Predicting European Enlargement Impacts: A Framework of Inter-regional General Equilibrium

Transcription

1 Predctng European Enlargement Impacts: A Framework of Inter-regonal General Equlbrum d Arts Kancs y June 2001 Abstract Although the Computable General Equlbrum (CGE) model s not a new tool for analysng polcy mpacts, t has not ganed a wde popularty n regonal applcatons such as rural economes yet. Ths study demonstrates how a regonal CGE model can be appled for analysng regonal mpacts of changng global economc condtons as well as for assessng nter-regonal and ntersectoral mplcatons of polcy changes wth even lmted computatonal resources and lackng a full range of regonal economc data requred by a formal CGE analyss. In the emprcal analyss we have found that rural economes n the Central and Eastern European (CEE) accesson countres may expect the largest welfare gans from ntegraton nto the European Unon (EU) f the EU Structural Fund and CAP support measures are mplemented mmedately but markets are opened gradually to foregn competton. Keywords: CGE, EU, regonal economes, economc ntegraton. JEL class caton: 51, R12, R13, R23. 1 Introducton The changes experenced today n the rural economy and socety are wde rangng and complex. To a large extent they stem from changes n the global economc envronment, market polcy, and country openness to trade as well as from changng soco-economc norms and values n socety (Ferman 1999). At the same tme, rural The author acknowledges helpful comments from John Whalley and Juergen Wandel as well as semnar partcpants at the Unversty of Kel and IIASA. y Correspondng address: Internatonal Insttute for Appled Systems Analyss (IIASA), Schlossplatz 1, A-2361 Laxenburg, Austra. E-mal: kancs@asa.ac.at. 1

2 economes have to meet socety s rsng envronmental and recreatonal requrements, whch fundamentally change both the structure and the level of goods and servces demanded n rural economes (rabenstott and Meeker 1999). The European Structural Funds, as one of the most mportant rural development polcy plannng tools n European Unon (EU), are currently undergong a fundamental reorganzaton to extend the EU assstance programs toward the Central and East European (CEE) accesson countres (Vanhove 1999). The reforms, whch the EU assstance programs seek to foster n the CEE economes, are deep and wdespread n ther nature. They may a ect the compettveness of rural economes across and wthn countres n several ways. For example, they may alter the sectoral structure and actvty levels n rural economes as well as the economc performance of local economes n general. In order to be able to account for these economy-wde mpacts of a changng nsttutonal framework n the rural development plannng, t s necessary to assess ther potental mpacts pror to ther mplementaton n rural economes (Johnson and Scott 1997). Although, the ongong reforms of regonal development polcy n the EU play a major role n reshapng the CEE economes, there s another mportant factor determnng the spatal dynamcs of rural economes, namely, the decreasng mportance of agrculture. Accordng to European Commsson (1999), almost all rural areas n Europe have experenced a declne n employment and output share n agrculture durng the 1990s. As a result, the economc structure of many rural economes ncreasngly reles on servce actvtes. Partcularly, n northern Europe the role of rural areas has been sgn cantly altered by ncreasng demand for rural servces, such as, rural toursm (European Commsson 1999). As a result, only few rural areas can now be de ned as fully dependent on agrculture (Lopez-Bazo et al. 1999). These nter-sectoral and structural shfts n the structure of rural economes hghlght the mportance of a sectorally spec c approach to regonal and rural development plannng. Another mportant cause of the rapdly changng relatonshps between economc actors, sectors, and regons are changes n the global economc envronment, producton technologes as well as the emergence of completely new producton technologes. The nformaton technology (IT) sector, a drvng force of the so-called new economy, a ects economc space and hence rural areas n mportant ways. For example, mprovng tele-commutng technologes allow to ncreasngly separate the places where people lve and work. Ths makes t more attractve for people to lve n a rural areas and to work n an urban areas (rabenstott and Meeker 1999). One consequence of these changes s a blurrng dstncton between rural and urban space and a concomtant change n the nature and extent of nterdependences that exst between rural and urban areas. The growng nter-regonal nterdependences when people lve n one regon, earn money n another, and spend t n yet another regon, however, have severe consequences for regonal development polcy plannng. In order to be able to account for these changes, a consstent regonal development polcy plannng re- 2

3 qures a mult-regonal and mult-sectoral general equlbrum approach (Schndler et al. 1997). In lght of these regonal polcy challenges, the man goal of our study s to develop a relable analytcal tool for assessng spatal, sectoral and socal mpacts of changng macro-economc envronment n rural economes. A further objectve of our study s to dentfy possble rural development polcy optons and to compare them wth the European Structural Fund polces n selected CEE accesson countres. Whle there s lttle doubt among economsts about the ncreasng mportance of these Funds on local economc development, only few quanttatve studes carred out to date have appled an nter-regonal and nter-sectoral framework to consstently analyze polcy mpacts on rural and urban regons as well as on varous sectors n local economes (Azs 1997a). 1 Usng the example of Latva our study contrbutes toward closng ths research gap - understandng and predctng the causal structure of fundamental changes experenced n regonal economes today n Europe and assessng potental mpacts of changes n European regonal polcy settng wthn the framework of nterregonal general equlbrum. 2 Methods of Regonal Polcy Analyss Regonal polcy analyss has been strongly domnated by a model orentaton n the past decades (Njkamp et al. 1986). espte a wde varety of goals pursued, scope covered, and technques appled, regonal economc models can be class ed along several dmensons. Among other approaches, three classes of models have ganed a consderable attenton n regonal applcatons: multpler models (economc base and ncome-expendture), commodty ow models (nput-output models and socal accountng matrces), and models based on actvty analyss (lnear and nonlnear programmng). Although, often appled n regonal analyss, these models su er from several weaknesses (see, for example, Mdmore and Harrson-May eld 1996, Isard et al. 1998). Takng nto account the spec c requrements to regonal polcy analyss tools dent ed n the prevous secton (the spatal dmenson, the nter-sectoral and nterregonal nterdependences of economc agents, and asymmetrc mpact of unform polcy changes), n ths secton we focus our attenton on a regonal analyss method, whch does satsfy these requrements. Our approach, whch accounts for both prce and volume changes smultaneously, s the computable general equlbrum (CGE) model (Takayama and Judge 1976). As the name suggests, CGE models are based on the dea of market clearng (equlbrum). They study all markets n an economy smultaneously (general). Lke other mathematcal models, computable general 1 Some of the few successful attempts of buldng a computable general equlbrum model for a regonal economy are those of Klkenny (1993), Azs (1997b), and Ando and Takanor (1997), who managed to bult a regonal CGE model even for a developng economy-chna. 3

4 equlbrum models express behavoral relatonshps among the varous sectors and markets of an economy as a set of mathematcal equatons that can be solved applyng approprate algorthms and software (computable) (Scarf and Shoven 1984). 2 The fundamental characterstcs of a general equlbrum analyss are, therefore, the dent caton of nterdependences between goods, factors, assets, and markets of an economy as well as of behavoral varables of economc agents and of market clearng (equlbrum) condtons (Gunnng and Keyzer 1995). Methodologcally, the CGE model s a generalzaton of the nput-output model. However, whle t s based on the same nput-output table and the same general equlbrum perspectves as the nput-output model, the CGE model makes consderably fewer lmtng assumptons n descrbng economc relatonshps (Koh et al. 1993). Moreover, n contrast to nput-output models, CGE models are bult on fundamental mcroeconomc prncples and nclude nonlnear feedback mechansms. Ths permts the analyss of all markets smultaneously and the modelng of complex nter-sectoral and nter-regonal nterdependences wthout the constrant of lnearty or the problems nvolved n modelng d erent markets separately from each other (Shoven and Whalley 1992). Instead, the CGE model can analyse varous nterrelated markets and examne complex market-based nteractons between d erent sectors and economc actors. It ncorporates fundamental general equlbrum lnkages between producton, ncomes, and the pattern of demand. Gven that t endogenzes both prces as well as quanttes, 3 a mutual determnaton of market outcomes, prces and quanttes, n many nterrelated markets or sectors of a regon s explctly emphaszed (Bandara 1991). Unfortunately for regonal development planners, the majorty of CGE models that have been bult to date have been appled solely at the natonal level (Treyz 1993). There are several reasons that may explan ths general neglect of regonal detals n CGE applcatons. One reason s that regonal CGE models requre consderably more regonal economc data than comparable alternatve modelng approaches. In order to construct a CGE model for a regonal economy, one needs nformaton about commodty and factor prces, the quanttes of traded and non-traded goods and servces, the avalablty of factors of producton, the regonal and sectoral tax structure, exports and mports, captal ows, and factor movements, whch are often not avalable at the regonal level (Mansur and Whalley 1984). They also requre nu- 2 Much of the research assocated wth CGE models has nvolved the development of hghly sophstcated algorthms sutable for the soluton of the necessarly complex models used to represent the nterdependences between markets. CGE models vary today from a closed sngle regon to comprehensve open mult-regonal models, from relatvely smple forms wth only a few equatons to models as comprehensve as the regonal socal accountng matrx, from lnear to nonlnear usng neoclasscal consumpton and producton functons (Scarf and Shoven 1984). 3 The producton functons are used to represent producton lnks that exst among the varous ndustres or sectors of a regon and to ndcate the possblty of factor substtuton, economes of scale, and productvty ncreases. Smlarly, consumpton functons are used to allow substtuton among the goods. 4

5 merous parameters, ncludng supply, demand, and substtuton elastctes for each sector and each regon, behavoral parameters for regonal consumpton and savng functons (Mansur and Whalley 1984). Another reason for the relatve neglect of the general equlbrum approach n regonal applcatons s that the CGE framework s often too expensve and unweldy when appled to small regonal economes wth only margnal changes n the economc structure (Anseln 1990). In order to compensate for ths drawback of hgher complexty, regonal CGE models frequently nvolve only a lmted number of regons and hghly aggregated sectors of producton. 4 Most often they are appled to polcy analyss, whch deals wth more macro-orented questons such as potental changes n prces, wages, and nterest rates, because they requre less sector-spec c nformaton (Spencer 1988). The present study attempts to employ the advantages of the general equlbrum framework, such as, the theoretcal consstency and mcroeconomc foundaton, n an emprcal applcaton to regonal economes. Moreover, ths study s one of the rst systematc attempts to adopt and emprcally mplement an nter-regonal CGE model to modellng regonal economes of the CEE transton economes, where the paucty of regonal economc data s partcularly actual. 3 Theoretcal Framework: An Inter-regonal CGE Model In ths secton we present the theoretcal framework, whch we employ n the emprcal analyss. Lke most CGE models, our nter-regonal CGE model s bult on the assumpton of perfect competton n the economy n whch the economc agents, that s, the consumers, producers, and government agents, act as maxmzers of ther utlty, pro t, or votes. Gven that we assume that the economc agents maxmze ther welfare subject to a gven set of constrants (costs, ncome, or voter satsfacton), each of the economc actors n the economy s represented by an objectve functon and the correspondng resource, budget, or ncome constrant. Ths allows us to account for varous substtuton possbltes of both nputs n the producton processes and consumer goods n nal demand. The model s formulated n a way that t explctly takes nto account nter-regonal lnkages that exst between d erent sectors and ndustres wthn and outsde the regon. 4 CGE models often tend to deal wth hghly aggregated ndustral sectors and thus are not really sutable for ndvdual sector analyss. Instead they are assocated mostly wth e cency questons and neoclasscal welfare analyss. Ther sze and complexty means that they have huge and detaled data requrements and thus are expensve to mantan and keep up to date, whch further reduces ther exblty (ervs et al. 1982). 5

6 3.1 Producton Structure We start wth descrbng the producton structure of our model. Two numercally equvalent methods of appled producton analyss can be used n the general equlbrum analyss - the prmal approach and the dual approach (Shoven and Whalley 1992). In our regonal CGE model, we employ the dual approach of an ndrect pro t functon nstead of specfyng the behavoral functon as a maxmzaton problem n terms of prces and quanttes under techncal constrants. Accordng to Chambers (1988), the ndrect pro t functon can be seen as a mathematcal representaton of the soluton to the pro t maxmzaton problem: P S = max P S Q S Q S where P S s the maxmzed pro t attanable at the gven vector of n output and nput prces P S, and Q S s a vector of the correspondng n nput and output quanttes. In order to be able to mpose the general restrctons of the classcal producton theory nto the producton system of our model, a exble form of the pro t functon - the Symmetrc Generalzed McFadden pro t functon - s adopted n the present study. 5 The Symmetrc Generalzed McFadden pro t functon s de ned as follows: 6 P S = X S P S P Pj S jp S P S j P S P S where S s a vector of predetermned parameters; S s a vector of parameters to be spec ed; and S j s an n m matrx of whch (n 2 + n) =2 parameters have to be determned and n (n 1) =2 parameters result from Young s theorem. 7 Parameter restrctons are enforced n order to ensure that the pro t functon s lnearly homogeneous n prces, nput demands are homogeneous of degree zero n prces, and symmetry holds. One advantage of ths functonal form s that t s homogeneous of degree one, rrespectve of the choce of parameters. Another mportant advantage of the Symmetrc Generalzed McFadden pro t functon n the context of the present study s that once convexty s mposed, the pro t functon s stll exble and the curvature condton s mantaned for any vector of prces, P S. 8 5 It was proposed by ewert and Wales (1988) n terms of a cost functon based on developments by McFadden (1978) and Lau (1978). 6 Throughout the paper small Greek letters denote parameters and Latn letters refer to varables. The superscrpt S refers to supply sde and the superscrpt refers to demand sde. 7 In mathematcs, the symmetry of second dervatves refers to the possblty of nterchangng the order of takng partal dervatves of a functon. In economc lterature t s known as Young s theorem. 8 In the lterature, ths property s usually referred to as global convexty. (1) 6

7 As usual, we apply Hotellng s lemma n order to derve the output supply and nput demand functons of good : P P P Q S P S = S j + S jpj S P S k j S kj P k SP j S j S j P j S 2 P 2 (2) j S j P j S The aggregate output, Q S, of the supply functon (2) s the composte product sold ether on the regonal markets or exported. Products sold on the regonal markets, Q S r, and those sold on the export markets (outsde the regon), Q S o, are assumed to be produced wth constant elastcty of transformaton S. 9 Thus, we assume that the representatve producer produces a composte good, Q S, whch combnes locally sold and exported components accordng to a constant elastcty of transformaton (CET) functon à la Armngton (1969). The Armngton assumpton allows for both mport and export ows n each sector a fact, whch has been ncreasngly observed n Latva s foregn trade data. As usual, the CET aggregaton functon wth the scalng parameter, a S, the share parameter, S, and the elastcty of transformaton S ( S = 1 ) has two arguments: the quantty sold locally, Q S r, and the quantty sold outsde the regon, Q S S 1 o: Q S = a S h S Q S r S + 1 S Q S o S 1 S (3) Pro t maxmzaton under the producton functon constrant yelds the rato of quantty of good sold on local markets and the quantty exported, Q S r and Q S o, dependng on ther relatve prces, Pr S and Po: S Q S r Q S o 1 = S S Pr S 1 S Po S Gven local and foregn prces and quanttes and knowng the elastcty of transformaton S, we can calculate the remanng parameters a S, S and S. 3.2 emand Structure Smlar to the producton sde, several alternatve formulatons can be used to represent household demand behavor wthn the general equlbrum framework. The partcular choce of the demand system that re ects consumpton behavor n our model s drven by two factors. Frst, we prefer to work wth a theoretcally consstent demand system, whch permts the mposton of the general restrctons stated 9 Subscrpts, j and k refer to goods (sectors), where denotes output and j denotes nput. Subscrpts o and r refer to regons, where r denotes the partcular domestc regon and o denotes the out-of-regon markets. 1 (4) 7

8 by the classcal demand theory (Lau 1978). 10 The second ssue concerns the emprcal mplementaton of the demand system. Accordng to the dualty theory, the utlty maxmzaton under a budget constrant s equvalent to expendture mnmzaton gven a certan level of utlty (eaton and Muellbauer 1996). Generally, the expendture functon represents a soluton to the expendture mnmzaton problem gven a certan utlty level: E U; P = mn P Q Q s:t:u = f Q where E U; P s the mnmzed expendture gven the vector of prces P facng a representatve consumer; Q s the vector of quanttes consumed by the representatve consumer; and U s the utlty level of the representatve consumer. In lght of these two consderatons, our choce of the expendture functon s form s comparable to that on the producton sde of the model. In partcular, we employ the Normalzed Quadratc expendture functon proposed by ewert and Wales (1988) for representng the household demand behavor. The Normalzed Quadratc expendture functon s gven by: P E U; P = X a P + U X P + U 2 Pj jp Pj P P where U denotes the level of utlty, P s the normalsed consumer prce of the consumed good ; s a vector of predetermned parameters. The remanng three sets of parameters, a and whch are vectors of parameters, and j whch s a matrx of parameters, need to be spec ed. We mpose the followng parameter restrctons to equaton (5): P P = 0, P a P = 1, a 1 0, j = j and P j j = 0. Gven these restrctons, the Normalzed Quadratc expendture functon s determned through 2n 1+n(n 1)=2 lnearly ndependent parameters, whch have to be spec ed. As usual, we apply Shephard s lemma and substtute out the unobservable utlty, U, n order to obtan a system of uncompensated household demand functons of good : Q P = a + + P j j P j + P j j P j P j j P j P k P j kj P k P j 2 P j j P j P Pk j kj P k P j 2 ( P j j P j ) 2 = 1 P j j P j (5) (6) 10 These restrctons are () addng-up: value of total demands equals total expendture, () homogenety: demands are homogeneous of degree zero n total expendture and prces, () symmetry: cross-prce dervatves of the Hcksan demands are symmetrc, and (v) negatvty: drect substtuton e ects are negatve for the Hcksan demands. 8

9 The representatve consumer consumes a composte good, Q. The composte demand, Q, s bought ether on the regonal market, r, or s mported from outsde the regon, o. Regonally produced goods, Q r, and mported goods, Q o, are assumed to be mperfect substtutes accordng to Armngton (1969). 11 The regonal and mported components, Q r and Q o, are combned accordng to a CES functon. The CES aggregaton functon wth a scalng parameter, a, a share parameter,, and an elastcty of substtuton S ( = 1 ) takes the same form as equaton (3) on 1 the supply sde: Q = a h Q r + 1 Q o 1 (7) As usual, utlty maxmzaton under the budget constrant allows us to derve the rato of Q r and Q o dependng on ther relatve prces, Pr and Po : Q r Q o 1 = Pr 1 Po Analogously to the supply sde, gven prces and quanttes of locally produced and mported goods, f we know the elastcty of substtuton, we can calculate the remanng parameters a, and. 3.3 Government emand Next, we specfy the government demand of goods and servces. Government demand s assumed to be exogenous to our model (both the level and the basket of goods and servces purchased). Ths assumpton s drven by two consderatons. On the one hand, a government demand basket that s kept constant allows us to better compare d erent polcy scenaros n terms of government budget expendture, whch s a more convenent method than lookng for quantty adjustments. On the other hand, we expect the demand elastcty of government to be farly low, suggestng that changes n prces for consumpton goods wll result n only margnal changes n government demand behavor. However, prces are endogenous n our model, and, hence, government expendture vares wth prce adjustments due to polcy changes. Smlarly to the household commodty demand, mported and regonally produced goods are mperfect substtutes n meetng the composte government demand. As above, the exogenous government 11 The Armngton assumpton mples that mperfect substtutes can have d erent prces n d erent countres. In the context of our study a major modellng advantage of the Armngton assumpton s that t permts prces of mmoble nput factors to d er across regons. If markets are compettve, as n our model, then d erences n nput prces lead to d erences n output prces, and the Armngton assumpton provdes an ntutve explanaton of why consumers do not buy output goods exclusvely from the regon wth the lowest prce (Armngton 1969). 1 (8) 9

10 commodty demand, Q G, from the two sources (regonal and mported) s aggregated accordng to the CES functon à la Armngton: Q G o Q G = a h Q G r + 1 where Q G r s government demand of mported goods. Q G 1 o (9) s government demand of regonally produced goods and servces and Agan, we calculate the quanttes of locally produced and mported goods demanded by regonal government by solvng the CES functon for the rst order condton. Ths yelds the demand rato of Q G r and Q G o, whch depends only on ther relatve prces: Q G r Q G o 1 = Pr G 1 Po G where Pr G s prce for regonally produced goods and servces and Po G for mported goods n the government demand. 1 (10) s prce 3.4 Investment emand emand for nvestment goods s modelled n a smlar way to the government demand, mplyng that the quantty demanded by nvestors s exogenous to our model. However, gven that the prce for nvestment goods s endogenous, the expendture for captal formaton vares wth prce changes. Smlarly to government and household commodty demand, mported and regonal nvestment goods are assumed to be mperfect substtutes n meetng the composte nvestment good demand. The aggregate demand for captal goods, Q I, from the two sources, regonal and mported, s gven by the same CES functon à la Armngton as above: Q I o Q I = a h Q I r + 1 Q I 1 o (11) where Q I r s quantty of nvestment goods demanded from regonal sources and s quantty of nvestment goods demanded from out of regon producers. As usual, by solvng the CES substtuton functon for the rst order condton we derve the rato of regonal and mport demand for nvestment goods as a functon of relatve prces: Q I r Q I o 1 = Pr I 1 Po I where Pr I s prce for regonally produced goods and servces and Po I s prce for mported nvestment goods. As above, gven prces and quanttes and knowng the elastcty of substtuton, we can calculate parameters a, and (12)

11 3.5 Inter-regonal Market Equlbrum In ths secton we de ne the market equlbrum condtons. Two ssues need to be consdered, when selectng the partcular model closure rules: 12 model solvablty and tme horzon of the study. Frst, gven the complextes ntroduced by endogenous prces n a mult-regonal and mult-sectoral general equlbrum settng, solvablty consderaton must be gven to denttes that close the model and lead to a soluton (Rckman 1992). Even f the model s solvable analytcally, t mght not be solvable when mplemented emprcally. For our study ths mght mply that we need to reduce the channels of adjustments between d erent equlbrum states, f the emprcal model turns out to be unsolvable. Second, closure rules provde a great exblty n that they allow the model to re ect a varety of stuatons rangng from very shortterm to ntermedate and long-term perspectves and from hghly closed to relatvely open markets wth respect to trade, ncludng trade n goods, labor, and captal, wth other regons (Harrgan and McGregor 1989). Ths suggests that model s equlbrum condtons need to be selected n the context of tme horzon of polcy nstruments to be smulated. In lght of these two consderatons, rst we de ne the commodty market equlbrum. Gven that n our model the aggregate commodty demand s a sum of ntermedate demand, nsttutonal demand, and export demand, and the aggregate commodty supply s a sum of regonal producton and mports, the commodty market equlbrum s gven by the followng dentty: Q S r Pr S + Q o Po + Q G P G + Q I Po I {z } = Q r Pr + Q G r Pr G + Q I r Pr I + Q S o Po S {z } regonal supply = demand f or local goods where Q S r denotes regonal producton, Q o, Q G o and Q I o capture consumer, government and nvestment mports respectvely. On the left hand sde the rst three terms, Q r, Q G r and Q I r are consumer, government and nvestment demand for locally produced goods respectvely, and Q S o s foregn (out of regon) demand for locally produced goods. As above, the assocated prces are denoted by P. Next we de ne factor market equlbrum condtons. Gven the consderable tme lags between the mplementaton of structural polcy measures untl takng of effect through the regonal economes, we study factors market behavor from a longrun perspectve. Ths mples the moblty of captal and labor between sectors and regons. Captal and labor are thus exposed to market forces that work toward nter-sectoral and nter-regonal equalzaton of wages and rents and an nter-regonal quantty-prce equlbrum of labor markets. Hgher wage rates and captal rents n 12 Model closure contans a set of assumptons whch partton model varables nto exogenous and endogenous varables such that the number of equatons and unknown endogenous varables are solvable. Beyond ths techncal descrpton, the spec caton of whch varables to hold exogenous also mples some underlyng behavoural hypothess beyond the core mechansms of the model. 11

12 the regon relatve to out-of-regon wages and rents encourage labor and captal mmgraton whle lower wages and rents nduce emgraton. As usual, factor markets are assumed to be perfectly compettve. In order to account for potental nter-regonal labor mgraton, we assume that labor mgraton, M or, from regon o to regon r s drven by d erences n the nomnal wage rate between regons: M or = L S Wo o log L or (13) where L S o s the ntal labor supply n regon o, W o s out of the regon wage rate, W r s wage rate n regon r, and L or s labor mgraton elastcty from regon o to regon r. Accordng to equaton (13), mgraton ows, M or, from regon o to regon r are ncreasng n nter-regonal wage d erences (log W o log W r ) as they are n labor mgraton elastcty, L. Labor market s n equlbrum when the quantty of labor suppled equals the quantty of labor demanded n all actvtes n each regon. L S r + M or = L r W r + L G r (14) where L r = P L r s the ndustry aggregate labor demand n regon r and L G r s exogenously xed labor demand by government agences. The aggregate household ncome n regon r from supplyng labor to producng rms and government agences s sum of the product of labor demanded and the wage rate n regon r: Yr L = W r L r + L G r where both regonal wage rate, W r, and ndustry labor demand, L r, are endogenous varables. The aggregate labor demand, L r, n regon r s derved from pro t maxmsaton subject to a vector of nput and output prces. 13 Regonal wage rate, W r, s determned by net labor supply and aggregate labor demand n regon r. The net (dsposable) labor ncome, Yr LN, s then calculated by subtractng the payroll tax from the gross labor ncome, Y L Y LN r r : = Yr L 1 t L r where t L r s the labor payroll tax rate n regon r. Captal demand has already been determned above. Captal supply n each regon s determned endogenously by adjustng the regonal captal stock such that the real rate of return to captal equals the steady state rate of return: Rr P r = o, where R r s captal rental rate, P r s the prce level n regon r and r s the steady state rate 13 For sake of smplcty, we assume that all factor rewards ow to the same regon, where prmary factors are employed n. (15) (16) 12

13 of return. Household ncome from captal rental s determned n the same way as ncome from labor supply. Land s regon-spec c and mmoble between regons. Moreover, the regonal endowment wth land s gven exogenously, whch captures the comparatve advantages of regons. Land market equlbrum s acheved through adjustments of rents, mplyng that land s fully employed n each regon at the equlbrum. In equlbrum land supply equals aggregate demand for land. 4 Model ata Base: An Inter-regonal Socal Accountng Matrx The model data base s organsed n form of a Socal Accountng Matrx (SAM). Accordng to Pyatt and Round (1985), SAM s an accountng tool whch gves a complete, consstent, and comprehensve snapshot of nteractons between varous actors, sectors and agents n an economy at a certan pont n tme. Lke an nputoutput table, each account s represented by both a row and a column where a matrx entry, r j, represents an expendture tem of account j and an ncome recept of account. Whle an nput-output table contans nformaton only on nteractons wthn the producton cycle of the economy, a socal accountng matrx extends the focus to the full crcular ow of ncome wthn the economy. In addton to the producton accounts, a typcal SAM contans also factor, household, government, captal, and rest of the world accounts (Pyatt and Round 1985). Usually, socal accountng matrces have been constructed for three reasons: () for reconclng d erent but overlappng sources of data wthn a consstent framework (Hewngs and Maden 1995); () as a descrptve mechansm for mpartng nformaton on the structure of an economy and the relatve mportance of nteractons that take place (Pyatt and Round 1985); and () for parameterzng appled partal and general equlbrum models (Mansur and Whalley 1984). The latter advantage of SAM conssts the man drvng force for buldng an nter-regonal SAM for the present study. Whle the constructon of natonal SAMs has become a common-place n recent years, examples of nter-regonal SAMs, especally for developng and transton economes, are stll relatvely few. Round (1988), as a poneer among these few examples, constructed a regonal SAM for Malaysa and llustrated how the data paucty ssues can be dealt wth n a consstent way. Followng ths example, we develop an nter-regonal SAM for Latva focusng on nter-regonal transactons between sectors, actors and economc agents. Followng the approach of Round (1988), the constructon of the nter-regonal SAM for Latva was carred out n three stages. Frst, a natonal SAM was bult for Latva as a whole. The prncpal sources of data used durng ths stage of the constructon process were the use, make, and mport matrxes from the

14 Latvan nput-output tables, data from the Industral and Agrcultural Census, and, nally, household expendture patterns and ncome sources from the Household Expendture Survey (HES). ata for the rest of the world account were ntegrated from the Global Trade Analyss Project (GTAP), verson 5 (maranan and Mcougall 2000) nto the SAM. 14 As a result, the obtaned SAM conssts of four producton actvtes, four correspondng commodtes, three factors of producton, and three economc agents: one representatve producer, one aggregate household, one government, and one aggregate exogenous account. Subsequently, followng the approach of Comer and Jackson (1997), the Latva SAM was splt nto two regons representng rural and urban Latva, respectvely. The de nton of rural and urban regons n Latva was drven by emprcal rather than conceptual crtera. e nng the rural Latva as beng an aggregate of four regons other than the dstrct of the cty of Rga allowed a relatvely straghtforward use of statstcs collected at a regonal level, thus mtgatng the need for extensve prmary survey work. Statstcal data for the regonal employment structures from the Census of Employment, whch provdes regonal-level sectoral employment data, and regonal household composton data from the 1998 Populaton Census was used to splt ndustry sub-matrces of the natonal SAM, whle mantanng the overall control totals for Latva as a whole. The nal stage of the SAM constructon process was the estmaton of trade and ncome ows between regons for whch lttle data or no data were avalable. For ths reason, survey data was used from a survey carred out n 1997 on tradng patterns between rms n Latva. In conjuncton wth nformaton on the relatve mportance of d erent producton sectors n rural and urban Latva, estmates were made about commodty ows between regons. An mplct assumpton wthn these calculatons was the assumpton that output levels from sectors n each regon were proportonal to employment levels. Moreover, no attempt was made to re ect possble d erences n producton technologes and nput demand between rural and urban rms wthn a sector. These are both areas n whch future research could substantally mprove the accuracy of nformaton contaned n the nter-regonal SAM for Latva. The obtaned SAM for Latva s reported n aggregate form n Table 1 n Appendx. Table 1 ndcates not only the economc weght of the rural and urban regons n Latva but also the sgn cance of d erent types of ows wthn and between the two regons n Accordng to Table 1, rms n rural Latva have produced LVL bllon output n the base year, whereas the aggregate output of urban rms amounts to LVL bllon. These numbers suggest that, although, two-thrds of Latva s populaton resde n rural areas, only 41 percent of the total output s produced n the rural regons. In contrast, fty-nne percent of Latva s GP s produced by only 14 The GTAP data base verson 5 corresponds to the global economy n 1997 (maranan and Mcougall 2000). In addton to GP, nput-output and trade data, t also contans tar coverage for agrculture and manufactures and agrcultural support. Ths nformaton was used to aggregate the exogenous rest of the world account and embeddng t nto the Latva SAM. 14

15 one-thrd of the total labor force, that s, those resdng n the urban regon. 5 Assessng Regonal Impacts of Alternatve Polces In ths secton we emprcally mplement the theoretcal nter-regonal general equlbrum model, whch we presented n secton 3. We start wth presentng the three polcy scenaros and the model base run assumptons, whch we assess n the followng sectons. Next, we solve the nter-regonal CGE model for the long-run nter-regonal general equlbrum under d erent sets of assumpton (scenaros). Ths allows us to assess mpacts of changes n rural development polces n the context of enlargement of the European Unon. 5.1 Polcy Scenaros In the context of European ntegraton and the world-wde globalsaton processes, we dentfy three possble economc development scenaros for Latva: the status quo, ntegraton nto the European Unon, and ntegraton nto world markets. The three polcy scenaro can be bre y summarsed as follows: n the base run there are no polcy changes, n the EU ntegraton scenaro all natonal polces are replaced by the EU Structural Fund and CAP support measures, and n the global lberalzaton scenaro all subsdes and other support measures (both natonal and EU) are removed. Base run: the counterfactual. Base run serves as a reference scenaro, where we assume that all sectoral, rural development and agrcultural polces observed n the base year 1998 do not change untl In other words, all market support measures, such as drect subsdes, nput subsdes, and per-farm subsdes, are kept at ther 1998 levels per unt of output. The nomnal rates of market protecton, de ned as the polcy-nduced percentage gaps between output and border prces, are assumed to be those observed for 1998 and are unform for all Latvan regons. Changes n mport (and export) prces between 1998 and 2007 are exogenous and are based on the OEC (2000) world market prce projectons. Assumptons about autonomous techncal progress are derved from the European Commsson (1999). Accordng to European Commsson (1999), the annual growth rates of techncal progress are n the range of 1 percent to 3 percent n the CEE accesson countres. In the present study we assume a constant growth rate of 2% per annum. Populaton (labor supply) growth and ncome growth rates are based on the OEC (2000) projectons. All these assumptons are subject to extensve senstvty analyss. EU ntegraton scenaro: SAPAR polces (EU). 15

16 Obvously, the EU ntegraton scenaro s of partcular nterest n the present study. In the EU ntegraton scenaro Latva s assumed to become fully ntegrated nto the EU, but remans protected from the world markets. In the EU ntegraton scenaro t s assumed that by 2007 Latva, as a full EU member, wll already have mplemented the SAPAR market regulatons as reformed by the Agenda 2000 decsons of the European Councl (European Commsson 1999) and that economc adjustments to these polcy changes wll be completed. All natonal polces from the base run are replaced by the EU Structural Fund and CAP support measures. The remanng assumptons on border prces, techncal progress, ncome, and populaton growth of the BR are mantaned n the EU ntegraton scenaro. Ths allows us to examne only those accesson mpacts whch can be attrbuted to changes n regonal and rural development polces. In order to be able to compare the mpact of rural development pre-accesson programs proposed n the Rural evelopment Plan for European Communty Support for Pre-Accesson Measures n Agrculture and Rural evelopment (SAPAR) wth alternatve nsttutonal settngs n Latva, we contrast the EU ntegraton scenaro wth two alternatve polcy scenaros - the base run wth no changes n the current polcy settng and global market lberalzaton. Lberalzaton scenaro: global market lberalzaton (LS). The globalsaton scenaro serves as a second pont of reference to the EU ntegraton scenaro. In the global market lberalzaton (globalsaton) scenaro Latva s assumed to be fully ntegrated nto the world economy. All market protecton s dsmantled both wth respect to the EU and to the rest of the world. More precsely, n the LS scenaro all sectoral and border protecton measures are completely abolshed, that s, the nomnal rates of protecton are set to the level of world average. We use the GTAP verson 5 tar data for agrculture and manufactures (maranan and Mcougall 2000) to obtan the average protecton rates for the world economy. The remanng assumptons regardng the technologcal progress, ncome, and populaton growth of the BR are mantaned n the lberalzaton scenaro. In the followng sectons we evaluate soco-economc mpacts of these three polcy scenaros accordng to three crtera: the gross regonal product (GRP), the household ncome proxyng the regonal household welfare and the regonal prce ndex. Frst, the EU ntegraton scenaro under full applcaton of the acqus communautare s compared to the base-run scenaro of unchanged contnuaton of the current rural development polces. Second, a scenaro of a global lberalzaton of regonal and agrcultural markets serves as a second pont of reference, to whch the EU accesson scenaro s contrasted. 16

17 5.2 Gross Regonal Product We start the scenaro analyss wth polcy mpacts on gross regonal product. Gross regonal product (GRP) s a comprehensve and one of the most often used measures capturng mpacts of polcy changes n regonal CGE models. In addton to changes n output quantty, GRP also accounts for the usage of prmary nputs and compensaton to each nput employed n the producton process. For example, t ncludes the ndrect busness tax pad by ndustry, the total compensaton for labor by ndustry, ncludng payroll taxes and employee bene ts, and the gross return to captal, ncludng pro ts, before deprecaton. The obtaned smulaton results are reported n Fgure Rural Urban BR 2007 LS 2007 EU Fgure 1: Gross regonal product (GRP), Mo LVL Accordng to the smulaton results reported n Fgure 1, polcy-nduced changes n the regonal welfare measured n terms of GRP, are surprsngly small n the rural economy, and, at the same tme, consderable n the urban regon. In all three scenaros, rural GRP growth rates d er only slghtly. The most moderate growth rate s smulated for the market lberalzaton scenaro, where the annual GRP growth rate was calculated at 0.71 percent (see 2007 LS n n Fgure 1). Assumng that the obtaned soluton s a medum-term equlbrum rather than a long-run equlbrum, these results are n lne wth predctons of the neoclasscal trade theory, whch suggests that a reducton n external protecton mght decrease the compettveness of local producers n the short to medum run. In the long run, however, the results mght be rather d erent, f allocaton and growth e ects were accounted for. The second concluson whch we can draw when consderng smulaton results reported n Fgure 1 s that the rural GP would grow most rapdly under the EUntegraton scenaro wth the GRP growth of about 1.9 percent per year (see 2007 EU 17

18 n n Fgure 1). The other two scenaros, the base-run and the market-lberalzaton scenaros (2007 BR and 2007 LS), suggest almost equal growth rates for Latva s rural economy n the reported perod. The presented smulaton results should, however, be treated wth cauton. One serous drawback of GRP as a regonal welfare measure s that GRP accounts only payments to resources employed n the regon rrespectve of where resource owners resde. In realty, however, factor payments mght ow to resource owners resdng both wthn the regon and outsde the regon. For example, f land owners lve n the urban regon, then rural welfare gans mght be even smaller that those suggested by n Fgure 1. In lght of these lmtatons, we am to ncrease the robustness of these smulaton results by performng welfare calculatons usng an alternatve welfare measure - household ncome. The results obtaned by comparng polcy-nduced changes n regonal household ncomes are presented next. 5.3 Household Income Household ncome s another wdely used measure of household welfare. 15 Most household ncome stems from factor, especally labor, renumeraton. Other sources of household ncome nclude nter-household transfers, government transfers, and net remttances from outsde the regon. The gross household ncome, Yr H, n regon r s calculated by addng up all these sources of ncome: Y Hr = Y L r + Y C r + Y T r + H r + T H r (17) where Yr L s gross labor ncome n regon r, Yr C s ncome from captal rents, Yr T s ncome from land rents n regon r, H r are rm pro ts n regon r, T Hr are government transfers to households n regon r and net transfers and remttances to households from outsde the regon, such as the EU farm ncome support measures. In contrast to factor payments, transfer payments do not depend on regonal resource ownershp and factor prces. Gven that gross factor payments are subject to government taxes and captal deprecaton, the total household earnngs are reduced by the applcable deductons avalable for the dstrbuton to households that own factors employed n the producton process. The dsposable (net) household ncome, Yr HN, n regon r s calculated smlar to the net labor ncome n equaton (16): Y HN r where t h r s household ncome tax rate n regon r. = Y H r (1 t hr ) (18) 15 Utlty measures for ndvduals and households are the result of preferences expressed through markets. Movng from one market result to another presumes a welfare change for households n most, f not all, regons. To measure ths change from a polcy or program change, welfare must be measurable. Because utlty s not drectly measurable, alternatve measures must be chosen. 18

19 Smulaton results for the rural household net ncome are reported n Fgure 2. Comparng Fgures 1 and 2 we may conclude that changes n regonal household ncome are consstent to those of GRP. These results ndcate that the results are robust wth respect to d erent welfare measures. The results reported n Fgure 2 also suggest that the total mpact on consumer welfare, whch here s measured by household ncome, s relatvely small. In the lberalzaton scenaro the rural household ncome wll grow at a yearly rate of 1.6 percent between 1998 and 2007 (see 2007 LS n Fgure 2). The base run and the EU ntegraton scenaros (2007 BR and 2007 EU), where the average growth rate of household ncome wll be about 2 percent yearly, seem to be slghtly more favorable to rural households Rural Urban BR 2007 LS 2007 EU Fgure 2: Rural and urban household ncome, Mo LVL As above, these results should be nterpreted cautously. For nstance, gven that n our model resource ownershp and ncome from transfers are held constant across scenaros, the regonal welfare measure of household ncome does not re ect changes n resource ownershp or changes n transfers to rural households. 5.4 Regonal Prce Level In ths secton we report changes n the regonal prce level under each scenaro. We calculate the regonal prce ndex for two purposes: () t can be used as an nverse measure the competton between supplyng rms n the regon; and () t serves as an ndrect measure of consumer welfare. The lower s the regonal prce ndex, ths hgher s rm competton for sales n the regon and the hgher s net ncome of households. Gven that the composte commodty prces are endogenous n our model, the prce growth n monetary terms may be caused by quantty changes and/or prce 19

20 adjustments to external (polcy) shock. The extent to whch these changes actually take place s determned by the elastcty of substtuton between locally produced and mported goods and servces. The overall regonal prce level may be calculated n two alternatve ways, ether as a weghted ndex of the composte commodty prces or as regonal output prces. We calculated several prce measures n order to obtan a robust and consstent pcture of the general prce level n the regon compared to the outsde-the-regon prce level: the net output prce, the tradable commodty output prce, and the commodty purchase (consumer) prce. Gven that all three prce measures yeld smlar results, we report only the polcy mpact on output prces. Fgure 3 reports producer prces for the three polcy scenaros relatve to the base year (1998) BR 2007 LS 2007 EU AGR FOO MAN LOCAL Fgure 3: Changes n output prces n rural regon n percent (1998=100) Accordng to Fgure 3, wth unchanged agrcultural and rural development polces, as assumed n the base run scenaro (2007 BR n Fgure 3), output prces mght develop more favorably for the food processng ndustry from 1998 to 2007 (see FOO n Fgure 3). These results are manly drven by the assumpton that the expendture share for demand of processed food compared to demand for raw agrcultural products wll contnue to grow untl The szable ncrease n relatve prces for manufactured food products s also drven by the fact, that food processng ndustry n CEE s not elgble for the nancal support from the European Structural Funds and CAP, mplyng that any ncrease n producton costs are forwarded to consumers n terms of hgher output prces. In contrast, accordng to our smulaton results, relatve output prces would decrease for agrcultural, manufacturng and non-tradable goods and servces (see AGR, MAN and LOCAL n Fgure 3). In the base run scenaro, because nomnal protecton 20

21 rates are assumed to stay at ther 1998 levels, the regonal output prces for tradable goods would change at the same rate as the projected world market prces between 1998 and The sharp ncrease n the producer prce for non-tradable goods under the EU scenaro stems from the nancal support of the European Structural Funds to locally produced and locally sold goods and servces, such as rural toursm (see LOCAL n Fgure 3). Generally, we may conclude that under ether scenaro the food processng ndustry would gan most compared to agrculture, manufacturng, as well as locally produced and consumed goods and servces. These results are new and have not been reported n the lterature before. However, they are n lne wth macroeconomc studes (e.g. OEC 2000), whch pont to the growng role of tertary ndustres n the CEE accesson countres. 6 Conclusons In ths artcle we have demonstrated that an nter-regonal CGE model can be a useful and relable tool for analyzng regonal mpacts of changes n global economc condtons and for assessng both nter-regonal and nter-sectoral mplcatons of potental polcy changes even f computatonal resources are lmted and a full range of regonal economc data requred by a formal CGE analyss are not avalable. In our emprcal analyss we found that the rural economy of Latva, whch was used as an example for the Central and East European accesson countres, mght gan most from ntegraton nto the EU, f the CAP and EU Structural Fund support measures are mplemented mmedately but markets are opened only gradually to foregn competton. These results are n lne wth the fact that market protecton n the EU s currently relatvely hgh compared to the CEE accesson countres. Gven that n the EU agrcultural markets are hghly protected, above all, rural households mght gan from ntegraton nto the EU, f the CAP support measures are equally adopted n the CEE accesson countres, as we assumed n our smulatons. Although we have been able to account for several mportant components and adjustment channels of the rural economy n our nter-regonal CGE model, t s mportant to focus future modelng actvtes on dealng wth remanng lmtng assumptons and ncorporatng more externaltes that rural economes face. Above all, one should consder market mperfectons, transportaton costs, and a dynamc rather than statc treatment of regonal economes. If all ths s ensured, then rural development planners wll have a relable plannng tool for the ex-post as well as the ex-ante evaluaton of regonal and rural development polces. 21