future utlty wth dscount rate ρ, whch s calbrated from the data. The deprecaton rate of captal, δ, s also calbrated from the data. Dsposable ncome of the representatve household s gven by YD= 1 t Y wl + 1 t K rk + TR+ erf REM (18) where L s the fxed labour supply, w sthewagerate,t Y s the ncome tax rate, t K s the captal rent tax rate, r s the rate of return to captal, TR s government transfer to households, FREM are foregn remttances, expressed n foregn currency, and er s the exogenous exchange rate, whch s chosen as numerare. The nstantaneous utlty functon s gven by the constant relatve rsk averson (CRRA) utlty functon: u (C) =lnc (19) 23

24 whch mples an elastcty of substtuton between consumpton at any two ponts n tme equal to 1. Solvng the above dynamc optmsaton problem yelds the Euler equaton Ċ 1 t Y C = 1 t K r ρ δ (20) P I Equatons (16) and (20) characterse the dynamcs of the model. Household aggregate consumpton s a Cobb-Douglas composte of consumpton sectoral goods C = Ω C N Y =1 c θc ; Ω C > 0; 0 < θ C < 1 (21) where c s prvate consumpton of good produced by sector, N =10sthe number of sectors n the Jordanan economy, Ω C s the shft parameter and θ C s the share parameter of good n the Cobb-Douglas consumpton functon. Solvng the statc problem max c Ω C NY =1 c θc (22) subject to the constrant P C C = NX =1 P X c (23) yelds the functons of demand for consumpton good produced by sector c = θ C P C C P X (24) where c s prvate consumpton demand for the good produced by sector, P C s the prvate consumpton prce ndex and P X s the prce of the fnal good produced by sector. Household consumpton of each good and servce c s are n turn compostes of domestc and mport goods, modelled through the Armngton (1969) assumpton of constant elastcty of substtuton (CES) between domestcally-produced consumpton good cd and mported consumpton good cm. The representatve household chooses the optmal level of each domestc and mport good and servce for a gven value of 24

25 total consumpton, by takng the Armngton specfcaton as constrant of the costmnmsaton statc problem: subject to mn P C cm,cd c = P M cm + P D 1+vat D cd (25) γ 1 c = Φ ε (cm ) γ +(1 ε )(cd ) 0 < ε < 1; γ > 0, γ 6=1 γ 1 γ γ 1 γ (26) where P M s the composte mport prce, nclusve of all taxes and mport dutes, P D s the prce of the domestc good (net of taxes), and vat D s the VAT rate that apples to domestc goods; γ s the elastcty of substtuton between domestc goods and mports, Φ s the shft parameter, ε s the mports share parameter, and the subscrpt s the ndex for sectors. The demand functons for mports and domestc goods resultng from the mnmsaton problem gven by (25) subject to constrant (26) are gven by µ cm =(Φ ) (γ 1) ε P X γ c (27) P M and cd =(Φ ) (γ (1 1) ε ) P X γ c (1 + vat D ) P D (28) Aggregate mports of consumpton goods are then dsaggregated across mports from the EU and from the rest of the world, through a Cobb-Douglas specfcaton. 3 The optmsaton problem for the households apples to each sectoral producton and s gven by: mn {cm j } P M cm = X j PM j cmj (29) s.t. cm = Φ M Q j cm j ε j ; X j ε j =1 (30) 3 Imports can be dsaggregated across several dfferent regons or countres (e.g. Arab countres, EFTA countres, USA), but for the purpose of ths work, the basc dsaggregaton between the EU and the rest of world s consdered. 25

26 where cm j s households consumpton of foregn good mported from regon j, PM j s the prce of good mported from regon j nclusve of all taxes, Φ M s the shft parameter, and ε j s the share parameter of mports of good from regon j, wth each ε j 0. The elastcty of substtuton between mports s therefore constant and equal to one, beng the Cobb-Douglas specfcaton a partcular case of CES functon. The soluton to the above mnmsaton problem yelds the demand functons for mports dsaggregated across foregn regons: cm j = εj P M PM j cm ; =1, 2,.., N; j = EU,RW (31) The domestc prces of mported goods are determned exogenously, snce they depend on the fxed world prce of mports, PW M, the mport tarff rate, tm j,the VATrateonmportedgoods,vat M, and the exchange rate er: PM j = erp W M 1+tm j 1+vat M ; j = EU,RW (32) 3.2 Frms On the supply sde, constant returns to scale and perfect competton are assumed. Sectoral output n the domestc economy Q s determned by a two-stage producton technology, whch exhbts at the top ter a Leontef fxed-proportons specfcaton between ntermedate nput q j, produced by sector j andusedntheproducton process of sector, and value-added output VA : ½ VA Q =mn, q ¾ j,,... (33) a 0, a j, where a 0, s the fxed requrements of valued-added output VA,anda j, s the fxed requrements of ntermedate nput q j, for producton of aggregate output Q. At the second ter, ntermedate nput q j, s an Armngton CES composte of domestc and foregn ntermedate consumpton goods, qd j, and qm j,.totalmport of ntermedate goods s n turn a Cobb-Douglas composte of ntermedate nput regonal mports. Value-addedproductonneachsector s determned by a technology charactersed by a constant elastcty of substtuton between the two prmary nputs, captal 26