Discussion Pape No. 2003/27 Efficiency Gains fom the Elimination of Global Restictions on Labou Mobility An Analysis using a Multiegional CGE Model Ana Maía Iegui * Mach 2003 Abstact This pape computes the woldwide efficiency gains fom the elimination of global estictions on labou mobility using a multiegional CGE model. A distinctive featue of ou analysis is the intoduction of a segmented labou maket, as two types of labou ae consideed: skilled and unskilled. Accoding to ou esults, the elimination of global estictions on the mobility of skilled and unskilled labou geneates woldwide efficiency gains that could be of consideable magnitude. When only skilled labou migates, the woldwide efficiency gains ae smalle, as this type of labou epesents a small faction of the labou foce in developing egions. Keywods: migation, applied CGE modelling, labou maket segmentation JEL classification: C68, F22, R13, R23 Copyight UNU/WIDER 2003 * Estudios Económicos, Banco de la República, Bogotá This study was pepaed fo the UNU/WIDER poject on Refugees, Intenational Migation and Povety which is co-diected by Geoge Bojas, Havad Univesity and Jeff Cisp, UNHCR. It was pesented at the UNU/WIDER development confeence on Povety, Intenational Migation and Asylum held on 27-28 Septembe 2002 in Helsinki, Finland. UNU/WIDER gatefully acknowledges the financial contibutions to its 2002-2003 eseach pogamme by the govenments of Denmak, Finland, Noway and Sweden.
Acknowledgements I would like to thank Chis Dawkins, Jesús Oteo, Jeff Round, and John Whalley fo helpful comments and suggestions. I have also benefited fom the comments of Jeffey Cisp, Geoge Bojas and the paticipants at the UNU/WIDER Confeence on Povety, Intenational Migation and Asylum held in Helsinki. The views expessed in the pape ae those of the autho and do not epesent those of the Boad of Diectos of the Banco de la República (Cental Bank of Colombia), o othe membes of its staff. The Wold Institute fo Development Economics Reseach (WIDER) was established by the United Nations Univesity (UNU) as its fist eseach and taining cente and stated wok in Helsinki, Finland in 1985. The Institute undetakes applied eseach and policy analysis on stuctual changes affecting the developing and tansitional economies, povides a foum fo the advocacy of policies leading to obust, equitable and envionmentally sustainable gowth, and pomotes capacity stengthening and taining in the field of economic and social policy making. Wok is caied out by staff eseaches and visiting scholas in Helsinki and though netwoks of collaboating scholas and institutions aound the wold. www.wide.unu.edu publications@wide.unu.edu UNU Wold Institute fo Development Economics Reseach (UNU/WIDER) Katajanokanlaitui 6 B, 00160 Helsinki, Finland Camea-eady typescipt pepaed by Liisa Roponen at UNU/WIDER Pinted at UNU/WIDER, Helsinki The views expessed in this publication ae those of the autho(s). Publication does not imply endosement by the Institute o the United Nations Univesity, no by the pogamme/poject sponsos, of any of the views expessed. ISSN 1609-5774 ISBN 92-9190-436-8 (pinted publication) ISBN 92-9190-437-6 (intenet publication)
1 Intoduction The classic economic agument in favou of labou migation is that people move in seach of highe wages, hence inceasing thei own poductivity.1 Howeve, as indicated by Layad et al. (1992), the decision to migate also depends upon othe economic, social and political consideations. Among the economic aspects, migants may take into account compaative wage levels (actual and expected); compaative unemployment ates and unemployment benefits; the availability of housing; and the cost of migation which includes tavel expenses, infomation costs, and the psychological cost of leaving fiends and family. Weyebock (1995) also indicates that political instability and civil wa may cause lage emigation flows than economic o demogaphic pessues. Recent empiical studies on intenational migation have mainly focused on US-Mexico migation pattens (Hill and Méndez 1984; Robinson et al. 1993; Levy and van Wijnbegen 1994), and migation flows fom Easten Euope and the fome Soviet Union into Westen Euope (Layad et al. 1992; Weyebock 1995). Hamilton and Whalley (1984) has been the only attempt to quantify the efficiency gains fom the emoval of global estictions on labou mobility. They use a patial equilibium famewok, in which the paametes of a CES poduction function ae estimated fo a seven-egion county classification. Then, the estimated paametes ae used to calculate the changes in labou allocation acoss egions afte the emoval of immigation contols. They assume that the woldwide labou supply is fixed, that full employment occus in all egions, and that diffeences in labou s maginal poduct acoss egions aise fom baies to inwad mobility of labou in high-wage counties. Hamilton and Whalley find lage efficiency gains fom the emoval of immigation contols; in most cases, these gains exceed wold wide, GNP geneated in the pesence of the contols. In addition, in labou-expoting egions wage ates ise and capital ownes ae made wose off; on the othe hand, in labou-eceiving egions wage ates fall and capital ownes ae made bette off. In this pape we compute the woldwide efficiency gains fom the elimination of estictions on labou mobility. In contast to Hamilton and Whalley (1984), we use a multiegional geneal equilibium model instead of a patial equilibium appoach, since the fome povides an ideal famewok to analyse the effects of policy changes on esouce allocation, the stuctue of distibution, and thus in economic welfae. A distinctive featue of ou analysis is that we conside a segmented labou maket (i.e., skilled and unskilled labou), which can be justified on the gounds that this facto is not homogeneous. The segmentation of the labou maket jointly with the geneal equilibium famewok allows us to examine the distibutional effects of migation between skilled and unskilled labou in each egion, and between these two and capital. Accoding to ou esults, the elimination of global estictions on labou mobility geneates woldwide efficiency gains that could be of consideable magnitude, anging fom 15 pe cent to 67 pe cent of wold GDP. With the intoduction of a segmented labou maket, welfae gains educe since the benefits and losses of migation ae not evenly distibuted within each county, anging fom 13 pe cent to 59 pe cent of wold 1 Layad et al. (1992) indicate that fee tade and intenational capital mobility can also aise poductivity, without labou migation. 1
GDP. And when only skilled labou migates, woldwide efficiency gains ae smalle anging fom 3 pe cent to 11 pe cent of wold GDP, since skilled labou epesents a small faction of the labou foce in developing egions. The pape poceeds as follows. Section 2 descibes the basic stuctue of ou multiegional geneal equilibium model. Section 3 contains the empiical implementation, including the desciption of the benchmak dataset and the calibation of the model. Section 4 pesents the esults of the model as well as the sensitivity analysis. Section 5 pesents model elaboations, including tansaction costs, intenational capital mobility, and selective mobility. Section 6 offes some concluding emaks. 2 The model In a wold economy chaacteized by counties with diffeent levels of income, individuals have incentives to migate to counties with highe wage ates. If labou wee allowed to move fom one county to anothe without estictions, it will do so until the maginal poduct of labou is the same in both low-income and high-income counties. Migation will educe the labou foce in the low-income county (souce egion), leading to an incease in wages,2 and a eduction in the demand fo labou. In addition, migation leads to a pocess of facto eallocation within the poo county: the emaining wokes gain though highe wages, but capital ownes lose since labou is now scace elative to capital. Convesely, in the high-income county (destination egion) the labou foce inceases, which leads to a eduction in the wage ate (assuming no igidities). This lowe wage will incease the demand fo labou and aggegate employment. Duing the tansition, wokes will lose though lowe wages and capital ownes will gain since labou is now less scace elative to capital (see e.g. Bhagwati et al. 1998; Layad et al. 1992). This analysis is based on the assumption that labou is a homogeneous facto of poduction, which implies that the benefits and loses of migation ae evenly distibuted within each county. Howeve, as ou analysis will show late on, this is not necessaily the case when thee ae many types of labou.3 2 The magnitude of the incease will depend on the elasticity of labou demand. The moe elastic the demand fo labou, the smalle the incease in wages. 3 Fo some tade theoists the issue of the emoval of estictions on labou mobility may not be of geat elevance because of the facto pice equalization theoem, accoding to which facto pices will be equalized by fee tade without intenationally mobile factos (see Samuelson 1948, 1949). Howeve, this theoem is based on vey estictive assumptions, such as identical technologies in diffeent counties, constant etuns to scale, pefect competition, no facto intensity evesals, no specialization, and that good pices ae equalized as a esult of tade. Moeove, facto pice equalization depends on the complete convegence of the pice of the goods. In eality, the pices of the goods ae not fully equalized because of both natual (e.g., tanspotation costs) and atificial baies to tade (e.g., impot taiffs, impot quotas, voluntay expot estaints). An additional eason why facto pice equalization may not be achieved is that counties exhibit diffeent technologies and esouces, so that they ae unlikely to emain unspecialized (see e.g. Layad and Waltes 1978; Kugman and Obstfeld 1994). 2
Figue 1 Poduction stuctue Expots to the othe 7 egions Expots Domestic consumption Output Labou Capital Skilled labou Unskilled labou The stuctue of the model follows the standad specification of a multiegional geneal equilibium model. The model is static, and consists of eight egions, each one with demand and poduction stuctues, linked though tade. Each egion contains one industy that poduces a single output, which is teated as heteogeneous acoss egions (Amington 1969). Thee is a epesentative consume in each egion and, fo simplicity, intemediate poduction is not consideed. Poduction in the model involves a CES value added function with capital (K) and labou (L) as pimay inputs. Thee ae two types of labou, skilled (Ls) and unskilled (Lu), and this labou maket segmentation is a distinctive featue of ou modelling execise in compaison to pevious wok by Hamilton and Whalley (1984). Figue 1 pesents the poduction stuctue of the model. The model uses two-stage CES poduction functions, which ae moe flexible since they allow us to have diffeent elasticity paametes in each stage of the poduction pocess. In the fist stage, Ls and Lu ae combined to poduce the aggegate labou input (L); that is, L = φ π Ls ( ς 1) / ς + (1 π ) Lu ( ς 1) / ς ( ς /( ς 1), = 1,...,8, whee L is the aggegate labou input used in egion ; Ls and Lu ae skilled and unskilled labou inputs in egion ; φ is a constant defining unit of measuement; π is a shae paamete; ς is the elasticity of substitution between skilled and unskilled labou in the poduction of the good in egion. Labou demand functions fo the two types of labou ae obtained fom cost minimization; that is, each industy selects an optimal level of Ls and Lu that minimizes the cost of poducing L units of the aggegate input. [1] 3
In the second stage the aggegate labou input and capital ae combined to poduce value added. In each egion the industy selects an optimal level of inputs that minimizes the cost of poducing value added. Futhe, the commodity poduced in each egion can be tansfomed eithe into a commodity sold on the domestic maket, o into an expot accoding to a constant elasticity of tansfomation (CET) function. Then, expots ae allocated acoss egions accoding to a CET function. Factos ae non-poduced commodities in fixed supply in each egion. Factos of poduction ae assumed to be intenationally immobile, although this assumption is elaxed late on fo L. Tuning to the demand side of the model, we assume that consumes within a egion have identical homothetic pefeences, which allows us to conside a epesentative consume, endowed with all the labou and capital in the egion. In this case, as thee is only one good, the egion s epesentative consume demands a composite of domestically poduced and impoted goods subject to the egion s budget constaint. Figue 2 pesents the demand stuctue of the model. Figue 2 Demand stuctue Utility Composite Domestic Impots Impots fom the othe 7 egions The budget constaint in each egion is given by income equal expenditue (I = E ). The egion s income is deived fom owneship of factos of poduction, govenment tansfes and the tade suplus (o deficit), that is: Ls I = PLu, Lu + PLs, + K, K + TR + P TB, [2] whee P Lu,, P Ls,, and P K, define the selling pices of the factos of poduction in egion ; Lu, Ls, and K coespond to the egion s endowment of unskilled labou, skilled labou, and capital, espectively; TR epesents tansfes fom the govenment; and TB coesponds to the egion s tade suplus (o deficit). On the othe hand, the egion s expenditue includes the amount spent on goods as well as taxes paid: E = P X + T, [3] 4
whee P and T coespond to the pice and taxes paid by the consume in egion, espectively. The model also incopoates tade and domestic tax policies. These include income, facto and consumption taxes, as well as impot taiffs, all of which ae modelled in ad valoem fom. All tax evenues aised ae assumed to be tansfeed back to consumes. Lastly, it is woth pointing out that some of the assumptions of the model may affect the outcome of the simulations. In global models it is usually assumed that capital is intenationally immobile. This assumption may not be vey ealistic since intenational capital makets ae becoming moe integated. Howeve, this assumption is fundamental to the stuctue of the model; if all factos of poduction ae allowed to move feely, the concept of egion is no longe clea. Hence the need fo a fixed facto in the specification of the model (in one of the extensions of the model, when capital is assumed to be intenationally mobile, unskilled labou is the fixed facto in the model).4 Regading labou, in the model it is assumed that diffeences in the maginal poduct of labou aise fom baies to inwad mobility of labou in high-wage counties. Thus, once baies to labou mobility ae eliminated, wage ates equalize acoss egions. The model also assumes that labou in one egion is the same as labou in anothe egion, so that diffeences in labou quality o human capital pe woke acoss counties ae ignoed. In the eal wold these diffeences ae not only pesent but may also be significant. Fo example, Lucas (1995) indicates that poduction pe woke in the US is about fifteen times what it is in India; afte coecting fo diffeences in human capital, each Ameican woke was estimated to be the equivalent of about five Indian wokes. Anothe impotant facto that may affect labou poductivity is the technology available in each egion. Thus, the elimination of estictions on labou mobility may not afte all eliminate diffeences in poductivity acoss egions. As can be seen, some of the assumptions used in the specification of the labou maket may be highly simplified; howeve, incopoating diffeences in the quality of labou acoss egions is seveely constained by data availability. Once the model has been specified, it can be solved fo an equilibium solution. Equilibium in the model is given by a set of goods and facto pices fo which all makets clea. That is, demand-supply equalities hold in each goods and factos makets; zeo pofit conditions hold fo each industy in each egion; and each egion is in extenal-secto balance (see Appendix A). Next we calculate the paametes of the model that ae consistent with the benchmak dataset; these paametes allow us to epoduce the dataset as an equilibium solution of the model. Then we compae countefactual equilibia with the benchmak equilibium geneated by the data. 3 Empiical implementation The model consists of eight egions, each of which engages in domestic and foeign tade activities. These egions wee chosen to eflect wold tade, and we use 1990 data fo the United States (USA), Japan (JAP), the Euopean Union (12-membe-EU), othe 4 Instead of having a fixed facto, a nontadable good could be intoduced, so that all poduction factos could be inte-egionally mobile. 5
development counties (ODC), developing Ameica (DAM), developing Afica (DAF), developing Asia (DAS), and developing Euope (DE).5 Appendix B pesents the gouping of individual counties. We assume that each egion poduces one commodity, and that each egion s domestically poduced and impoted goods ae qualitatively diffeent (Amington 1969). We conside one commodity as ou analysis focuses on the efficiency gains fom the elimination of estictions on labou mobility. The intoduction of a segmented labou maket is a vey impotant featue of ou model, so that we conside two types of labou: skilled and unskilled. This chaacteistic allows as to analyse the distibutional effects that the migation of skilled labou has on unskilled labou, since the assumption of homogeneous labou implies that the benefits and losses of migation ae evenly distibuted within each egion. Lastly, the pice of the composite commodity demanded by the consume in USA is chosen as the numeaie. 3.1 Benchmak dataset The benchmak dataset involves domestic activity data and extenal secto data fo each egion in 1990. Domestic activity data involve data on value added by component, the segmentation of the labou maket as well as domestic taxes. Extenal secto data includes data on foeign tade and impot taiffs. The size of the eight egions is given by thei espective GDP in 1990 US dollas, consistent with the Wold Tables (1995). The benchmak dataset satisfies the equilibium conditions of the model in the pesence of the existing policies. We use data fom national accounts as compiled by the United Nations, Wold Tables poduced by the Wold Bank, and the Govenment Finance Statistics Yeabook of the Intenational Monetay Fund. Regading foeign tade statistics, we use infomation fom UNCTAD (1995) and the GATT-tade policy eview fo vaious counties. The dataset used was based on a dataset peviously assembled by the autho, in which each egion poduced thee goods, namely pimay commodities, manufactued goods, and sevices. Fo the pupose of this pape, these thee goods wee aggegated into a single commodity. We use infomation fom (vaious issues of) the Yeabook of Labou Statistics of the Intenational Labou Office (ILO) to calculate the pecentages of skilled and unskilled labou in each egion; the following pecentages wee obtained:6 USA JAP EU ODC DAM DAF DAS DE Ls, % 30.2 17.4 20.6 25.5 15.7 14.3 9.1 13.5 Lu, % 69.8 82.6 79.4 74.5 84.3 85.7 90.9 86.5 5 Initially, developing Oceania (which included Fiji, Kiibati, Papua New Guinea, Samoa, Solomon Islands, and Vanuatu) was included as a ninth egion. At the time of solving the model we encounteed numeical poblems because this egion was vey small compaed to the othes (in 1990 its GDP accounted fo only 0.2 pe cent of wold GDP). Hence, it was excluded fom the analysis. 6 An appendix with the souces and the pocedue followed to assemble the dataset is available fom the autho upon equest. 6
As can be seen, these pecentages indicate that moe than 17 pe cent of the labou foce in developed egions is skilled, while in developing egions this pecentage is less than 16 pe cent. National accounts, fom which the wage bill is taken, epot the emuneation of employees without distinguishing between types of labou. The pecentages epoted above ae theefoe impotant since they ae used to split the wage bill into emuneation to skilled and unskilled labou in each egion. The esulting wage bills ae (figues in $ billions): USA JAP EU ODC DAM DAF DAS DE Ls 999.8 281.0 643.5 297.7 61.4 10.2 48.6 110.6 Lu 2,313.8 1,330.8 2,473.0 871.7 330.2 60.8 485.8 706.8 Once we have assembled the dataset, some paamete values such as shae paametes and scale paametes, can be diectly calculated fom the equilibium conditions of the model, following the pocedue descibed in Mansu and Whalley (1984). Because of the functional foms used in the model, we need to specify paamete values fo the elasticities of substitution and tansfomation that ae not contained in the dataset. Once these paametes have been specified, shae paametes can be obtained fom demand functions. On the supply side, shae and scale paametes can be obtained fom cost functions. 3.2 Elasticities The key elasticities in ou model ae the skilled-unskilled labou substitution elasticity, as well as the elasticity of substitution between capital and the aggegate labou input. The degee of substitutability between skilled and unskilled labou detemines the change in elative wages once a policy change is intoduced. On the demand side of the model, the most impotant ae the elasticities contolling substitution between impot types in foming impot composites, and those contolling substitution between compaable domestic goods and aggegate impots. The majoity of studies on labou-labou substitution use a disaggegation by occupation to sepaate the labou foce; in paticula, the disaggegation most widely used is between poduction and non-poduction wokes, because of data availability. Thee does not seem to be consensus as to an appoximate value fo the labou-labou substitution elasticity, and this is eflected by the fact that thee is a athe lage ange of vaiation in the elasticity estimates, fom 0.14 to 7.5 (Hamemesh and Gant 1979).7 The big diffeences in the elasticity estimates can be the esult of majo methodological diffeences, such as the choice of estimating a cost o a poduction function, the choice of functional foms, the choice of data (time-seies vesus coss-section), and the disaggegation of the labou foce accoding to vaious citeia, among othes. The estimate of the elasticity of substitution between non-poduction-poduction wokes was chosen as poxy fo the elasticity of substitution between skilled and unskilled labou. We use a value of 0.9 in ou cental case, and this value is used fo all egions, 7 Hamemesh (1993: 65), howeve, points out that the substitution elationship between poduction and non-poduction wokes tells us little about the substitution between high- and low-skilled wokes because thee is a emakably lage ovelap in the eanings of these two goups. 7
since estimates fo each egion wee not available. Sensitivity analysis is pefomed aound the value chosen in the ange 0.5 to 2.5.8 In the case of the value added functions, the key paametes ae the CES elasticities of substitution between the aggegate labou input and capital.9 We use elasticities of facto substitution based on those used by Whalley (1985). Because of the lack of detailed egional data, ou elasticities ae almost identical acoss egions. On the demand side of the model, two diffeent types of elasticities ae involved with the CES foms used: those contolling substitution between impot types in foming impot composites, and those contolling substitution between compaable domestic goods and aggegate impots. In this model, elasticities of substitution in consumption ae not needed because each epesentative consume demands one good only, which is a composite of compaable domestic and impoted (composite) goods. Table 1 Elasticities in the model Elast. USA JAP EU ODC DAM DAF DAS DE ς 0.900 0.900 0.900 0.900 0.900 0.900 0.900 0.900 σ 0.830 0.800 0.820 0.840 0.850 0.860 0.840 0.840 π 0.920 0.930 0.859 0.948 1.263 1.019 1.546 2.715 ζ 0.990 0.930 0.919 1.130 0.544 0.572 1.227 1.410 Notes: ς is the labou-labou substitution elasticity. σ is the elasticity of substitution between capital and the aggegate labou input; based on estimates pesented in Whalley (1985). π is the elasticity of substitution between domestic and impoted goods. The values used ae based on impot pice elasticities. Fo USA and JAP the souce is Maquez (1990). Fo EU we use an aveage of the elasticities of Gemany and the United Kingdom (Maquez 1990); Fance, Belgium-Luxemboug, Denmak, Ieland, Italy, and the Nethelands (Sten et al. 1976); and Potugal (Houthakke and Magee 1969). Fo ODC we use an aveage of the elasticities of Canada (Maquez 1990); Austia, Finland, Noway, Sweden, Switzeland, Austalia, and New Zealand (Sten et al. 1976). Fo DAM we use an aveage of the elasticities of Agentina, Bazil, Chile, Colombia, Costa Rica, Ecuado, Peu and Uuguay (Khan 1974). Fo DAF we use an aveage of the elasticities of Ghana and Moocco (Khan 1974). Fo DAS we use an aveage of the elasticities of India, the Philippines and Si Lanka (Khan 1974); and Pakistan and Bangladesh (Nguyen and Bhuyan 1977). Fo DE we use the elasticity fo Tukey estimated by Khan (1974). ζ is the elasticity of substitution between egional impots. The values used ae based on expot pice elasticities. Fo USA and JAP the souce is Maquez (1990). Fo EU we use an aveage of the elasticities of Gemany and the United Kingdom (Maquez 1990); Fance, Belgium- Luxemboug, Denmak, Ieland, Italy, and the Nethelands (Sten et al. 1976); and Potugal (Houthakke and Magee 1969). Fo ODC we use an aveage of the elasticities of Canada (Maquez 1990); Austia, Finland, Noway, Sweden, Switzeland, Austalia, and New Zealand (Sten et al. 1976). Fo DAM we use an aveage of the elasticities of Agentina, Bazil, Chile, Colombia, Costa Rica, Ecuado, and Peu (Khan 1974). Fo DAF we use an aveage of the elasticities of Ghana and Moocco (Khan 1974). Fo DAS we use an aveage of the elasticities of Pakistan, India, Bangladesh and Si Lanka (Nguyen and Bhuyan 1977). Fo DE we use the elasticity fo Tukey estimated by Khan (1974). 8 It was also tied to use elasticity values geate than 2.5, but we encounteed numeical poblems when solving the model. 9 Whalley (1985) points out that thee is no consensus as to the quantitative odes of magnitude involved, since most time-seies estimates of the aggegate substitution elasticity ae in the neighbouhood of unity, and coss-section estimates ae often aound 0.5. 8
Regading tade elasticities, the most impotant ae impot-pice elasticities and expotpice elasticities. Substitution elasticities between impot types making up any composite detemine the expot-pice elasticities faced by egions. Substitution elasticities between impot composites and compaable domestic poducts eflect impot-pice elasticity estimates in the liteatue, since it was not possible to find any econometic estimate of elasticities of substitution. The elasticities used in the model (cental case) ae pesented in Table 1. 3.3 Calibation Once the dataset has been assembled, and elasticity paametes have been specified, shae and scale paametes can be calculated fom the equilibium conditions of the model, following the pocedue descibed in Mansu and Whalley (1984). The benchmak dataset povides infomation on equilibium tansactions in value tems. The fist step of the calibation pocedue involves the sepaation of these tansactions into pice and quantity obsevations. In ode to do this, a units convention is widely used, in which it is assumed that a physical unit of each good and facto is the amount that sells fo one dolla. That is, both goods and factos have a pice of unity in the benchmak equilibium. Howeve, this appoach is not applicable in the case of the labou maket, because we assume diffeent maginal poducts of labou, esulting fom baies to inwad mobility of labou in high-wage counties (that is, wages ae diffeent fom one). In addition, we conside two types of labou, skilled and unskilled, each one with a diffeent poductivity and, as a esult, a diffeent pice within each egion. Thee is no ageement as to how to calculate the aveage wage ate. Hence, we conside six altenative measues, which ae the most widely used. Fist, we use the wage bill fo each egion (WB), as taken fom national accounts, and divide it by total population (TOTP), as taken fom the UN Demogaphic Yeabook. Total population, howeve, exceeds the wokfoce in each egion. Theefoe, we use as a second measue of the aveage wage ate the wage bill divided by the economically active population (EAP).10 The thid and fouth measues use GDP pe capita using TOTP and EAP, espectively. The fifth and sixth measues of the aveage wage ate use GDP pe capita using TOTP and EAP, whee the GDP has been adjusted by the exchange ate deviation index, that coects fo the diffeence between the official and the puchasing powe paity exchange ates (AGDP pc ) (Kavis et al. 1982). The wage measues based on GDP pe capita wee included fo compaison puposes, since Hamilton and Whalley (1984) used this measue in thei calculations. Howeve, GDP pe capita is only an appoximate measue of aveage wages as it is a measue of economic activity, and not a measue of income. Futhemoe, in the poduction of domestic output, labou is not the only facto of poduction involved; physical capital and human capital ae also involved. Fom GDPpc it is not possible to isolate the labou component. Since thee is labou maket segmentation, we need to calculate the aveage wage ates of skilled and unskilled labou in each egion. Given that in pactice such data ae not 10 ILO (1996: 5) defines the economically active population as all pesons of eithe sex who funish the supply of labou fo the poduction of goods and sevices duing a specified time-efeence peiod. 9
available, we use aveage eanings pe woke in finance, insuance, eal state and business sevices as poxy fo skilled labou wages, while aveage eanings pe woke in wholesale and etail tade, estauants and hotels as poxy fo unskilled labou wages. The atio between high and low wages is then used to infe the aveage wage ates fo skilled and unskilled labou in each egion. The esulting elative wage ates fo the two types of labou ae epoted in Table 2. As can be seen, egadless of how the wage ates ae calculated, USA, JAP, EU and ODC have highe wage ates than the developing wold (i.e., DAM, DAF, DAS and DE). The final step in the calibation pocedue is to use the pice-quantity data to calculate paametes fo demand and poduction functions fom the benchmak equilibium obsevations, given the equied values of pe-specified paametes such as elasticities and tax ates. In ode to do this, we use the equilibium conditions togethe with fistode conditions (fom utility maximization and cost minimization), to solve fo function paamete values using equilibium pices and quantities. Calibation allows us to test the solution pocedue, and ensues the consistency of agents behaviou with the benchmak dataset. The model was solved using a outine we wote in the Geneal Algebaic Modelling System (GAMS) softwae. Table 2 Relative wage ates (1990 US$) Regions Wage measues USA JAP EU ODC DAM DAF DAS DE WB/TOTP Unskilled labou 91.3 96.3 60.2 72.9 6.7 1.0 1.5 13.8 Skilled labou 150.0 158.1 122.2 97.6 13.1 1.8 2.2 17.3 WB/EAP Unskilled labou 68.5 71.3 52.9 61.6 7.1 1.0 1.6 11.3 Skilled labou 112.6 117.0 107.5 82.4 13.8 1.8 2.3 14.1 GDP pc (TOTP) Unskilled labou 31.9 51.0 23.8 25.4 4.2 1.0 0.9 5.1 Skilled labou 52.4 83.7 48.3 33.9 8.1 1.8 1.3 6.4 GDP pc (EAP) Unskilled labou 23.9 37.7 20.9 21.4 4.4 1.0 1.0 4.2 Skilled labou 39.3 62.0 42.5 28.6 8.6 1.8 1.4 5.2 AGDP pc (TOTP) Unskilled labou 16.0 28.1 11.8 12.6 4.0 1.0 1.2 3.8 Skilled labou 26.2 46.1 23.9 16.8 7.7 1.8 1.8 4.8 AGDP pc (EAP) Unskilled labou 12.0 20.8 10.4 10.6 4.2 1.0 1.3 3.1 Skilled labou 19.7 34.1 21.1 14.2 8.2 1.8 1.8 3.9 10
4 Model esults The model descibed above was used to calculate the woldwide efficiency gains fom fee mobility of labou (the esults ae pesented fo the six measues of wages mentioned peviously). We conside two scenaios: in the fist one labou is a homogeneous facto of poduction; this scenaio is included in ode to compae the esults with those obtained by Hamilton and Whalley (1984). In the second scenaio, labou is classified as skilled and unskilled. In the latte scenaio, we conside two cases: (i) both skilled and unskilled labou migate; and (ii) skilled labou is the only facto that migates. We did not conside the case whee unskilled labou is the only facto that migates, since unskilled labou is usually involved in illegal migation and the model does not conside this type of migation. The emoval of estictions on labou mobility modifies the maket cleaing condition that detemines the equilibium wage ate. In paticula, when labou is homogeneous the equilibium condition is given by 8 8 L = L, [4] = 1 = 1 whee L coesponds to the egion s endowment of labou. In the heteogeneous case we have and 8 Ls = 8 = 1 = 1 8 Lu = 8 = 1 = 1 Ls Lu [5a], [5b] whee Ls and Lu coespond to the egion s endowment of the two types of labou. In the model intenational capital tansfes ae not consideed, since it is assumed that migant wokes do not bing capital with them no send capital back home. Capital flows and tansfes may alleviate the negative effects of migation on wages. In addition, the model assumes that all migant labou entes the labou maket (some migants such as childen and eldely people will not actually wok). Once immigation contols ae emoved, labou migates fom low-wage egions to high-wage egions. The souce egions ae DAM, DAF, DAS, and DE, while the destination egions ae USA, JAP, EU, and ODC. Howeve, when the aveage wage ate is measued as wage bill divided by EAP, and wage bill divided by TOTP, DE becomes a destination egion fo the homogeneous labou case. When labou is heteogeneous, and both skilled and unskilled labou migate, DE becomes a destination egion fo unskilled labou, and a souce egion fo skilled labou. Regadless of whethe labou is homogeneous o heteogeneous, the amount of the facto enteing DE is not consideable. 11
Table 3 quantifies the effects of the emoval of immigation contols on welfae, as measued by the aggegate equivalent vaiation.11 In the homogeneous labou case, thee is a eduction in poduction in all sectos in the souce egions. This is accompanied by a eduction in expots and an incease in impots which compensate fo the eduction in domestic output. Convesely, in the destination egions, thee is an incease in poduction in all sectos accompanied by an incease in expots and a eduction in impots fom developing egions. In this case, thee ae lage gains fom the emoval of global immigation contols, anging fom 15 pe cent to 67 pe cent of wold GDP. These gains ae not as lage as those obtained by Hamilton and Whalley (1984), whee in some cases the gains exceeded the woldwide economy GNP. The diffeences may be the esult of the modelling famewoks (i.e. patial equilibium vesus geneal equilibium), the flows of labou leaving low-wage egions, o units of measuement as Hamilton and Whalley (1984) use population, and we use units of labou. Table 3 also pesents the welfae effects of the emoval of immigation contols when labou is a heteogeneous facto. In this case, as in the pevious scenaio with homogeneous labou, thee is an incease in domestic output in developed egions, wheeas output educes in developing counties; the eduction in domestic output is compensated by a eduction in expots and an incease in impots fom developed egions. When both skilled and unskilled labou migate, efficiency gains ange fom 13 pe cent to 59 pe cent of wold GDP. The gains ae smalle than in the homogeneous case as a esult of the technological constaint imposed by the substitutability between skilled and unskilled labou. Thus, with a segmented labou maket skilled and unskilled labou have less oppotunity to eallocate. When only skilled labou migates, woldwide welfae gains ae much smalle than in the pevious two cases (fom 3 pe cent to 11 pe cent of wold GDP) because skilled labou epesents a small faction of the labou foce in the souce egions (i.e. 14 pe cent in DAM, 10 pe cent in DAF, 5 pe cent in DAS, and 14 pe cent in DE). The segmentation of the labou maket also allows us to examine the distibutional effects of immigation between skilled and unskilled labou in each egion. Tables 4 and 5 Table 3 Welfae effects of the emoval of immigation contols (Equivalent vaiation as a pecentage of wold GDP) Homogeneous labou Heteogeneous labou Both Ls and Lu migate WB/TOTP 67 59 11 WB/EAP 54 48 9 GDPpc TOTP) 45 41 8 GDPpc(EAP) 36 32 7 AGDPpc(TOTP) 19 17 4 AGDPpc (EAP) 15 13 3 Note: Ls and Lu denote skilled and unskilled labou, espectively. Only Ls migates 11 The equivalent vaiation (EV) is a measue of welfae change. It is defined as the amount of money a paticula change, that has taken place between equilibia, is equivalent to. In this case, an aithmetic sum of EVs, summed acoss egions is used. 12
pesent the distibutional impacts of the emoval of immigation contols fo the six measues of wages consideed. A pioi one would expect that labou migation fom the souce egions inceases the labou supply in the destination egions, educing the aveage wage ate (assuming no igidities), and benefiting capital ownes. In the souce egions, the emoval of immigation contols is expected to educe the labou supply, inceasing the aveage wage ate. As a esult, capital is less scace elative to labou, so that a eduction in the etun to capital is expected. When both skilled and unskilled labou migate (see Table 4), aveage wages incease in the souce egions because labou is less abundant elative to capital, and the etun to capital deceases. The emoval of immigation contols benefits skilled labou moe than unskilled labou, because the fome is a small popotion of the total labou foce, and afte migation this facto is moe scace in developing egions. In the destination egions, aveage wages educe fo both skilled and unskilled labou, since labou is now less scace elative to capital, and the etun to capital inceases. When only skilled labou migates (see Table 5), thee is a substantial incease in the emuneation of this type of labou in the souce egions, since this facto of poduction is not abundant in these egions. Unskilled wokes and capital ownes ae wose off as a esult of migation, despite the fact that thee is an incease in thei emuneation. As to the destination egions, the inflow of skilled labou inceases the supply of this type of labou, hence educing its aveage wage ate. As we would expect, the aveage wage of unskilled labou and the etun to capital incease. Skilled labou is wose off. The flexibility of wages allows the labou maket to absob labou immigation. Lowe wages induce an incease in labou demand and in aggegate employment. The amount of labou leaving the souce egions vaies depending on the measue used to calculate aveage wages (see Table 6). Fo example, when these ae measued as the wage bill divided by TOTP, 53 pe cent of the labou endowment of developing egions migate to developed egions; when the aveage wage ate is measued as adjusted GDP pe capita using EAP, this pecentage educes to 37 pe cent. On the othe hand, when both skilled and unskilled labou migate in the heteogeneous labou case, the pecentage of labou leaving the souce egions vaies fom 35 pe cent (aveage wage ate measued as adjusted GDP pe capita using EAP) to 50 pe cent (aveage wage ate measued as the wage bill divided by TOTP). When only skilled labou migates, between 59 pe cent and 73 pe cent of the skilled labou endowment of developing egions migate, depending on how the aveage wage ate is calculated. Regading the costs of migation, these ae estimated between 5 pe cent and 8 pe cent of GDP (depending on the wage measue) when labou is assumed to be homogeneous and also when both skilled and unskilled labou migate; see Table 7 These costs ae clealy smalle when compaed to the welfae gains epoted in Table 3. On the contay, when only skilled labou migates, the estimated costs of migation ae lage than the welfae gains, which can be explained by the fact that a much lage pecentage of skilled labou leaves the developing egions. Thus, the net welfae effects of the emoval of estictions on labou mobility ae negative. In summay, migation leads to facto eallocation, and duing this pocess thee ae winnes and loses. In the souce egions, labou becomes moe scace elative to capital (between 37 pe cent and 53 pe cent of the labou endowment of developing egions migate to developed egions, depending on the wage measue used), and capital ownes lose. Howeve, not all wokes ae bette off, since labou is a heteogeneous facto. 13
Table 4 Distibutional impact of the emoval of immigation contols Heteogeneous labou: both skilled and unskilled labou migate (% change) USA JAP EU ODC DAM DAF DAS DE WB/TOTP P LS -18-22 0 26 837 6,671 5,446 608 P LU -47-50 -20-34 620 4,697 3,022 249 P K 191 236 181 164 263 1,351 900 329 WB/EAP P LS -12-15 -7 21 619 5,377 4,246 604 P LU -40-43 -23-34 478 3,981 2,467 263 P K 145 179 142 129 363 1,130 755 312 GDP pc (TOTP) P LS -7-42 1 43 499 2,580 3,535 662 P LU -34-59 -12-17 406 2,005 2,161 313 P K 113 146 116 113 308 615 655 328 GDP pc (EAP) P LS -3-39 -10 33 343 1,998 2,654 631 P LU -28-54 -17-19 294 1,633 1,701 317 P K 83 100 85 85 229 494 529 307 AGDP pc (TOTP) P LS -0-43 9 56 238 1,342 1,374 446 P LU -17-53 12 5 234 1,220 969 245 P K 44 38 66 63 174 364 332 234 AGDP pc (EAP) P LS 1-42 -6 40 143 993 982 407 P LU -13-50 0-2 149 942 716 234 P K 31 17 42 43 115 250 250 210 Note: P LS coesponds to the aveage wage ate of skilled labou; P LU is the aveage wage ate of unskilled labou; and P K efes to the etun to capital. Table 5 Distibutional impact of the emoval of immigation contols Heteogeneous labou: skilled labou migates only (% change) USA JAP EU ODC DAM DAF DAS DE WB/TOTP P LS -60-62 -51-38 361 2,234 2,631 249 P LU 24 55 48 40 117 217 189 129 P K 27 60 51 42 115 211 184 128 WB/EAP P LS -52-54 -50-35 289 2,862 2,250 281 P LU 19 46 39 34 97 190 164 129 P K 22 41 42 36 96 185 160 128 GDP pc (TOTP) P LS -46-66 -41-16 250 1,464 2,022 345 P LU 16 35 35 34 86 116 146 139 P K 18 39 37 35 85 113 143 138 GDP pc (EAP) P LS -38-61 -43-15 183 1,239 1,657 366 P LU 13 25 26 27 67 97 121 135 P K 14 28 28 28 66 94 118 134 AGDP pc (TOTP) P LS -27-59 -20 14 147 954 978 299 P LU 8 9 23 26 54 76 78 111 P K 9 12 24 27 53 74 76 109 AGDP pc (EAP) P LS -21-54 -26 10 91 761 752 299 P LU 6 3 14 19 37 58 59 102 P K 7 4 15 20 36 56 57 101 Note: See note to Table 4. 14
Table 6 Migation flows as % of developing egions labou endowment Homogeneous labou Both Ls and Lu migate Heteogeneous labou Only Ls migates WB/TOTP 53 50 73 WB/EAP 51 48 72 GDPpc TOTP) 48 46 70 GDPpc(EAP) 47 44 68 AGDPpc(TOTP) 39 37 61 AGDPpc (EAP) 37 35 59 Note: Ls and Lu denote skilled and unskilled labou, espectively. Table 7 Migation costs as % of wold GDP Homogeneous labou Both Ls and Lu migate Heteogeneous labou Only Ls migates WB/TOTP 8 8 12 WB/EAP 8 8 11 GDPpc TOTP) 8 7 11 GDPpc(EAP) 7 7 11 AGDPpc(TOTP) 6 6 10 AGDPpc (EAP) 6 5 9 Note: Migation costs wee calculated as the numbe of people migating multiplied by the cost of moving. The numbe of people that migates is calculated as the economically active population in developing egions multiplied by the pecentage of labou moving fom developing to developed egions. As to migation costs, we use Conley and Ligon (2002) to obtain estimates of tanspot costs. They epot aifaes between counties capitals as the cost of tanspoting embodied human capital and the cost of shipping a 20 kg expess package between capital cities. We used thei data base, as taken fom http://ae.bekeley.edu/~ligon/papes/distance.tgz, to calculate an appoximate aveage cost of moving fom developing to developed egions. Fom the model it is possible to establish that people migate fom developing to developed egions, but not to which paticula developed egion. Emigation will benefit wokes whose skills ae substitute to those of migant labou, wheeas it will hut those wokes whose skills ae complementay to those of migant wokes. On the othe hand, in the destination egions, labou becomes moe abundant (less scace) elative to capital, so that capital ownes benefit. Howeve, not all wokes ae wose off, because labou is a heteogeneous facto. Immigation will benefit those wokes whose skills ae complementay to those of the immigant woke, wheeas immigation will hut those wokes whose skills ae substitute to those of immigant wokes. We also pefomed a sensitivity analysis on the key elasticities of the model (these esults ae not epoted hee). In paticula, in a fist set of simulations the elasticity of labou-labou substitution was vaied fom 0.5 to 2.5. This elasticity is vey impotant in ou model since it includes a segmented labou maket, a featue that has not been consideed in pevious woks. In a second set of simulations, the elasticities of substitution in the poduction of value added wee set at values between 0.5 and 1.5 in all egions; this elasticity coesponds to the elasticity of substitution between the 15
aggegate labou input and capital. We conclude that the esults ae obust to the elasticity choice, in the sense that the elimination of immigation contols geneates woldwide efficiency gains. In addition, in the destination egions capital ownes benefit fom labou immigation, and wokes lose because of lowe wages. In the souce egions, capital ownes ae wose off and wokes ae bette off. The sensitivity analysis also confims that migation of skilled labou huts unskilled labou in the souce egions. 5 Model extensions This section intoduces thee new featues to the model: (i) tansaction costs; (ii) intenational capital mobility; and (iii) selective labou mobility. Fo bevity we only epot the esults of two out of the six measues of aveage wages consideed: aveage wage ate as measued by the wage bill divided by TOTP, and as adjusted GDP pe capita using EAP. These two measues wee chosen as they povided the exteme esults. 5.1 Tansaction costs The fist elaboation of the model is the intoduction of tansaction costs. This extension of the model seems appopiate, since migation is a costly pocess. Thee ae costs involved in the pocess of moving fom one egion to anothe, such as tanspot costs, the costs of settling in the othe egion, the costs of finding a new job, and the costs of leaving fiends and family behind. With the elimination of estictions on labou mobility, labou will move until the maginal poduct of labou equals the cost of hiing labou; hence, a single maket cleaing wage no longe chaacteizes the equilibium. Tansaction costs thus dive a wedge between wages in developed and developing counties. These wee modelled as a tax without evenue, whose ate is exogenously detemined. The pice eceived by ownes of labou in each egion coesponds to a pecentage of the maket cleaing pice when estictions to labou mobility ae eliminated. That is, the pice of labou in each egion is given by, P = W( 1 TC ), [6] L whee W coesponds to the wold pice of labou, and TC coesponds to egional tansaction costs. Tansaction costs ae difficult to quantify since thee ae no measues available. As mentioned ealie, thee ae costs associated with migation fom low-wage to highwage egions. In the case of developing egions, these costs could be vey high. Taking into account the substantial diffeences in elative wages among the egions, we assume the following values fo tansaction costs: 0.9 fo DAF and DAS; 0.8 fo DAM; and 0.7 fo DE. The tansaction costs fo developed egions (USA, JAP, EU, and ODC) ae assumed to be much smalle (i.e. 0.1), since wokes in these egions have little o no incentive to move to low-wage egions. The intoduction of tansaction costs educes migation flows (see Table 8). Fo example, when the aveage wage is measue as WB/TOTP, and labou is homogeneous, migation educes fom 53 pe cent of the developing egions endowment of labou to 16
24 pe cent. In the heteogeneous labou scenaio, migation educes fom 50 pe cent of the developing egions endowment of labou to 32 pe cent when the two types of labou ae allowed to migate and fom 73 pe cent of the developing egions endowment of skilled labou to 53 pe cent when only skilled labou migates. Welfae gains (see Table 9) and migation costs (see Table 10) ae also smalle in the pesence of tansaction costs, because less people ae moving fom developing to developed egions. Without tansaction costs Table 8 Migation flows in the pesence of tansaction costs (Migation as a pecentage of developing egions labou endowment) Homogeneous labou Both Ls and Lu migate Heteogeneous labou Only Ls migates WB/TOTP 53 50 73 AGDP pc (EAP) 37 35 59 With tansaction costs WB/TOTP 24 32 53 AGDP pc (EAP) 4 2 9 Note: Ls and Lu denote skilled and unskilled labou, espectively. Table 9 Welfae effects of the emoval of immigation contols in the pesence of tansaction costs (Equivalent vaiation as a pecentage of wold GDP) Without tansaction costs Homogeneous labou Both Ls and Lu migate Heteogeneous labou Only Ls migates WB/TOTP 67 59 11 AGDP pc (EAP) 15 13 3 With tansaction costs WB/TOTP 31 26 6 AGDP pc (EAP) 1 1 1 Note: See Table 8. Without tansaction costs Table 10 Migation costs in the pesence of tansaction costs as % of wold GDP Homogeneous labou Both Ls and Lu migate Heteogeneous labou Only Ls migates WB/TOTP 8 8 12 AGDP pc (EAP) 6 5 9 With tansaction costs WB/TOTP 4 5 8 AGDP pc (EAP) 1 0 1 Note: See Table 8. 17
Regading the distibutional effects, esults not epoted hee indicate that when the labou maket is segmented, skilled labou benefits elative to unskilled labou in the souce egions; in the destination egions the two types of labou lose, but unskilled wokes ae hut even moe when both skilled and unskilled labou migate. Finally, migation, welfae gains and migation costs incease as the tansaction costs fo the developing egions ae educed (these esults ae not epoted hee though). This is the case since tansaction costs distot the labou maket, specially in developing egions, and as the distotion is educed, efficiency inceases and the wage gap educes. 5.2 Capital mobility In the second elaboation of the model we intoduce intenational capital mobility. Although this featue is usually ignoed in global models (see e.g., Whalley 1985; Shoven and Whalley 1992), it seems inteesting to include it in the model since capital makets ae becoming moe integated intenationally. In this case, the etun to capital equalizes acoss egions. Theefoe, a single maket cleaing ental ate chaacteizes the equilibium; that is, the maket cleaing condition fo the maket of the capital facto is given by, 8 K = 8 = 1 = 1 K [7] that is the sum of the demand fo capital in each egion must equal the global endowment of the facto. Simulations wee caied out fo the scenaio in which only skilled labou migates, since we need a fixed facto (in this case unskilled labou). If all factos of poduction ae allowed to move feely, the concept of egion is no longe clea. When we emove the estictions to skilled labou mobility, we obseve that labou moves fom egions with low wages (DAM, DAF, DAS, and DE) to egions with high wages (USA, JAP, EU, and ODC). Capital moves fom egions whee it is abundant elative to labou (USA, JAP, EU, and ODC) to egions whee it is scace elative to Table 11 Distibutional impact of the emoval of immigation contols in the pesence of capital mobility (% change) USA JAP EU ODC DAM DAF DAS DE WB/TOTP P LS -62-64 -54-42 332 3,018 2,454 226 P LU 14 29 28 24 69 136 131 76 P K 48 48 48 48 48 48 48 48 AGDPpc(EAP) P LS -24-56 -29 5 82 720 712 280 P LU 3 1 7 11 20 37 40 57 P K 15 15 15 15 15 15 15 15 18