Provided in Cooperation with: Ifo Institute Leibniz Institute for Economic Research at the University of Munich

Transcription

1 econstor Der Open-Access-Publkatonsserver der ZBW Lebnz-Informatonszentrum Wrtschaft The Open Access Publcaton Server of the ZBW Lebnz Informaton Centre for Economcs Fernández-Huertas oraga, Jesús; Rapoport, Hllel Workng Paper Tradable mmgraton quotas CESfo Workng Paper: Publc Fnance, No Provded n Cooperaton wth: Ifo Insttute Lebnz Insttute for Economc Research at the Unversty of unch Suggested Ctaton: Fernández-Huertas oraga, Jesús; Rapoport, Hllel 2013 : Tradable mmgraton quotas, CESfo Workng Paper: Publc Fnance, No Ths Verson s avalable at: Standard-Nutzungsbedngungen: De Dokumente auf EconStor dürfen zu egenen wssenschaftlchen Zwecken und zum Prvatgebrauch gespechert und kopert werden. Se dürfen de Dokumente ncht für öffentlche oder kommerzelle Zwecke vervelfältgen, öffentlch ausstellen, öffentlch zugänglch machen, vertreben oder anderwetg nutzen. Sofern de Verfasser de Dokumente unter Open-Content-Lzenzen nsbesondere CC-Lzenzen zur Verfügung gestellt haben sollten, gelten abwechend von desen Nutzungsbedngungen de n der dort genannten Lzenz gewährten Nutzungsrechte. Terms of use: Documents n EconStor may be saved and coped for your personal and scholarly purposes. You are not to copy documents for publc or commercal purposes, to exhbt the documents publcly, to make them publcly avalable on the nternet, or to dstrbute or otherwse use the documents n publc. If the documents have been made avalable under an Open Content Lcence especally Creatve Commons Lcences, you may exercse further usage rghts as specfed n the ndcated lcence. zbw Lebnz-Informatonszentrum Wrtschaft Lebnz Informaton Centre for Economcs

2 Tradable Immgraton Quotas Jesús Fernández-Huertas oraga Hllel Rapoport CESIFO WORKING PAPER NO CATEGORY 1: PUBLIC FINANCE JANUARY 2013 An electronc verson of the paper may be downloaded from the SSRN webste: from the RePEc webste: from the CESfo webste: Twww.CESfo-group.org/wpT

3 CESfo Workng Paper No Tradable Immgraton Quotas Abstract Internatonal mgraton s maybe the sngle most effectve way to allevate global poverty. When a gven host country allows more mmgrants n, ths creates costs and benefts for that partcular country as well as a postve externalty for ndvduals and governments who care about world poverty. Ths mples that the exstng nternatonal mgraton regme s neffcent as t fals to nternalze such externalty. In addton, host countres qute often restrct mmgraton due to ts apparently unbearable socal and poltcal costs, however these costs are never measured and made comparable across countres. In ths paper we frst dscuss theoretcally how tradable mmgraton quotas TIQs can reveal countres comparatve advantage n hostng mmgrants and, once coupled wth a matchng mechansm takng mgrants preferences over destnatons and countres preferences over mgrants types nto account, generate substantal welfare gans. We then dscuss two potental applcatons: a market for the resettlement of nternatonal e.g., clmate change refugees, and the creaton of an OECD poverty-reducton vsa program adapted from the US green card lottery. JEL-Code: F220, F500, H870, I300, K330, O190. Keywords: mmgraton, mmgraton polcy, tradable quotas, refugees resettlement, clmate change, nternatonal publc goods. Jesús Fernández-Huertas oraga FEDEA & IAE CSIC Jorge Juan, 46 Span adrd jfernandezhuertas@fedea.es Hllel Rapoport Center for Internatonal Development Kennedy School of Government Harvard Unversty 79 JFK Street USA Cambrdge, A hllel.rapoport@hks.harvard.edu January 10, 2013

4 I am dstressed by the sght of poverty; I am benefted by ts allevaton; but I am benefted equally whether I or someone else pays for ts allevaton lton Fredman, Captalsm and Freedom, 1962, page Introducton People care about poverty out of altrusm.e., genune concern for others well-beng and self-nterest, because they fear for ther securty, health, and property. To the same extent that domestc poverty allevaton has the dmenson of a domestc publc good, nternatonal poverty allevaton has the dmenson of an nternatonal publc good. Whenever a gven country ncreases ts foregn ad to one of the countres where many of the world s poor lve, ths generates a postve externalty for all those n the world, ndvduals and governments, who care about nternatonal poverty reducton. Whenever a gven country chooses to let ther people come Prtchett, 2006, ncreasng the number of mmgraton vsas granted to natonals of one of the countres where many of the world poor lve, ths generates a postve externalty for all those who care about nternatonal poverty reducton. In both cases and gven the publc good nature of poverty allevaton, free rdng s lkely to preval and result n global under-provson of foregn ad, debt relef programs, and mmgraton vsas. Whle the nternatonal communty has establshed nternatonal organzatons and set up nsttutons to coordnate foregn ad and debt relef efforts, no such nsttutonal settng exsts for nternatonal mgraton. In realty, hgh-ncome countres qute often restrct mmgraton of poor people from 1 We thank Ran Abramtzky, Smone Bertol, Omer Bran, George Borjas, Bran Burgoon, Roberto Burguet, aría Calle, Alessandra Casella, Kfr Elaz, Avner Gref, Tmothy Hatton, Glenn Loury, Igal lchtach, Panu Poutvaara, Lant Prtchett, Dan Rodrk, Robert Stavns, Eko Thelemann, Davd Wel, partcpants at the 4 th INSIDE Workshop Barcelona, the 3 rd AFD-World Bank graton and Development Conference, the 2010 eetng of the Spansh Economc Assocaton, the 2012 Envronment and Development Conference at CERDI, the Thrd TEPO Conference on Internatonal graton at the IAB n Nuremberg, the frst CEIR conference at CESfo, unch, December 2012, and semnar audences at IAE CSIC, Llle, Pars School of Economcs, Toulouse School of Economcs, Harvard Kennedy School, Harvard Center for Populaton and Development Studes, Scence Po, Stanford, Louvan, UCL and Brown, for comments and suggestons. We are partcularly ndebted to Davd De la Crox and Frédérc Docquer for stmulatng exchanges and to Lída Brun for helpful research assstance. Jesús Fernández-Huertas oraga receved fnancal support from the ECO project funded by the Spansh nstry for Scence and Innovaton. 2

5 poor countres due to ts apparently unbearable socal and poltcal costs. However, these non-economc costs potental threats to natonal dentty and socal coheson - are never measured and made comparable across countres. 2 Dfferences n the perceved costs of mmgraton may come from dfferent demographc structures e.g., dependency ratos, hstores of prevous mmgraton, or preferences for ethnc, relgous and cultural dversty. Whatever the source, these dfferences mply that some countres have a comparatve advantage n hostng mmgrants. A system of tradable mmgraton quotas TIQs, therefore, creates an opportunty to reveal and explot such comparatve advantages. 3 We frst show ths theoretcally n Secton 2 where we adapt a generc model of tradable quotas to mgraton and supplement t wth a matchng mechansm takng mgrants and countres preferences nto account. Ths matchng component s essental because, n contrast for example to polluton partcles a well-know applcaton of tradable quotas, mgrants have preferences over ther destnaton and destnaton countres have preferences over the orgn and, possbly, other characterstcs of mmgrants. Takng these preferences nto account creates opportuntes for strategc behavor that may undermne the effcency of the proposed system but that can also boost ts attractveness. Fnally, the feasblty of a TIQs system s evaluated n Secton 3 aganst the background of two potental applcatons: a market for the resettlement of nternatonal ncludng clmate change refugees, and an OECD poverty-reducton vsa program nspred by the US dversty lottery. The rest of ths ntroducton dscusses the poverty-allevaton effects of nternatonal mgraton, the dea that vsas can substtute for ad, and the related lterature. 1.1 Internatonal mgraton and poverty reducton Globalzaton s qute advanced for goods and captal but stll very mperfect for low-skll labor moblty. Partly due to ths asymmetry n the extents of globalzaton at dfferent 2 Clearly, negatve atttudes toward mmgraton consttute a strong poltcal barrer to more labor moblty. See for example O Rourke and Snnott 2006, Dustmann and Preston 2007, Hanson, Scheve and Slaughter 2007, ayda 2008, Facchn and ayda 2008, or Hanmueller and Hscox Usng survey data from 21 European countres, Card, Dustmann and Preston 2012 fnd that non-economc costs, whch they term compostonal concerns, are two to fve tmes more mportant n explanng atttudes towards mmgraton than concerns about wages and net fscal contrbutons. 3 See Renhart 2000 for evdence on the fact that frms learn ther cost functons when ncentvzed to do so through a system of tradable polluton permts. 3

6 margns, the potental gans from even a small lberalzaton of nternatonal mgraton are orders of magntude hgher than, say, a full lberalzaton of trade n goods and servces, a comprehensve full debt relef program, or a doublng of offcal development ad Prtchett, 2006, 2010; Clemens, In the words of Rodrk 2007, p. 240, therefore, allowng for more nternatonal moblty of workers today s really gong for the real gans. The povertyreducton effect of these gans, obvously, depends on ther dstrbutonal mpact across and wthn countres. Recent lterature suggests that ths mpact s substantal both drectly that s, through the extracton of mgrants out of poverty and ndrectly, through the developmental mpact of mgraton on source countres. The man dffculty n measurng the ncome gans accrung to mgrants as a result of mgraton s to produce sensble counterfactuals of domestc earnngs for mgrants, accountng not just for ther observable characterstcs but also for unobservable characterstcs such as motvaton at work, atttudes toward rsk, cogntve ablty, etc. ckenze, Gbson and Stllman 2010 use the New Zealand mgraton lottery program to clean ncome gans estmates from such self-selecton effects. Comparng lottery-wnnng mgrants to lotterylosng non-mgrants, they fnd mgraton ncreases mgrants earnngs by a factor of four. Ths s consstent wth the non-expermental results of Clemens, ontenegro and Prtchett 2008 who compare workers n developng countres to workers from the same countres workng n the Unted States. After controllng for workers characterstcs, mgraton s found to rase real wages by 200 percent, 250 percent and 680 percent respectvely for Guatemalans, Flpnos and Hatans. These ncome gans would seem to exceed the potental gans of any n stu development polcy by orders of magntude. For example, they calculate that the total present value of access to a lfetme of mcro-credt s equvalent to the wage dfference of just four work weeks of the same worker n the US versus n Bangladesh, or that the present value of a lfetme wage ncrement of one addtonal year of schoolng obtaned at no cost s equvalent to 11 weeks work of the same worker n the US versus n Bolva. The effects of mgraton on poverty reducton through the drect extracton of mgrants out of poverty can be llustrated by fgures put together by Clemens and Prtchett 2008 usng three poverty standards at US$1, 2 and 10 per day n PPP. Respectvely 50, 75 and 93 percent of all Hatan naturals people born n Hat lve below the $1, 2 and 10 poverty lnes. Out of the 25 percent of all Hatans between the frst two lnes, 26 percent are US mmgrants. Out of the 18 percent between the last two lnes, 82 percent are US mmgrants. 4

7 By the latter measure, among the 56 percent of all excans between the last two lnes, 43 percent are US mmgrants. Whle t would be an abuse of language to nterpret these fgures as ndcatve of the share of people escapng poverty thanks to mgraton, they are clearly suggestve of large drect effects of mgraton on poverty reducton. 4 In addton, we know that whle mgrants ntally come from the mddle of the ncome and wealth dstrbutons, network and other dynamc effects act to reduce mgraton costs, makng mgraton affordable to households down the ncome ladder. Ths generates poverty and nequalty reducng effects both drectly, through mgrants self-selecton patterns, and ndrectly, through general equlbrum effects and dstrbutonal effects of remttances gradually reachng poorer households ckenze and Rapoport, 2007, 2010; Shen et al., Vsas, Not Ad! Alternatve strateges for nternatonal poverty reducton The dea that vsas can be used as part of an ad relef strategy frst materalzed when the US Temporary Protected Status TPS mechansm, enacted n 1990, was appled to thousands of Hondurans and Ncaraguans n the aftermath of Hurrcane tch n 1998 UNHCR, TPS was also granted to llegal Salvadoran mmgrants followng the earthquakes that devastated El Salvador n The decson was made by then Presdent George W. Bush at the request of the Salvadoran Presdent, Francsco Flores, durng a Whte House meetng. The status allowed 150,000 undocumented Salvadorans to legally reman n the Unted States for eghteen months. 5 ore recently, TPS was also granted to tens of thousands of llegal Hatan mmgrants followng the earthquake n Hat n 2010, and n 2012 Hat has been added to the lst of countres elgble to partcpate n the H-2 vsa program for temporary and seasonal workers. As Clemens put t, whle there have been cases ncludng Hat where deportaton polcy was relaxed n response to a dsaster..., endng Hat s H-2 vsa excluson s the frst tme n at least half a century that US admsson polcy 4 They may even be seen as conservatve. In the case of exco, they neglect the nduced effects of mgraton on poverty through ncreased wages for low-skll workers shra, 2007, consumpton of remttance ncome, and the fact that there s evdence of negatve selecton nto mgraton both on observables and unobservables, meanng that mgrants would on average earn less n exco f they had not mgrated than those who dd not mgrate Fernández-Huertas oraga, Presdent Bush was quoted sayng: Ths wll allow them to contnue to work here and to remt some of ther wages back home to support El Salvador s recovery efforts. The New York Tmes, arch 3,

8 was changed n response to ncreased need n the wake of a dsaster. 6 To the same extent that the nternatonal communty has repeatedly called for the rch natons to contrbute to development ad and assstance on a far bass, settng quanttatve objectves such as 0.7 percent of GDP, one may ask whether some countres contrbute more than others to global poverty reducton through ther welcomng more low-skll mmgrants from poor countres. As can be seen from Table 1, hgh-ncome countres contrbuted on average 0.2 percent of ther GDP to foregn ad n The only countres that met the 0.7 percent mark were Denmark, Luxembourg, Netherlands, Norway and Sweden. The Unted States contrbuted just 0.1 percent, the lowest fgure among the group of hgh-ncome countres. However, the US were host to 45 percent of all low-skll mmgrants orgnatng from Low-Income Countres n 2000 and receved more than 50 percent of the flow of such mmgrants between 1990 and Countres such as Australa, Canada or the UK welcomed more than twce as many low-skll mmgrants from poor countres as what an allocaton based on GDP shares would predct, and conversely for countres such as Germany, Australa or Japan 50 percent or less. 1.3 Related lterature The dea of usng market-based systems n the context of mmgraton s not new. It can be traced back at least to Gary Becker s proposal to aucton mmgraton vsas as an effectve way of screenng workers under a quantty constrant Becker, Orrenus and Zavodny 2010 and Per 2012 have recently ntegrated varants of such auctonng of vsas to employers of mmgrants rather than to the workers themselves as part of ther proposals for large mmgraton polcy reforms. However, these proposals address some of the neffcences of the current mmgraton system e.g., ts llegal and crmnal dmensons see also Aurol and esnard, 2012, or the mperfect matchng between workers and jobs, but not the type of postve nternatonal externaltes we address n ths paper. 7 On the other hand, the dea that ssung mmgraton vsas can be analogous to contrbutng to an nternatonal publc good s not fully new ether as t was frst dscussed n the case of refugees by scholars n 6 See 7 See also Djajc et al for a theoretcal framework where home and host countres decde on ther educaton polcy and vsa polcy strategcally. 6

9 Table 1: Contrbutons to ODA and to mmgraton from LICs. GDP GDP ODA Low-skll Low-skll ODA per cap. n n g. Stock Net flow share/ Countres from LICs from LICs GDP US=100 % % 2000 % % share g. stock share/ GDP share Australa Austra Belgum Canada Denmark Fnland France Germany Greece Iceland Ireland Italy Japan Luxembourg Netherlands New Zealand Norway Portugal Span Sweden Swtzerland UK US Total $24 tr. $54 b. 2.5 m. 1 m. Notes: graton data from Docquer and arfouk 2006 refer to mmgrants aged 25 or more. Low ncome countres are defned accordng to the World Bank classfcaton. 7

10 the feld of nternatonal law Schuck, 1997; Hathaway and Neve, The central concept n ths lterature was and stll s, see Hatton, 2012 that of burden sharng. However, these scholars fell short of proposng a TIQs system as they envsoned the possblty for countres to trade refugees for money on a strctly blateral bass. ore recently, De la Crox and Gosseres 2007 suggested that tradable mmgraton quotas could be used for temporary unsklled mmgrants; however, they dd not model ths proposal formally and coupled t wth a source-country market for emgraton rghts among sklled mgrants as an alternatve to a Bhagwat tax. Smlarly, Prtchett 2006 reflects on the mmgraton experence of Gulf countres to dscuss possble varants of guest-worker programs; the man tradeoffs faced are between the lack of prospects for permanent ntegraton as mmgrants are offered only temporary vsas and no poltcal rghts and the number of people that can beneft from the program thanks to the mpled turnover. The closest related paper s certanly De la Crox and Docquer 2010, who also stress that a hgher level of low-skll mmgraton than currently observed would contrbute to reduce world poverty and propose a tax-subsdy scheme to encourage rch countres to accept more low-skll mmgrants than they would unlaterally admt n a way that ensures voluntary partcpaton. The tax would consst of contrbutons to a global fund that would then be refunded through a subsdy as countres accept more mmgrants. Ther focus s on partcpaton constrants to ensure the poltcal feasblty of ther proposal. However, effcency as well as feasblty crucally hnge on a correct determnaton of the approprate tax and subsdy levels, for whch the nformatonal requrements of ther model may seen excessve. The mechansm proposed n ths paper s less demandng n terms of nformatonal requrements snce the revelaton of the opportunty costs of mmgraton that s, of countres comparatve advantage n hostng mmgrants s precsely one of ts man objectves. It may equally apply to temporary, humantaran or permanent economc mgraton. Dfferently from De la Crox and Docquer 2010, we ntally leave partcpaton constrants outsde of the model. However, our model can satsfy partcpaton constrants through the manpulaton of ntal quotas see Appendx A; ths requres knowledge about the net cost of mgraton for all the countres nvolved, an nformaton whch s unknown ex-ante but can be revealed over tme through the market. Another advantage of the mechansm we propose s that t takes nto account the fact that mgrants have preferences over locatons and that countres have preferences over the type of mgrants they receve, somethng that has been 8

11 gnored by prevous lterature. 2 Theoretcal framework There s a vast lterature on tradable quotas markets, startng wth the semnal contrbuton of Coase The lterature has manly concerned applcatons related to cap-and-trade systems to reduce polluton levels 8 and the creaton of a market for tradable emsson permts s stll proposed as one of the most promsng nstruments to address clmate change challenges Stavns, Hahn and Stavns 2010a, 2010b dscuss the pros and cons of tradable quotas and try to explan why they are so popular among poltcans. The man reason they put forward s the ablty of such schemes to acheve effcency and cost-effectveness even n the presence of well known problems such as market power and poltcal barganng. Whle n practce tradable quotas have not been used except for envronmental polcy, they have also been proposed as a potental soluton to a varety of externalty stuatons such as defct reducton Casella, 1999 or fertlty controls Bouldng, 1964; De la Crox and Gosseres, In ths secton we extend the dea of tradable quotas to mmgraton. Ths extenson requres adaptng the mechansm to the peculartes of mmgraton. We start by ntroducng tradable mmgraton quotas under the assumptons that prospectve mgrants are ndfferent between potental destnatons and that countres are ndfferent wth respect to the dentty of the mgrants and care only about ther quantty, what we call the homogenous case secton 2.1. We next extend the model frst by takng mmgrants preferences nto account secton 2.2 and then by ntroducng destnaton countres preferences over the types of mgrants they receve secton TIQs wth homogenous agents Consder that each ndvdual country faces a decson about the number of mmgrants m to let n and that the net cost of recevng these mmgrants s descrbed by the net total cost functon c m. It s a reduced form takng nto account dverse components such as the drect cost of recevng mmgrants, admnstratve costs of processng ther vsa applcatons, socal costs nherent to dversty, lower trust and socal coheson n heterogeneous populatons 8 Crocker 1966, Dales 1968 and ontgomery 1972 are the earlest contrbutors. 9

12 Alesna and La Ferrara, 2000, 2002, poltcal costs assocated wth xenophobc sentments, as well as all the economc and socal cost and benefts that mgrants may brng about e.g., the mmgraton surplus, mmgrants net fscal contrbuton, and the value for country of ts contrbuton to reducng global poverty by recevng m mmgrants. 9 It s assumed that c m s a convex dfferentable functon n the number of mgrants wth an nteror postve mnmum. Nothng s sad about the sgn of the cost functon to allow for the possblty that mmgrants are consdered ether a net burden total postve cost or are postvely valued total negatve cost by the destnaton country. In ths secton we treat mgrants as homogenous n the sense that we assume all potental mmgrants to be ndfferent between gong to any of the N possble destnaton countres, an assumpton that wll be relaxed n Secton 2.2. Immgrants are also consdered homogeneous from the pont of vew of the recevng countres. Therefore, the margnal cost of an mmgrant can be nterpreted as the expected net cost of a typcal or average mgrant. For smplcty we assume that destnaton countres can effectvely choose the number of mmgrants they want we can consder enforcement costs as another part of the reduced form cost functon c m. A destnaton country would therefore solve the followng maxmzaton problem: max m g c m 1 where g captures the externalty for country from mmgraton to other rch countres and = j m j. Gven that the externalty s lnked to the nduced reducton of poverty at the world level, we assume that g > 0. The optmal soluton s: c m NC = 0 2 where N C stands for the non-cooperatve soluton. As explaned n the prevous secton, mmgraton to one country generates a postve externalty for the other countres. Hence, the non-cooperatve equlbrum does not satsfy a general optmal level GO. To see why ths s the case, consder the global optmal problem: 9 On the mmgraton surplus and mmgrants net fscal contrbuton, see for example Borjas 1994, On prce and wage effects, see Cortes 2008 and, for a recent reassessment, Ottavano and Per

13 m NC The frst order condtons are: c m GO max {m } N =1 N [g c m ] 3 =1 + Snce g > 0, we have c m GO. The Nash soluton m NC at the world level: 10 j g j GO j = 0 = 1...N 4 GO j > 0 = c m NC and thus m GO = j g j > clearly mples a lower level of mgraton than t s desrable NC N =1 m NC < N =1 m GO GO 5 Notwthstandng partcpaton constrants dscussed below, assume now that N countres sgn a multlateral agreement, or a central authorty steps n to coordnate these countres towards a hgher level of total nternatonal mgraton decded outsde of the model such that: NC < GO That s, the agreement would go part of the way towards achevng an optmal global level of nternatonal mgraton. The problem that must be solved by ths central authorty can be wrtten as: The frst order condtons are: max {m } N =1 s.t. N [g c m ] 6 =1 N m =1 10 For a more explct modelng of the externalty n the case of refugee protecton, see Barbou des Places and Deffans 2004, Hatton 2004, 2011, Hatton and Wllamson 2004 or Bubb, Kremer and Levne In the case of mmgraton, see De la Crox and Docquer

14 c m cm + g j cm j + π = 0 = 1...N 7 j where cm stands for constraned maxmzaton and π s the multpler assocated to the constrant. It must also be true that: π N =1 m cm = 0 8 Snce g > 0 and the multpler assocated to the constrant π 0, we know as before that c m cm = j g j cm j + π > 0 = c m NC so that m cm > m NC and N =1 mcm > NC. Snce N =1 mnc = NC <, we can then be sure that π > 0, so that: = N =1 m cm 9 The optmal soluton to ths constraned maxmzaton problem equates the margnal cost of acceptng one addtonal mmgrant c m cm for a gven number of mmgrants to the shadow prce of ncreasng the sze of the program π plus the sum of the welfare gans obtaned by the other countres when country decdes to accept one addtonal mmgrant j g j cm j. If we assume that these welfare gans are common to all partcpatng countres for example, take g = β, the optmal soluton for ths constraned maxmzaton problem wll completely equalze margnal costs across destnaton countres. In the lnear example, the soluton would be: c m cm = N 1 β + π λ. Let us now assume that the above soluton s mplemented by creatng a market for mmgraton quotas that would open for a lmted tme, after whch mmgrants receve vsas to ther fnal destnatons. Under ths system, each country s assgned an ntal quota of mmgrants m 0 that can then be traded n a market n whch the prce for acceptng one addtonal mmgrant wll be represented by p. 11 The ntal dstrbuton of quotas must be agreed upon by the countres partcpatng n the multlateral agreement or establshed by a central authorty and be such that: 11 The noton of payng a prce to avod mgraton mght not appear very attractve poltcally. Ths can easly be crcumvented by wordng the proposal so that countres have two means for contrbutng to poverty reducton through mmgraton: bddng vsas, or bddng money to fund the settlement of mmgrants n thrd countres. 12

15 N = m 0 10 =1 It s assumed that the cost functons are expressed n monetary unts and that the market s compettve so that all countres behave as prce-takers we dscuss ths assumpton below. The problem that each country must solve n ths case s: mn m c m p m m 0 11 If the market s compettve, the frst order condton wll be: c m = p 12 The margnal costs of acceptng one addtonal mmgrant wll then be equalzed across destnatons through the market. In addton, the market must clear: = N m 0 = =1 N m 13 =1 Proposton 1 The market soluton to 6 s effcent. Proof. The market soluton wll be effcent m = m cm as long as t can be proved that p = λ. To see that ths s the case, suppose p λ. There are two possbltes: p < λ. From the frst order condtons n both problems, ths mples: c m < c m cm so that m < m cm for all because of the convexty of c m. But then = N =1 m < N, a contradcton. =1 mcm p > λ. Followng the same reasonng, ths mples m > m cm for all so that = N =1 m > N =1 mcm =, contradctng the soluton to the constraned maxmzaton problem. It s clear, therefore, that a TIQs system s able to replcate the constraned maxmzaton soluton and that the ntal dstrbuton of quotas only has redstrbutve consequences as long as the market s compettve. 13

16 The compettveness of a market where destnaton countres operate as buyers or sellers deserves some attenton. Casella 1999, when proposng the creaton of a market for tradable defct permts n the European onetary Unon just 12 countres at the tme, advocated establshng a computerzed contnuous double aucton mechansm, wth sellers and buyers repeatedly submttng bd and ask prces, a market desgn employed n many exchanges around the world e.g., the New York Stock Exchange. Fredman and Ostroy 1995, among others, have shown expermentally that such a mechansm converges to compettve equlbrum outcomes even n the presence of just a few large buyers or sellers. The ntuton s that the mechansm promotes a Bertrand-type competton. Coaltons are easly formed n experments but they quckly unravel towards the competttve equlbrum prce. However, authors such as uller et al suggest that the presence of a monopolst or a monopsonst can cause the prce to dverge from the compettve equlbrum case. Stll, even n these cases, they show that effcency s not affected sgnfcantly, whch can be explaned by the ablty of the monopsonst or monopolst to successfully prce dscrmnate. These results are consstent wth the emprcal observatons of Hahn and Stavns 2010a and 2010b dscussed above. Another natural queston that arses s why the market should be used to solve the externalty problem nstead of relyng on other mechansms such as taxaton. The answer follows the reasonng of Baumol and Oates The market for tradable quotas and an approprate Pgouvan tax/subsdy are equvalent n an envronment of perfect nformaton. However, f we assume that the cost functons of ndvdual countres are only known to the countres themselves, then a market for tradable quotas s superor to a Pgouvan tax/subsdy snce the tax would only ensure a certan level of margnal cost whereas the market makes sure that the fnal objectve.e., achevng the agreed upon number of mmgrants s attaned. 12 From a Coasan perspectve, we must also assume that transacton costs are low relatvely to the gans from the market. Ths s why mplementng a TIQs system s more lkely where organzatonal costs have already been ncurred for other purposes, whch s the case for example for nternatonal refugees see our frst applcaton n Secton 3, or where the potental gans at stake are huge, as s the case wth reducng global poverty. 12 Accordng to Wetzman s 1974 termnology, the margnal beneft of the externalty s perfectly nelastc so that the quanttatve restrcton the market s preferred over the prce restrcton Pgouvan tax. We could add that, n our vew, learnng one s cost functon s a dscovery process, whch s more lkely to occur n a market. 14

17 In any event, tradable quotas solutons to externalty problems need to address the challenge of partcpaton to the same extent that taxes and subsdes schemes do. De la Crox and Docquer 2010, whose mechansm s Pgouvan, concentrate on a desgn that satsfes partcpaton constrants by compensatng the ntal tax payments wth subsdes that ensure that no country s worse off. We show n Appendx A that a TIQs system can also be made ndvdually ratonal for every country through the manpulaton of ntal quotas. The problem, both for De la Crox and Docquer 2010 and for us f we were to ntroduce partcpaton constrants, s the unreallstcally large nformatonal requrement mpled by the determnaton of the rght subsdy or ntal quota for each country. Whle one can always set the ntal allocaton of quotas so that there wll only be wnners, takng partcpaton constrants nto account requres usng nformaton that countres have usually no ncentve to reveal. In the absence of such readly avalable nformaton, we can only note that countres do sgn nternatonal agreements, even f they lose, on ssues such as the envronment, refugee protecton, or ban of whales huntng, when there s a clear sense that an mportant nternatonal publc good s provded leadng to strong nternatonal pressure on ndvdual countres to sgn and there s a percepton that contrbutons are shared farly. We also note that even f just a few countres ntally agree to partcpate, they can generate a Pareto mprovement and nduce more countres to jon n the future as n Courtos and Haernguer, 2012, or Petraks and Xepapadeas, Ths process s abundantly documented n the poltcal scence and poltcal economy of nternatonal organzatons lterature e.g., Keohane, 1982; Fnnemore and Skknk, 1998; Da, 2007; Gllgan, 2009, whch descrbe t as resultng from a combnaton of normatve entrenchment, expressve behavor Hllman, 2010 and name t and shame t schemes. ore prosacally, we also suggest that the ssue of partcpaton constrants s mtgated when a TIQs system s supplemented by a matchng mechansm where countres can express ther preferences over the types of mgrants they receve, as dscussed n Secton 2.3 below. 2.2 Takng grants Preferences nto Account So far we have assumed that an nternatonal agency or a multlateral agreement determnes that mmgrants must be dstrbuted among N countres who agreed to host them accordng to some pre-determned tradable quotas. At ths pont, we have a sequence { m } N =1 15

18 of mmgrant assgnments for each of the potental destnaton countres. The problem s now to assgn ndvsble tems rghts for a mgrant to enter a gven destnaton country, or vsas to agents mgrants takng nto account the preferences of the latter. In ths sense, the problem s exactly analogous to assgnng houses to tenants Abdulkadroglu and Sonmez, The soluton proposed by Abdulkadroglu and Sonmez 1999 s the use of the top tradng cycles mechansm, whch n our case, where no mmgrant has prevous rghts to enter a partcular country, s equvalent to a random seral dctatorshp. 13 The applcaton of the top tradng cycles mechansm to the problem at hand would work as follows: 1. Each mmgrant ranks all potental destnaton countres, specfyng those to whch she would not want to go at all. 2. An orderng of mmgrants s randomly chosen from a gven dstrbuton of orderngs. 3. For any gven rankng of countres done by the mmgrants and orderng of mmgrants, assgn the frst mmgrant her frst choce, the second mmgrant her frst choce and so on untl an mmgrant chooses frst a country whose quota s flled. In that case, assgn that mmgrant her second choce or, f that one s also flled, her thrd choce and so on. If all the quotas are flled for the countres for whch the mmgrant would be wllng to go, that partcular ndvdual s taken out of the mechansm and substtuted for another one ntally out of the total number. The descrbed mechansm s ndvdually ratonal as t ensures every eventual mmgrant a vsa that s at least as good as the possblty of stayng n her orgnal country. It s also ncentve compatble no mmgrant has an ncentve to msrepresent her preferences whatever the strateges others use and Pareto effcent n the sense that there s no possblty for mmgrants to beneft from a mutual exchange of assgned vsas, at least ex-ante. In realty, t could well be that two relatves prefer to be together n a less preferred destnaton than alone n a more preferred one. To ths end, more complex matchng mechansms, such as those descrbed by Roth 2002, could be adopted to prevent famles from beng dvded nto dfferent destnaton countres. 13 See however Appendx B for an applcaton to refugee resettlement where mgrants have pre-exstng rghts. 16

19 If such a matchng mechansm s ntroduced, the problem that a central authorty would need to solve n order to mnmze the total costs of dstrbutng mgrants over N destnatons s completely equvalent to the smple maxmzaton model of the prevous secton. The soluton would just equalze the margnal costs of acceptng an addtonal mmgrant across countres. A potental dffculty arses f one of the N partcpatng destnaton countres s such an undesrable destnaton that only a fracton or none of the potental mmgrants wllng to apply for a vsa would consder gong there. In such a case, equaton 13 mght no longer be verfed. Notwthstandng ths dffculty that wll be addressed below, the planner s problem once the above-descrbed matchng mechansm s ntroduced becomes: mn {m } N =1 N =1 c m 14 s.t. N =1 m m = F m 1, m 2,..., m N = 1...N The last set of constrants embeds the matchng mechansm. The sequence {F } N =1 of functons F : [0, ] N [0, ] transforms an allocaton of vsas {m } N =1 decded by the central planner as f countres were homogenous from the mgrants perspectve nto another } N =1 allocaton { m mechansm denoted. dscussed above. that does take nto account mgrants preferences through the matchng Notce that t wll always be the case that m m, as The functons n the sequence {F } N =1 can be approxmated by dfferentable functons, for example by nterpolatng a polynomal that wll take exactly the same values where the matchng functon s defned. In such a case, the soluton to the total mnmum cost problem above can be obtaned from the followng frst order condtons: N j=1 F j c m j m j λ = 0 = 1...N 15 where λ s the multpler assocated wth the frst constrant. In ths case, the equalzaton of margnal costs across countres s just one possble soluton. For example, consder the case n whch the matchng mechansm does not alter the planner s allocaton. Ths mples: 17

20 F j m = 0 j 16 F m = 1 17 We now ask whether the ntroducton of a matchng mechansm takng mgrants preferences nto account affects the ablty of a TIQs system to replcate the soluton to the total mnmum cost problem. Ths wll depend on the desgn of the market. If countres are compensated dependng on the number of mgrants for whom they bd n the market m, a TIQs system wll be unable to replcate the total mnmum cost soluton. However, ths s not the case f they are compensated by the number of mgrants they actually receve: m. The dfference between these two compensaton schemes s what we term the mplct penalty. Ths penalty prevents countres from havng an nterest n becomng undesrable locatons. to go. Consder what happens f there are countres to whch just a few or no mgrants want Suppose that the matchng mechansm establshes an mplct penalty for those countres that are not attractve to mgrants. The key s that countres would pay dependng on the fnal outcome of the matchng mechansm m sellng decsons adopted n the market m. p m c m m 0 rather than c m for countres that turn out to be undesrable destnatons. 14 rather than on the buyng and In other words, ther objectve functon s p m m 0. Ths generates a penalty To see ths more clearly, specfy the functons F m 1, m 2,..., m N as: m j = F j m 1, m 2,..., m N = m j j m = F m 1, m 2,..., m N = m f m m 18 = m otherwse The nterpretaton s that only m ndvduals n the world are wllng to go to country even as a last resort. 14 As a practcal matter, ths penalty could be collected by the nternatonal organzaton or body n charge of settng up the market. Smlarly to the Clarke s tax lterature, the only formal requrement s that the tax cannot be redstrbuted among the partcpatng countres. 18

21 Frst, suppose country cannot affect m, the soluton denoted by a star would then be characterzed by the followng frst order condtons: c m = p f m m 19 m = m otherwse Of course, the second soluton s the nterestng one n the sense that the frst soluton smply replcates the results n secton 2.1. It must be noted that, wth m = m, we wll have c m < p. If country were able to manpulate m, t would try to ncrease t up to the pont where the margnal cost equates the prce. It would have no ncentve to decrease ts attractveness as a destnaton for mgrants snce t would then have to pay or stop earnng f m > m 0 the market prce p nstead of ncurrng a by defnton lower margnal cost for all nframargnal mgrants. If anythng, a country that s not attractve to mgrants would have ncentves to become more attractve. Indeed, suppose that a country receves an ntal quota m 0 such that c m 0 = p. For that country, t makes no sense to buy or sell n the market. But now suppose that m 0 > m. If countres were compensated accordng to market outcomes, country s cost would be c m p m 0 m 0 = c m. Because of the penalty, country s cost s c m p m m 0 > c m. The amount p m 0 m would be the penalty for not complyng wth the number of mgrants that the country was supposed to take. It s easy to see that ths result extends to any case n whch the result from the market s greater than the number of mgrants who would actually be wllng to go to country : m > m. It s also clear that the penalty remans mplct as long as countres have perfect knowledge about m see the soluton to the problem below. As t wll be seen n the applcatons, ths s not an extreme assumpton snce we can consder that the preferences of refugees or mgrants are collected before the market opens. As a way to prevent prce manpulatons e.g., one country offerng to host all the mgrants at prce 0 knowng that t would only receve m, we can even restrct partcpatng countres to bd up to m. Formally, we can nclude ths constrant n the functons F m 1, m 2,..., m N. In general terms, we can wrte the problem that a representatve country would face as: mn m c m p m m 0 + p m m 20 19

22 s.t. m = F m 1, m 2,..., m N where p m m represents the penalty. Notce that the objectve functon smplfes to c m p m m 0. The frst order condton assocated wth ths problem s: F c m m p = 0 21 Once the problem s correctly wrtten, the followng proposton holds: Proposton 2 Let 20 represent the structure of the problem solved by country partcpatng n a market for tradable mmgraton quotas. In partcular, the setup of the market s such that country pays s compensated for the actual number of mgrants receved m rather than by the number of mgrants bd n the market m, so that p m m can be consdered as a penalty assocated to the outcome of the matchng mechansm, denoted by m = F m 1, m 2,..., m N. Under ths defnton of a market for tradable mmgraton quotas, at least one of the market solutons to 14 s effcent, specfcally margnal cost equalzaton across countres. Proof. One soluton to 21 s c m = p. Substtutng p wth λ, t s easy to see that ths soluton would also solve equaton Takng Countres Preferences nto Account Countres may also have preferences over the types of mgrants they receve. Indeed, our current formulaton of the margnal cost functon c m can be nterpreted as the margnal cost over ex-ante dentcal mmgrants. In prncple, however, we could allow countres to choose mmgrants n the same way n whch we allowed mmgrants to choose destnaton countres. Before we deal wth ths ssue formally, one may want to ask whether t s an mportant one n practce. It s certanly the case that countres have preferences over the ethnc, relgous or natonal orgn of mmgrants due to common lngustc or cultural characterstcs and to shared hstores leadng, for example, to the consttuton of mgraton and daspora networks. To a large extent, ths tends to concde wth the mgrants preferences and to be reflected n ther rankng of preferred destnatons. Stll, the correspondence may be far from perfect, and recevng countres do have preferences over certan types of mgrants 20

23 such as sklled ndvduals, or ndvduals wth specfc sklls. For labor/economc mgraton n general, and even for relatvely unsklled mmgraton, t s therefore lkely that countres wll have strong preferences over the types of mgrants they receve. Ths s less lkely to be the case for the types of mmgraton we focus on n secton 3.1, that s, humantaran and refugee mgraton, as t can more easly be argued that refugees and asylum seekers are ex-ante dentcal from the vewpont of the recevng countres. But agan, ths does not mply that some countres wll not favor certan types of refugees. In any event, there are two ways of allowng for heterogenety of countres wth respect to mgrants characterstcs/types. The smplest and less nterestng one s to create a dfferent market for each type of mgrants that countres can consder. Then, the results presented so far automatcally go through wth as many markets as mmgrant types we want to consder. A second possblty s to group dfferent types of mgrants nto the same market even f they are heterogeneous. To ths end, we frst need to defne types, whch we ndex by k. Then we can redefne the total cost functon as c m where m s a vector of K elements types of mgrants denoted m k. Assumng that the cost functon s convex n each of the elements of the vector m, 15 we can recover the results proven above. Indeed, consder the total mnmum cost problem: mn { {m k } K k=1 } N =1 N =1 c m 22 s.t. N =1 K k=1 m k, m k, = F k m 1, m 2,..., m N = 1...N; k = 1...K For the soluton, we would have N K frst order condtons: N K F l j c j m k j=1 l=1 m l j λ = 0 = 1...N; k = 1...K 23 We would also have a condton assocated to the constrant: 15 The cost functons must be such that the global optmal soluton mples hgher levels of mgraton than the non-cooperatve soluton so that we can use total cost mnmzaton as an alternatve to utlty maxmzaton. For example, ths s true f we assume the cost functons are separable n the elements of the vector m. 21

24 λ N =1 K k=1 m k, = 0 24 The matchng mechansm creates a dependence between the mgrants of each type accepted by each country and those of the same or other types accepted by other countres. Stll, margnal cost equalzaton across mgrant types and across countres remans a soluton to the problem. In cases where the matchng mechansm does not affect the market outcome that s: F k m k = 1 and F l j m k c m k = 0, j, l k, the frst order condtons would smplfy to: m k, = λ = 1...N; k = 1...K 25 In terms of the market for TIQs wth heterogeneous country preferences, the formulaton of the problem that each ndvdual country would face s the followng: mn {m k } K k=1 c m K p m k m 0 + k=1 K p m k m k, k=1 26 s.t. m k, = F k m 1, m 2,..., m N k = 1...K The frst order condtons assocated wth ths problem are: K l=1 F l m k c m l p = 0 k = 1...K 27 Proposton 3 Let 26 represent the structure of the problem solved by country partcpatng n a market for tradable mmgraton quotas. In partcular, the setup of the market s such that country pays s compensated for the actual number of mgrants receved m k, rather than by the number of mgrants bd n the market m k, so that K k=1 p m k m k, can be consdered as a penalty assocated to the outcomes of the matchng mechansm, denoted by m k, = F k m 1, m 2,..., m N. Under ths defnton of a market for tradable mmgraton quotas, at least one of the market solutons to 22 s effcent, specfcally margnal cost equalzaton across countres. Proof. One soluton to 27 s c = p. Substtutng p wth λ, t s easy to see that ths m l soluton would also solve equaton

25 F k m k In partcular, when the matchng mechansm does not affect the outcome of the market = 1 and F l j m k = 0, l k, 16 t would be the only soluton: c m k m k, = p k = 1...K 28 Summng up, the heterogenety of mgrants reflected n countres preferences does not alter the effcency propertes of the market. Ths result does not extend to the matchng mechansm. If we allow countres to choose ther preferred types of mgrants, the approprate matchng mechansm s no longer equvalent to assgnng houses to tenants wthout exstng rghts. The closest analogy, nstead, s the college admssons problem, that s, the allocaton of students to colleges Gale and Shapley, 1962; Roth Our context has specfc features, however, that make t dfferent; notably, countres are assgned a quota whch they can trade and pay a penalty for the unflled slots. Stll, there s a lot to learn from the college admssons lterature for our purpose. In partcular, the matchng lterature 17 presents a very negatve result concernng the allocaton of students mgrants to colleges countres: the mpossblty of havng a stable Pareto-effcent matchng mechansm n whch countres colleges reveal ther preferences truthfully over the type of mgrants students they want. As demonstrated by Roth 1985, there exsts no stable mechansm that makes t a domnant strategy for each school to state ts preferences over the students truthfully. Despte the dfferences, ths mpossblty theorem remans true n our specfc context. 18 Beyond ths negatve result, here are two man soluton concepts that have been proposed for the college admsson problem: the school-proposng country-proposng and the studentproposng mgrant-proposng deferred acceptance algorthm. They consst of one sde of the matchng countres or mgrants submttng ther preferences and the other sde acceptng or rejectng them n as many steps as necessary untl there are no mgrants or countres left to assgn. Both algorthms converge to the most desrable outcome for countres and mgrants respectvely. In a recent paper, Azevedo and Leshno 2012 show that, usng a contnuum to approxmate the number of mgrants wth a fnte number of countres, both algorthms converge to the same unque stable soluton. In the context of the market for TIQs and ndependently of whether a country-proposng or a mgrant-proposng deferred acceptance algorthm s ntroduced, the market wll equally 16 In addton, ths mplctly assumes no lmts on the number of potental mgrants of each type. 17 See the revew by Sonmez and Unver The exstence of the market does not nvaldate the counter-example n Roth

26 tend to margnal cost equalzaton across mgrant types and quanttes. Thus, the externalty that justfes the creaton of the market n the frst place s correctly addressed. What changes are the desrable propertes of the fnal outcome for both mgrants and countres and these desrable propertes wll be assocated to the choce of the matchng mechansm. For example, Sonmez and Unver 2011 report how a desgn based on the student-proposng deferred acceptance algorthm was ntroduced n the hosptal-ntern entry labor market n the US on the bass of the fact that ths algorthm ensures truth-tellng by the nterns whereas the college-proposng algorthm does not ensure truth-tellng even by the hosptals. On the contrary, a recent contrbuton by Pathak and Sonmez 2011 defnes a new concept of manpulablty and supports a college-proposng deferred acceptance mechansm on the grounds that t s less manpulable than the student-proposng one. 19 As ths ssue s stll unsettled n the matchng lterature, we just pont out to the fact that any mechansm s compatble wth our market for TIQs. We fnd the manpulablty concept n Pathak and Sonmez 2011 convncng and would suggest the use of a countryproposng deferred acceptance mechansm. Stll, allowng for takng nto account countres preferences over mgrants types could ntroduce a trade-off. On the one hand, as we show below, cost-effcency s ncreased n the sense that countres mght be wllng both to take more mmgrants n, and to partcpate more n the market f they can pck ther preferred mmgrants and even more so f the matchng mechansm employs an algorthm that favors them. On the other hand, the matchng mechansm could create addtonal uncertanty f we assume that the uncertanty assocated to not knowng ex ante whch types of mgrants wll be receved and dong nothng about t as n subsectons 2.1 and 2.2 s smaller than the uncertanty assocated wth dong somethng about t by tryng to select preferred mgrants through the matchng mechansm. In prncple, one could argue that uncertanty could also be reduced by the possblty of choce but ths s ultmately an emprcal queston. Theoretcally, however, t s relevant to consder ths ssue because hgh uncertanty s known to affect transacton costs and decrease the desrablty of tradable quotas Hahn and Stavns, 2010a Under some condtons, Azevedo and Budsh 2012 argue that deferred acceptance algorthms are strategy-proof n the large, meanng that the ncentves to manpulate them dsappear as the sze of the matchng market grows. 20 We do not attempt to model ths problem n a dynamc settng, whch would allow for updatng of belefs over the mmgrants mx to be receved. 24