Document de treball de l IEB 2009/8

Document de treball de l IEB 2009/8 IMMIGRANT WAGES IN THE SPANISH LABOUR MARKET: DOES THE ORIGIN OF HUMAN CAPITAL MATTER? Esteban Sanromà, Raúl Ramos, Hpólto Smón Ctes and Innovaton

Documents de Treball de l IEB 2009/8 IMMIGRANT WAGES IN THE SPANISH LABOUR MARKET: DOES THE ORIGIN OF HUMAN CAPITAL MATTER? Esteban Sanromà, Raúl Ramos, Hpólto Smón The IEB research program n Ctes and Innovaton ams at promotng research n the Economcs of Ctes and Regons. The man objectve of ths program s to contrbute to a better understandng of agglomeraton economes and 'knowledge spllovers'. The effects of agglomeraton economes and 'knowledge spllovers' on the Locaton of economc Actvtes, Innovaton, the Labor Market and the Role of Unverstes n the transfer of Knowledge and Human Captal are partcularly relevant to the program. The effects of Publc Polcy on the Economcs of Ctes are also consdered to be of nterest. Ths program puts specal emphass on appled research and on work that sheds lght on polcy-desgn ssues. Research that s partcularly polcy-relevant from a Spansh perspectve s gven specal consderaton. Dssemnatng research fndngs to a broader audence s also an am of the program. The program enjoys the support from the IEB- Foundaton. The Barcelona Insttute of Economcs (IEB) s a research centre at the Unversty of Barcelona whch specalzes n the feld of appled economcs. Through the IEB- Foundaton, several prvate nsttutons (Caxa Catalunya, Aberts, La Caxa, Gas Natural and Applus) support several research programs. Postal Address: Insttut d Economa de Barcelona Facultat d Economa Empresa Unverstat de Barcelona C/ Tnent Coronel Valenzuela, 1-11 (08034) Barcelona, Span Tel.: + 34 93 403 46 46 Fax: + 34 93 403 98 32 eb@ub.edu http://www.eb.ub.edu The IEB workng papers represent ongong research that s crculated to encourage dscusson and has not undergone a peer revew process. Any opnons expressed here are those of the author(s) and not those of IEB.

Documents de Treball de l IEB 2009/8 IMMIGRANT WAGES IN THE SPANISH LABOUR MARKET: DOES THE ORIGIN OF HUMAN CAPITAL MATTER? * Esteban Sanromà, Raúl Ramos, Hpólto Smón ABSTRACT: The am of ths paper s to analyse the role played by the dfferent components of human captal n the wage determnaton of recent mmgrants wthn the Spansh labour market. Usng mcrodata from the Encuesta Naconal de Inmgrantes 2007, the paper examnes returns to human captal of mmgrants, dstngushng between human captal accumulated n ther home countres and n Span. It also examnes the mpact on wages of the legal status. The evdence shows that returns to host country sources of human captal are hgher than returns to foregn human captal, reflectng the lmted nternatonal transferablty of the latter. The only excepton occurs n the case of mmgrants from developed countres and mmgrants who have studed n Span. Whatever ther home country, they obtan relatvely hgh wage returns to educaton, ncludng the part not acqured n the host country. Havng legal status n Span s assocated wth a substantal wage premum of around 15%. Lastly, the overall evdence confrms the presence of a strong heterogenety n wage returns to dfferent knds of human captal and n the wage premum assocated to the legal status as a functon of the mmgrants area of orgn. JEL Codes: J15, J24, J31, J61 Keywords: mmgraton, wages, human captal Esteban Sanromà Unverstat de Barcelona Facultat d Economa Empresa Avda. Dagonal, 690 08034 Barcelona (Span) E-mal: esanroma@ub.edu Raúl Ramos Unverstat de Barcelona AQR-IREA, IZA Avda. Dagonal, 690 08034 Barcelona (Span) E-mal: rramos@ub.edu Hpólto Smón Unversdad de Alcante Facultad de Cencas Económcas Campus San Vcente del Raspeg 36900 San Vcente del Raspeg (Alcante) E-mal: hsmon@ua.es * Esteban Sanromà, Raúl Ramos and Hpólto Smón wsh to express ther grattude for support receved from INSIDE Span and from the Spansh Mnstry of Educaton and Scence through the followng projects: SEJ2007-65086/ECON, SEJ2005-04348/ECON and SEJ2007-67767-C04-02, respectvely.

1. Introducton Human captal has been consdered a key factor n the determnaton of ndvdual wages and ther growth over tme (Card, 1999; Psacharopoulos and Patrnos, 2002). Consstent wth ths perspectve, the analyss of the stuaton of mmgrants wthn ther host countres labour markets has focused on ther human captal as well. The two man emprcal results reached from several decades of academc effort regardng wages of mmgrants the presence of a sgnfcant ntal wage gap relatve to natve-born workers and the rapd wage growth from the moment of arrval can bascally be explaned by ther human captal. Thus, the wage dsadvantage experenced by mmgrants when they arrve n a new country can generally be attrbuted to the lmted transferablty of the human captal they have acqured n ther home country. The reason may le n the lower qualty of the educatonal system there or n ther nsuffcent destnaton language sklls. Whatever the case may be, the relevant fact s that newly arrved mmgrants lack suffcent human captal for ther host country s labour market (Chswck, 1978; Chswck and Mller, 1985, 2007; Fredberg, 2000). On the other hand, the explanatory factor behnd the rapd growth over tme n mmgrant wage levels can be found n ther accumulaton of dfferent types of human captal n the host country, whch s partcularly sgnfcant n the frst years of resdence n the host country. It s noteworthy that t s ths rapd growth n wage levels that generally leads to wage assmlaton wth the natve populaton (nter ala, Chswck, 1978; Baker and Benjamn, 1994; Chswck and Mller, 1995 and Bell, 1997). Concernng the general analyss of mmgrant wages n host countres and the role played by human captal as explanatory factor, major advances have been made by dfferentatng the effect of the dfferent components of human captal. Consequently, studes focusng on mmgrants wage returns to schoolng have shown the relevance of dstngushng between educaton completed n home and n host countres, gven that ther wage effects dffer sgnfcantly (Schaafsma and Sweetman, 2001; Bratsberg and Ragan, 2002; Ferrer and Rddell, 2003). Moreover, ther fndngs suggest that wage returns to educaton also dffer across home countres and that, n partcular, the level of economc development of the countres postvely affects the transferablty of studes completed there (Bratsberg and Ragan, 2002). In addton, analyses that have addressed non-lnearty n mmgrants years of schoolng show that age and the educatonal level attaned could be sgnfcant at the tme of emgratng (Ferrer, Green and Rddell, 2006; 2

Hartog and Zorlu, 2009). Smlarly, t has been also consdered necessary to separate years of foregn experence from years of experence obtaned n the host country, as long as returns to the former are generally zero or at least consderably lower than the latter (Chswck and Mller, 1985; Kossoudj, 1989; Fredberg, 2000; Schaafsma and Sweetman, 2001). Along the same lnes, the recent lterature on wage progress n host countres and the process of assmlaton (thrd-generaton models) also supports the usefulness of breakng down educaton and experence nto the components related to home and host countres (Fredberg, 2000; Skuterud and Su, 2008; Clark and Lndley, 2009). Dong so, the lmtatons of the frst emprcal models, n whch usng the varable years snce mgraton dd not allow to dstngush among dfferent types of actvtes pursued after the mmgrant s arrval, can be overcomed. An addtonal mportant advance n a few recent studes has been to hghlght the relevance of effectve work experence, gven that the accumulaton of human captal specfc to the new country s not necessarly the same f the mmgrant holds employment or s jobless (Chswck, Lee and Mller 2005; Skuterud and Su, 2008; Galloway, 2008). The man objectve of the paper s to analyse the role played by the varous components of human captal on mmgrant wages n the Spansh labour market and, n partcular, on the relevance of the dfferent orgn of human captal -home vs. host country-, an ssue, that to our knowledge, has not been studed before. Studyng mmgraton n the Spansh labour market s a matter of great nterest, because Span has become n a relatvely short perod of tme a country wth sgnfcant mgraton flows n the nternatonal context (OECD, 2008). Span ranks second among OECD countres after the Unted States n absolute numbers of annual mmgraton, and t stands thrd after Luxembourg (41.6%) and Swtzerland (20.3%) n percentage of foregners out of the entre populaton (10.3%). Those fgures put Span ahead of all other European Unon members. The growth of the stock of mmgrants was partcularly ntense between 1995 and 2007, rsng steeply from 542,300 foregners (1.4% of the populaton) n 1995 to 5,268,800 (11.4%) n 2007. Gven the magntude of the phenomenon, extensve research s needed to devse strateges and mmgraton polces to guarantee economc well-beng and socal stablty. In the sense, the factors explanng mmgrant wages are of specal nterest, partcularly returns to ther endowments of human captal, whch s ther prmary and, n many cases, only asset. The recent nature of mmgraton n Span, however, has 3

made t dffcult to obtan approprate statstcal nformaton, whch has n turn lmted and condtoned studes on mmgraton n the Spansh labour market. Consequently, because of the absence of wage data, Amuedo-Dorantes and de la Rca (2007), Fernández and Ortega (2008) and Sanromá, Ramos and Smón (2005, 2008) have analysed mmgrant assmlaton n terms of employment status, occupaton, contract type and over-educaton. In a smlar ven, Smón, Sanromá and Ramos (2008) analysed dfferences n wages structures between natve and mmgrant workers usng mcrodata drawn from the 2002 Spansh Wage Structure Survey (Encuesta de Estructura Salaral), whch do not nclude nformaton on the tme of arrval n Span (and consequently on educaton and work experence acqured n the home country). Last, Izquerdo, Lacuesta and Vegas (2009) have analysed assmlaton of mmgrants usng the Contnuous Sample of Workng Lves (Muestra Contnua de Vdas Laborales) whch, n addton to lackng the year of arrval n Span, used the capped earnngs from Socal Securty as a proxy for wage levels. The current paper overcomes earler lmtatons by makng use of mcrodata from the Spansh Natonal Immgrant Survey 2007 (Encuesta Naconal de Inmgrantes 2007 hereafter, ENI-) -), whch was conducted by the Spansh Natonal Statstcs Insttute between 2006 and 2007. The ENI collected a wde range of statstcal nformaton about mmgrants, ncludng wages. Moreover, t allows to dstngush, followng the suggestons n recent lterature, between educaton completed n home and host countres and to break down years of experence between home and host countres as well. In addton, the ENI enables a good approxmaton of effectve work experence n Span to be calculated, as well as years wthout employment (dle years). Lastly, t also provdes nformaton on mmgrants home countres, whch facltates the estmaton of returns to human captal by economcally dfferent areas of orgn. The ENI comprses a sngle cross-secton, whch rules out longtudnal analyss. It also mpedes the constructon of a pseudo-panel combnng nformaton from dfferent cross-sectons, a hghly useful approach n the lterature snce the contrbuton of Borjas (1985). However, workng wth a cross-secton can lead to bas n the estmaton of returns to human captal. The bas could result from three dfferent causes: changes n the composton or qualty of the mmgrants arrvng at dfferent ponts n tme (Borjas, 1985, 1995); the effect of the busness cycle on the wages of ndvduals enterng the labour market at dfferent tmes (Aslund and Rooth, 2007); and the exstence of return mgraton (or onward mgraton to a thrd country) (Constant and 4

Massey, 2003; Dustmann and Wes, 2007; Lubotsky, 2007). To mnmse these problems, the emprcal analyses have been conducted on mmgrants arrved n Span between 1997 and 2007. The selecton of ths specfc group of mmgrants s ntended to address the three problems set out above. Frstly, mmgrants pror to 1997 mostly came from developed countres or from a number of relatvely advanced Latn Amercan countres, whle the bulk of the mmgrant populaton arrvng subsequently have come from Latn Amerca (bascally the Andean countres), Eastern Europe and Afrca (Reher et al., 2008). Workng wth entre ENI sample would have nvolved a change n the qualty and composton of the mmgrants arrvng at dfferent ponts n tme, whle the selecton actually used consderably reduces ths problem. Moreover, n order to control the heterogenety of the mmgrant populaton more effectvely, separate estmatons have been performed by regon of orgn. In a smlar ven, n the nterests of greater homogenety, mmgrants wth Spansh natonalty from brth have been excluded. Secondly, the perod 1997-2007 s a homogeneous perod of sustaned growth and ntense job creaton, wth the consequence of mnmsng the persstent effects of the economc cycle on wages. Lastly, t would seem reasonable to argue that return mgraton s not quanttatvely sgnfcant when workng solely wth a perod of economc expanson. The busness cycle change observed n 2008 does appear to have resulted n rsng levels of return mgraton and the measures encouragng return mgraton approved by the government durng that year seem to be another factor at work. However, as the ENI was conducted n late 2006 and early 2007, the data should not be affected by ether the cyclcal change or government nterventon 1. The results obtaned show sgnfcant dfferences n returns to dstnct components of mmgrants human captal. Of partcular mportance s the orgn of human captal. In general, returns to schoolng n Span exceed returns to foregn schoolng, except n the case of mmgrants from developed countres. Along the same lnes, earler educaton pursued by mmgrants who have then completed ther schoolng n Span present notably hgher returns, whch are greater n all cases than returns for mmgrants who have only studed n ther home country. In addton, wage progress occurs for mmgrants as a functon of the length of ther stay n Span, because Spansh experence s more valuable than experence abroad, whch has a lmted transferablty (agan, except n the case of mmgrants from developed countres). In general, the 1 It should be noted, however, that ths soluton s not optmal. As a result, the results presented n the paper must be nterpreted wth cauton. 5

analyss shows apprecable dfferences n returns to human captal and the pace of wage progress as a functon of the area from whch mmgrants come. The dfferences are sgnfcant when comparng mmgrants from developed and less developed countres, but also when comparng the dfferent geographc areas to whch the less developed countres belong. The dfferences appear to depend on ther economc and cultural dstance from Span. Lastly, legal status s assocated wth a substantal wage premum. The remander of the paper s structured n three parts. Below, the second secton sets out the prncpal characterstcs of the database used n the emprcal analyss. It also descrbes how the prncpal varables of nterest have been constructed n the study. The thrd secton descrbes the methodology appled and shows the results obtaned. Lastly, the fourth secton summarses the man conclusons of the paper. 2. The Natonal Immgrant Survey 2007 The ENI (Encuesta Naconal de Inmgrantes 2007) s a survey prepared by the Spansh Natonal Statstcs Insttute n order to obtan detaled nformaton on the nternatonal nature of mmgraton n Span, supplementng nformaton gathered from regular sources of data (such as the Padrón Muncpal, the Encuesta de Varacones Resdencales, the Encuesta de Poblacón Actva o the Censo de poblacón), whch provde partal nformaton on the characterstcs of mmgraton. The scope of the ENI covers all of the natonal terrtory of Span and the data collecton was conducted between November 2006 and February 2007 based on the Padrón Muncpal, usng the week pror to the ntervew as the reference perod 2. The orgnal survey sample comprses approxmately 15,500 ndvduals. The ENI provdes detaled nformaton on the socodemographc characterstcs of mmgrants (e.g., age, gender, natonalty, country of brth, martal status, legal status, knowledge of languages and year of arrval n Span) and on ther current work stuaton (as well as nformaton on the characterstcs of ther frst job n Span, although to a lesser extent than ther current job). The range of questons on mmgraton covered by the survey s very wde comprsng, among others, mmgrant household structure and accommodaton characterstcs; famly and socal networks; prevous stuaton n ther 2 More detaled nformaton on the contents of the ENI, the sample desgn and the data collecton procedure used s avalable at the web page of the Natonal Statstcs Insttute (www.ne.es). 6

home countres and ther current relatonshp to those countres, and varous aspects of ther mgraton experence. The ENI defnes mmgrants as any ndvduals born abroad (regardless of whether they have Spansh natonalty or not) who at the tme of dong the ntervew had reached at least 16 years of age and had resded n a home for a year or longer (or, alternatvely, n the case of ndvduals wth less than one year s resdence n Span, had the ntenton to reman here for at least a year). The only excepton s ndvduals born outsde Span who have possessed Spansh natonalty from brth, but had not reached two years of age by the tme of arrval n Span. In that case, Span was consdered as ther country of orgn. Ths defnton of mmgrant meant, among other crcumstances, that ndvduals born abroad but wth Spansh natonalty are consdered mmgrants, whle foregn natonals born n Span are not. Hence, ths approach excludes ndvduals born n Span of foregn mmgrants, even f ther natonalty s not Spansh. It also excludes Spansh emgrants who have returned to Span 3. Immgrants wages are expressed n monthly terms and correspond to the pay receved n ther prncpal job n net terms (.e., after deductons, contrbutons and other related payments), ncludng the proportonal monthly part correspondng to extraordnary payments and other extraordnary ncome receved on a regular bass 4. If surveyed ndvduals chose not to provde the exact value of ther wages durng the ENI ntervew, they were gven the opton to dentfy ther wages answerng a closed nterval queston. For ndvduals provdng alternatve nformaton of ths sort, ther wages have been calculated at the mdpont of the correspondng wage nterval. It should be noted that the total number of ndvduals choosng ths approach consttuted only about 15% of the effectve sample of wage earners used (Table A.1) and the general results of the emprcal analyss are robust to the excluson of ths group. As emphassed n the ntroducton, the central aspect examned n ths research s how dfferent forms of human captal affect mmgrant wages. Ths requres 3 Usng a defnton based on country of brth contrasts wth the alternatve defnton based on natonalty, whch has generally been used n prevous studes on mmgraton and ts effects on the Spansh labour market (see, for example, Amuedo-Dorantes and la Rca, 2007; Carrasco, Jmeno and Ortega, 2008; and Smón, Sanromá and Ramos, 2008). 4 Even when wages correspond to the man job, a dummy varable has been ntroduced nto the emprcal estmatons, whch measures whether the ndvdual has more than one job or not. The purpose s to control for any possble effect on wages. Wth a few exceptons, the varable tends to have a negatve coeffcent and to be statstcally sgnfcant. 7

dfferentatng between educaton and work experence and knowng whether these types of human captal have been acqured n the mmgrants home countres or n Span. As the ENI lacks precse nformaton on the age at whch mmgrants have fnshed ther schoolng, the breakdown of human captal nto foregn and domestc components s based on a standard approach n the lterature. Therefore, after recodng nformaton on schoolng levels nto years 5, the approxmaton assumes that the perod of educaton has been pursued contnuously from the frst year of entry nto the school system at the age of sx untl the ndvdual reaches the maxmum declared level. Ths makes t possble to approxmate the age at whch schoolng s concluded as the total number of years n educaton plus sx 6. Potental work experence corresponds to the dfference between the ndvdual s age and the age at whch schoolng was concluded. In addton, knowng the year of arrval n Span makes t possble to dfferentate whch porton of an mmgrant s human captal has been acqured n the home country and n Span, n the case of both educaton and work experence. Moreover, Spansh potental work experence can be dfferentated nto effectve work experence and dle years 7. Wth respect to the labour market n the home country, the nformaton contaned n the ENI s nsuffcent to calculate effectve work experence. For ths reason, only a varable that measures whether the mmgrant has worked n the home country at some tme pror to emgratng to Span (and consequently whether he has effectve work experence pror to arrval) has been calculated. 5 Followng common practce, years of schoolng have been measured as follows: 0 years for ndvduals wthout any formal educaton; 3 years for ncomplete prmary educaton; 6 years for prmary educaton; 10 years for completng lower secondary educaton; 12 years for completng hgher secondary educaton; 15 years for the frst cycle of unversty educaton; and 17 years for the second cycle of unversty educaton. 6 In the absence of nformaton on the exact age of fnshng schoolng, ths approach reflects a standard approxmaton n the lterature (see, for example, Fredberg, 2000). In order to nterpret the results, however, t s necessary to bear n mnd that ths fgure tends to overestmate (underestmate) the years of schoolng n the home country (host country). Skuterud and Su (2008) provde a thorough revew of the varous approaches used to calculate foregn and domestc human captal and determne ther nfluence on emprcal estmatons. 7 The dstncton rests on several assumptons, based on the varables avalable n the ENI whch measure whether the mmgrant stll holds the frst job obtaned n Span; the tme requred to obtan the frst job; the number of tmes unemployed; whether the mmgrant has been unemployed for more than a month snce arrvng n Span; and the longest perod of unemployment. In ths way, the calculaton of effectve experence for all mmgrants has broadly dscounted tme requred to obtan the frst job from potental work experence. In the specfc case of mmgrants who do not reman n ther job, tme spent unemployed has also been dscounted by multplyng the longest perod of unemployment by the number of tmes unemployed n Span (gven that the length of the longest perod of unemployment s measured usng ntervals, then the mdpont of the correspondng nterval has been used n the calculaton). The exact defnton of the varables used n the analyss s avalable from the authors on request. 8

In order to break down the nformaton by area of orgn, mmgrants have been grouped by country of brth, dstngushng between developed and less developed countres. Developed countres nclude the EU-15 countres, Norway, Swtzerland, Iceland, Cyprus, Malta, the small European prncpaltes, the Unted States, Canada, Israel, Japan, Australa and New Zealand. All other countres have been consdered less developed, dstngushng three man areas: Latn Amerca, Eastern Europe and the rest of the world. In addton, a varable has been devsed to capture mmgrants legal status, reflectng whether or not they have the permts to become legally contracted employees under current Spansh law. The varable s dchotomous and reflects whether mmgrants state that they have any of the followng documents: permanent resdency authorsaton; temporary resdency authorsaton 8 ; EU resdence permt (except n the case of Romanan and Bulgaran workers who, despte beng EU ctzens, could not become legally contracted workers n Span temporarly at the tme of the ENI); refugee status or asylum applcaton. Ths category also ncludes mmgrants whose natonalty s Spansh, from other EU member state (excludng Bulgara and Romana) or from non-eu members of the European Free Trade Assocaton (.e., Lechtensten, Iceland, Swtzerland and Norway). Alternatvely, mmgrants not consdered to have documentaton to work legally as contracted employees nclude Romanan and Bulgaran resdents; mmgrants wth student vsas; mmgrants who have resdency applcatons pendng or have not yet submtted ther applcatons; mmgrants who state that they have none of the documents lsted above and mmgrants who respond that they do not know whch documents they possess. Other varables employed n the emprcal analyss nclude monthly workng hours (calculated based on usual weekly workng hours), gender, martal status, controls by regon, the number of chldren n the household, and Spansh profcency s a dchotomous varable whch takes a value of 1 for ndvduals whose mother tongue s Spansh or, f not, who state that they can speak Spansh well or very well. Observatons have been excluded from the orgnal sample for ndvduals wth lackng nformaton concernng the varables of nterest; ndvdual who are under 16 or 8 In the case of temporary resdency authorsaton, the mmgrant does not necessarly receve a work permt n Span. The mmgrant s applcaton and the ssung of a work permt by the authortes are dscretonary. However, the overall results of the emprcal analyss do not show sgnfcant varatons, regardless of whch of the two categores of the dchotomous varable on legal status s assgned to mmgrants n ths stuaton. 9

over 65 years of age; ndvduals wth net monthly wages below 200 euros or usual weekly workng hours less than 10 hours or more than 90. The fnal sample also excludes mmgrants wth Spansh natonalty at brth and mmgrants who arrved before 1997. The fnal sample comprses 4,885 mmgrants workng as employees. Table A.1 n the annex shows some descrptve statstcs for the ENI sample. The characterstcs of the mmgrants covered by the survey generally ft the profle charactersng recent mmgraton to Span. Wthout ntendng to be exhaustve 9, the mmgrants n the sample are bascally from less developed countres (92.2%), partcularly Latn Amerca (52.9%) and Eastern Europe (26.3%). They are mostly men (53.4%) and of ntermedate age (the average age s approxmately 34 years old). The mmgrants n the sample also acqured the bulk of ther human captal n ther home country (10.95 of ther 11.1 years of educaton, on average, corresponds to ther home countres, and n the case of potental work experence, 12.67 of the 16.87 years on average). For the most part, ther length of stay n Span (4.35 years on average) leads to ther accumulatng potental work experence (4.2 years of whch 3.57 correspond to effectve work experence and 0.63 reflects dle years), whle ther accumulaton of educaton s very lmted (only 0.15 years). Ths largely reflects the low percentage of mmgrants who have completed schoolng n Span. Immgrants who have studed n Span make up 5.5% of the total sample. They are typcally younger and have arrved n Span at an earler age than other mmgrants. They have slghtly hgher wages (1,017 euros). They come largely from developed countres and Latn Amerca. They have lower levels of work experence n both ther home and host countres, and they possess hgh endowments of educaton, averagng 12.2 schoolng years, of whch 2.66 correspond to years of study n Span. The vast majorty of the mmgrants have fluency n Spansh (82.9%), legal status to work (87.5%), and prevous work experence n ther home country (85.2%). In addton, the overall group of mmgrants shows a strong heterogenety n ther characterstcs as a functon of area of orgn. For example, whle the average monthly net salary s 995 euros, t reaches 1,402 euros for mmgrants from developed countres and 960 euros for mmgrants from less developed countres. Along the same lnes, notable dfferences can be observed n educatonal endowments (wth comparatvely lower endowments for mmgrants from the rest of the world); n legal 9 For a more detaled descrpton of the mmgrants n the ENI, read the report Informe Encuesta Naconal de Inmgrantes (ENI-2007) that s avalable at the web page of the Natonal Statstcs Insttute (www.ne.es). 10

status (wth a less stable stuaton for mmgrants from Eastern Europe); and n fluency n Spansh (partcularly low for mmgrants from Eastern Europe or from the rest of the world). 3. Emprcal evdence The frst model used n ths study to analyse mmgrants wages s a semlogarthmc Mnceran wage equaton wth the form: w = α + δ ysm + β potexp + γ X + ε (1) f 1 sch + β2 potexp + β3 f 2 where w corresponds to the wage logarthm for ndvdual, the varable ysm ndcates the number of years snce arrval n the host country, the varable sch represents the number of years of schoolng and the varable potexp f denotes the number of years of foregn potental experence whch s squared, as s usual n the lterature. X s a vector that represents other ndvdual characterstcs whch have an nfluence on wages, whle ε s a random error term. Chswck, Lee and Mller (2005) have estmated equaton (1) for a sample of mmgrants to Australa. Ther results show a sgnfcant, postve effect on wages from the amount of tme that has passed snce arrval n Australa. The sgnfcant economc progress acheved by mmgrants n ther study s consstent wth the hypothess of wage assmlaton. Ths specfcaton has also been used n varous other studes on mmgrant wages (see, for example, Accetturo and Infante, 2008). Table 1 shows the result of estmatng equaton 1 usng ordnary least squares on the mmgrant sample descrbed n the prevous secton, treatng the logarthm of monthly wages as the endogenous varable. In addton to human captal, the remanng varables capturng mmgrant characterstcs nclude gender, martal status, legal status n Span, geographc area of brth, and regon of resdence 10. As can be seen n the frst column of Table 1, the results obtaned show that years snce mgraton have a postve and sgnfcant effect on mmgrant wages. More 10 Another controls ncluded n the regresson are monthly workng hours (n logarthms), f the mmgrant holds more than one job and f he has provded wage nformaton accordng to predetermned bands. The complete results for all estmatons presented n the paper are avalable from the authors on request. 11

specfcally, each year of Spansh resdence ncreases wages ceters parbus n a 1.4%, an ncrease that could be nterpreted as evdence supportng the noton of economc progress for mmgrants: a longer perod of resdence n Span mples an mprovement wth respect to the ntal wages. Although the returns are dmnshng, a year of potental experence n the home country also has a postve, but modest, effect on mmgrant wages. The results form estmatng equaton (1) also show that schoolng have a postve and sgnfcant effect on mmgrant wages. More specfcally, each year of schoolng results n a wage ncrement of 1.8%. In comparatve terms, the fgure s notably lower than returns to schoolng for natve workers, whch accordng to estmatons obtaned from the Wage Structure Survey 2006 (EES) are approxmately 4% 11. As ndcated by Chswck (1978), Chswck and Mller (1985) and Fredberg (2000), among many others, the lower returns to schoolng for mmgrants could be explaned by the lower (real or perceved) qualty of mmgrants educaton, an mperfect transferablty of ther human captal or an nsuffcent command of the host country language. One result whch s new to the Spansh lterature on mmgraton s the possblty of quantfyng the wage premum receved by mmgrants who work wth the necessary permts. The results n the frst column of Table 1 show that the gap s 15.1% between mmgrants wth documentaton and mmgrants of smlar characterstcs who lack documentaton. Accetturo and Infante (2008) have found a somewhat hgher wage gap (roughly 20 percent ponts) n the case of Lombardy. In the Spansh case, lower remuneraton must be related to the characterstcs of unsklled agrcultural and constructon jobs to whch most mmgrants wthout work permts must turn for employment. Another factor s ther lower barganng power. A thrd factor n explanng the gap may well be the greater adaptablty of legal mmgrants to the Spansh labour market. 11 Gven the lack of recent emprcal lterature on the returns to schoolng n Span, the estmaton mentoned above has been obtaned by usng the Wage Structure Survey 2006 (EES) and controls smlar to the ones ncorporated nto the estmaton based on the ENI. Comparsons between the two sets of results should be made wth cauton as the two surveys reflect statstcal operatons wth dstnct purposes and methodologes. For example, the ENI s a household survey, whle the EES s a company survey gatherng data on employees. In addton, the wage concepts used n the two surveys dffer. Whle the ENI provdes nformaton on net wages, the EES gves nformaton on gross wages. Nonetheless, the estmatons of returns to schoolng are very smlar for mmgrants n both cases: 1.8% accordng to the ENI and 2.0% accordng to the EES. 12

Table 1 Logarthm of monthly wages Model (1) Model (2) Model (3) Model (3) Schoolng n Span Model (3) No schoolng n Span Legal status 0.141*** 0.141*** 0.142*** -0.0709 0.150*** [0.0152] [0.0156] [0.0155] [0.101] [0.0157] Years snce mgraton 0.0141*** [0.00250] Schoolng years 0.0177*** [0.00166] Schoolng years n Span 0.0332*** 0.0369*** 0.0871*** [0.00747] [0.00746] [0.0218] Schoolng years n home country 0.0176*** 0.0172*** 0.0445*** 0.0160*** [0.00166] [0.00166] [0.00869] [0.00170] Potental experence n Span 0.0140* [0.00817] Potental experence n Span 2 9.76E-06 [0.000879] Effectve experence n Span 0.0239*** -0.00387 0.0206*** [0.00736] [0.0353] [0.00769] Effectve experence n Span 2-0.00094 0.00734-0.000665 [0.000863] [0.00646] [0.000892] Idle years n Span -0.0036-0.0256-0.00414 [0.00501] [0.0184] [0.00524] Potental experence n home country 0.00676*** 0.00679*** 0.00622*** 0.0163** 0.00535*** [0.00177] [0.00177] [0.00176] [0.00737] [0.00182] Potental experence n home country 2-0.000190*** -0.000191*** - 0.000185*** -0.000380* -0.000166*** [5.08e-05] [5.10e-05] [5.04e-05] [0.000202] [5.22e-05] Labour experence n home country 0.0313** -0.106 0.0301** [0.0142] [0.0724] [0.0143] Number of observatons 4.885 4.885 4.885 271 4.614 Adjusted R 2 0.453 0.453 0.456 0.582 0.455 Notes: OLS estmates wth controls related to gender, cvl status, geographcal area of brth, monthly worked hours, havng more than one job, the regon of resdence and Heckman s lamda. ***, ** and * ndcate that the estmated coeffcent s statstcally dfferent from zero at 1%, 5%, and 10% sgnfcance levels, respectvely. 13

The results obtaned from the remanng controls reflect what has typcally been found n the lterature. There s a favourable wage dfferental for men n relaton to women and there s a wage premum for marred mmgrants. In addton, wage dfferences are sgnfcant by area of orgn as a functon of the economc and cultural dstance of each area from Span. In partcularly, there s a negatve dfferental of 23.3% for mmgrants from Latn Amerca wth respect to mmgrants to developed countres. It s 20.9% for mmgrants from Eastern Europe and 29.7% for mmgrants from the rest of the world, who therefore face the most severe wage gap n the Spansh labour market. A fnal ssue to emphasse s that the estmaton of the model has taken nto account the possble exstence of bas n employment selecton. As a result, the twostage procedure proposed by Heckman has been appled. The results obtaned after applyng the frst stage of the procedure s show n Table A.2 n the annex. The varables ncluded n the dscrete-choce selecton model, whch act as excluson restrctons, have been the number of chldren lvng n the household and profcency n Spansh. Heckman s lambda (obtaned from prevous results as the nverse Mlls rato) has been ntroduced as another explanatory varable n equaton (1). Yet, nether ths model nor any of the other estmated models has provded favourable evdence on the statstcal sgnfcance of ths varable. Ths s a common result n the mmgraton lterature and could be explaned by the lqudty restrctons of recently arrved mmgrants, leadng them to accept avalable employment wthout beng able to exercse choce (see, for example, Fredberg, 2000). Gven the mportance of human captal n the explanaton of mmgrant wages and wage progress, the remander of the paper explores ths central aspect n greater depth. A key queston that could affect the nterpretaton of the results assocated wth the varables related to human captal n equaton (1) s the presence of a close relatonshp between mmgrants years of resdence n Span, ther years of schoolng and ther years of potental experence. More specfcally, as ndcated by Borjas (1999), Fredberg (2000) and more recently Skuterud and Su (2008), the equaton s a restrcted specfcaton of a broader model that break downs returns to schoolng and experence accordng to whether they have been acqured n the home or host countres. The coeffcent δ n model (1) captures the effect of human captal nvestment n the host country (n ths case, Span), whereas the coeffcents assocated wth years of schoolng 14

and potental experence are affected by the relatve composton of human captal n home and host countres. For ths reason, t s useful to expand equaton (1) as follows: w = α + β sch f 2 h 1 h + β potexp + β sch f f 1 f 3 f + β + β potexp h 2 h + β h 3 f 2 ( potexp ) + γ X + h ( potexp ) ε 2 + (2) where the superscrpt h refers to human captal of any knd acqured n the host country and the superscrpt f refers to foregn human captal. The second column of Table 1 represents the results from estmatng equaton (2). Based on these estmatons, the margnal returns to a year of schoolng n Span (3.3%) are hgher than the margnal returns to a year of foregn schoolng (1.8%), and the dfference between the two coeffcents s statstcally sgnfcant at 5%. Notably, mmgrants margnal returns to schoolng n Span are relatvely much closer to the 4% estmated for natve workers usng mcrodata n the EES. In any case, the lower return to foregn formal educaton ndcates that home country schoolng have lmted transferablty to the Spansh labour market. Ths concluson s consstent wth prevous work by Sanromá, Ramos and Smón (2008). The returns to a year of potental experence n Span are 1.4%, whch would support the exstence of wage progress. Potental experence n the home country has a margnal return of 0.7% n the Spansh labour market. Ths fgure s lower than returns to experence accumulated n Span, whch supports the noton that the transferablty of foregn job experence s lmted wth respect to the Spansh labour market. Nevertheless, ths result can be seen as favourable n lght of contrastng evdence obtaned n countres lke Israel (Fredberg, 2000), Canada (Schaafsma and Sweetman, 2001) and the Unted States (Kossoudj, 1989), whch ponts to zero or near zero returns to foregn experence. Recent studes such as Skuterud and Su (2008) have ponted to the value of also dstngushng between effectve and potental experence. As ndcated prevously, the avalablty of nformaton n the ENI on mmgrant work hstores after ther arrval n Span makes t possble to break down years of Spansh potental experence nto years of effectve experence (effexp h ) and dle years (dle h ). In addton, t provdes nformaton on whether an mmgrant has been employed n the home country, permttng the ntroducton of a dummy varable (job f ) to try to represent ths effect on 15

wages. In ths way, the emprcal model can be expanded to reflect the addtonal breakdown: w = α + β sch + β sch f 2 h 1 h + β potexp f f 1 + β f 3 f h h h h 2 h + β2 effexp + β3 ( effexp ) + β4 f 2 f f ( potexp ) + β4 job + γ X + ε dle + h (3) The thrd column of Table 1 shows the results from estmatng ths equaton. The frst new contrbuton of ths estmaton s that t breaks down the effect of years actually worked n Span from dle years. The frst result to hghlght n ths sense s that returns to effectve experence n Span would appear to be greater than returns to potental experence: 2.4% and 1.4%, respectvely. (The dfference between the two coeffcents, however, s not statstcally sgnfcant at the usual levels.) The result would seem to suggest that the wage mprovement experenced by mmgrants durng ther perod of stay n Span s largely assocated wth workng and that ths s bascally the factor whch enables them to accumulate knowledge and develop sklls that are useful and adapted to the Spansh labour market. Perods of unemployment or nactvty (dle years) n Span do not appear to have any statstcally sgnfcant mpact on mmgrant wages. That would seem to ndcate that sklls acquston and the development of socal relatonshps take place most promnently n the workng envronment. Although the sgn of ths varable s negatve and not statstcally sgnfcant at the conventonal levels, t does not seem to reflect a wage penalty, whch contrasts wth what mght be expected from the evdence obtaned for other countres (Bratsberg, Barth and Raaum, 2006). One possble explanaton for ths fndng s that the obsolescence effect may be mnmal because the unemployment spells are generally short as a result of the hgh labour turnover n the Spansh labour market, the hgher job search ntensty of mmgrants n relaton to natve workers and the perod of ntense hrng occurrng durng the tmeframe of our study. The second dfference between ths model and the prevous one s the presence of a dummy varable to reflect whether mmgrants have held employment n ther home countres. The varable s statstcally sgnfcant at conventonal levels, clearly showng that mmgrants wth foregn work experence obtan an addtonal wage ncrement of 3.2%. The exstence of a wage premum ndcates hgher productvty as a result of effectve experence ganed n the home country, but ts lmted magntude draws 16

attenton agan to the lmted transferablty of most sklls acqured n settngs other than the Spansh labour market. Wth respect to educaton, the thrd model offers a result smlar to the prevous one. Returns to schoolng completed n Span (3.7%) are hgher than returns to foregn schoolng (1.7%), and the dfference s statstcally sgnfcant at conventonal levels. In model (3), a restrcton has been mposed so that the coeffcent for years of schoolng n the home country s equal for all mmgrants. By contrast, the evdence contrbuted by Bratsberg and Ragan (2002) suggests that the coeffcent for years of schoolng n the home country could be dfferent for mmgrants who have also studed n Span and for mmgrants who have only studed n ther home country and then come drectly to Span to work. In order to overcome ths restrcton and test whether schoolng n Span mprove returns to foregn educaton, model (3) has been estmated separately for the two groups: mmgrants that have fnshed ther schoolng n Span and those who have not. The results n columns 4 and 5 of Table 1 clearly show that there s effectvely a notable dfference (statstcally sgnfcant) for the two groups wth respect to the returns to schoolng completed n the home country. Returns to schoolng are much lower for mmgrants who have only studed n ther home country (1.6%) than for mmgrants who have also studed n Span (4.5%). Ths result suggests that returns to foregn schoolng for mmgrants who contnue to study n Span mght not be substantally dfferent from the fgure estmated for natves. In addton, the group of mmgrants who have completed ther schoolng n Span also present a very hgh returns. Ths evdence ponts n the same drecton as the fndngs of Bratsberg and Ragan (2002) and t seems to confrm ther concluson that pursung schoolng n the host country serves to revaldate formal educaton obtaned n the home country, makng t valuable for the labour market of the host country. One aspect of partcular nterest n the lterature s the exstence of wage dfferences and other work-related results among mmgrants as a functon of ther geographc areas of orgn. In order to evaluate whether these dfferences are also related to dstnct returns to the varous components of human captal, equaton (3) has been estmated separately accordng to mmgrant s areas of orgn. Developed and less developed economes have been dstngushed. Wthn the category of less developed economes, a further breakdown has dstngushed among mmgrants from Latn 17

Amerca, Eastern Europe and the rest of the world, whch are the only three categores wth a suffcent sample sze. The results appear n Table 2. The results for mmgrants from developed countres show hgh returns to schoolng, both for schoolng n Span (4.9%) and foregn schoolng (roughly 6%). These fgures, partcularly the latter one, are slghtly hgher than the fgure obtaned for natves from estmatons based on the EES-2006 data. The evdence reveals a very hgh transferablty of educaton from developed countres n Western Europe and North Amerca and sgnfcant returns to schoolng completed n Span. For ths group of mmgrants, however, a statstcally sgnfcant postve effect s not observed from Spansh effectve experence. Nor s there a clear wage penalty as a result of perods of unemployment or nactvty (dle years). The result s consstent wth the fact that returns to schoolng n the home country are hgher for ths group than for natves. Indeed, Smón, Sanromá and Ramos (2008) have found that the wage gap between natves and mmgrants from developed economes s favourable to the mmgrant group. Therefore, t seems reasonable to expect that ther economc progress would not occur or would be less mportant than for other groups. Ths result s not new n the lterature, but s rather known as ds-assmlaton or negatve assmlaton. For example, Chswck and Mller (2008) have found that the stuaton of mmgrants from Englsh-speakng countres who mmgrate to the Unted States actually declnes over tme wth respect to ther stuaton on arrval. Bell (1997), Dustman et al. (2003) and Clark and Lndley (2009) have obtaned smlar evdence for whte mmgrants to the Unted Kngdom. Lastly, experence accumulated n the home country, by contrast, presents sgnfcant postve returns, clearly supportng the complete transferablty of foregn experence to the Spansh labour market, although no dfferences are detected as a result of havng held job n the home country. The results for mmgrants from less developed countres show a postve wage effect from years of schoolng n both home and host countres, as well as from effectve experence n Span and potental experence n the home country. However, returns to each component of human captal are dstnct n the Spansh labour market. Specfcally, returns to schoolng n Span (3.8%) are hghly equvalent to returns for natves based on the EES, but they are much hgher than returns to foregn schoolng (1.5%). The low margnal returns to schoolng n less developed countres reflect ther lower transferablty. The same outcome appears when comparng foregn experence wth (effectve) Spansh experence: a year of work n Span results n greater wage returns 18

than a year of foregn experence. Nevertheless, havng held employment n ther home country s related to mmgrants earnng wages whch are 4.4% hgher wthn the Spansh labour market. The results for the three geographc groupngs of less developed countres clearly show that the margnal returns to schoolng completed n Span are hgher than returns to foregn educaton, revealng ther lmted transferablty. In comparatve terms, the evdence demonstrates that returns to schoolng n Span are greater for mmgrants from Latn Amerca (4.4%) and Eastern Europe (3.6%) than for mmgrants from rest of the world (2.4%). Exactly the same result arses wth respect to returns to foregn educaton. For mmgrants from the rest of the world, these returns are nonexstent. Some of the possble explanatons for ths result could be that the vast majorty of Latn Amercan mmgrants speak Spansh and that the cultural dstance s smaller for mmgrants from Easter Europe than for mmgrants from the rest of the world. In addton, dfferences n returns to schoolng n Span could also be related to the exstence of dscrmnaton n the Spansh labour market. Equally, the gap could reflect the lmtng or determnant effect that low-qualty educaton n the home country could have on the ablty of mmgrants from the rest of world to beneft effectvely from any schoolng subsequently completed n Span. (Recall that the coeffcent assocated wth ths varable s not sgnfcant for mmgrants from the rest of the world.) 19

Table 2 Logarthm of monthly wages Developed countres Less developed countres Latn Amerca Eastern Europe Rest of the world Legal status 0.182** 0.130*** 0.136*** 0.133*** 0.245*** [0.0813] [0.0156] [0.0214] [0.0239] [0.0457] Schoolng years n Span 0.0486* 0.0381*** 0.0435*** 0.0357** 0.0243* [0.0278] [0.00804] [0.00989] [0.0179] [0.0146] Schoolng years n home country 0.0596*** 0.0148*** 0.0175*** 0.0109*** 0.00417 [0.00766] [0.00167] [0.00213] [0.00329] [0.00329] Effectve experence n Span 0.0311 0.0242*** 0.0354*** 0.0122-0.00979 [0.0303] [0.00778] [0.00997] [0.0150] [0.0178] Effectve experence n Span 2-0.00176-0.00111-0.0019 0.000148 0.00222 [0.00357] [0.000883] [0.00120] [0.00182] [0.00192] Idle years n Span -0.111*** -0.00373 0.00539-0.016-0.000421 [0.0244] [0.00486] [0.00629] [0.00976] [0.00938] Potental experence n home country 0.0178** 0.00391** 0.00932*** 0.00217-0.00667 [0.00865] [0.00175] [0.00236] [0.00343] [0.00493] Potental experence n home country 2-0.00019-0.000130*** -0.000258*** -0.000126 0.000116 [0.000252] [4.93e-05] [6.59e-05] [9.75e-05] [0.000149] Labour experence n home country -0.0274 0.0426*** 0.0436** 0.0334 0.0396 [0.0622] [0.0155] [0.0207] [0.0288] [0.0309] Number of observatons 381 4.504 2.586 1.286 632 Adjustd R 2 0.506 0.446 0.437 0.545 0.384 Notes: OLS estmates wth controls related to gender, cvl status, geographcal area of brth, monthly worked hours, havng more than one job, the regon of resdence and Heckman s lamda. ***, ** and * ndcate that the estmated coeffcent s statstcally dfferent from zero at 1%, 5%, and 10% sgnfcance levels, respectvely. 20

Wth respect to foregn work experence, a postve wage effect s found only for Latn Amercan mmgrants and not for mmgrants from Eastern Europe or the rest of the world. In addton, the fact of havng actually worked n the home country s only related to hgher wages (4.5%) n the case of Latn Amercan mmgrants. As already noted, t s common n the lterature to fnd zero returns to foregn experence both effectve and potental n the case of mmgrants from less developed countres. Zero transferablty s also a common result. The economc, technologcal, cultural and lngustc dstance between home and host countres translates nto knowledge and sklls that do not match the requrements of a developed country s labour market, the Spansh economy n ths case. The explanaton could also le at least partly n the dfferent profcency n Spansh of mmgrants from Latn Amerca, who experence lmted but postve returns, when compared wth mmgrants from other less developed countres. Ths result has already been hghlghted n prevous studes (e.g., Sanromá, Ramos and Smón, 2008). Effectve labour experence n Span only presents postve and sgnfcant margnal returns for mmgrants from Latn Amerca 12. By contrast, effectve experence n Span s not statstcally dfferent from zero for the remanng mmgrants. Sanromá, Ramos and Smón (2008) have already obtaned some evdence of the absence of assmlaton n the levels of over-educaton found among sub-saharan and Asan mmgrants, so t s not surprsng to fnd no wage progress over tme as mmgrants from the rest of the world gan experence after arrval n Span. By contrast, the absence of wage progress and, therefore, of assmlaton n the Spansh labour market s nconclusve n the lterature on mmgrants from Eastern Europe. Whle Fernández and Ortega (2008) have found no assmlaton n terms of over-educaton, Sanromá, Ramos and Smón (2008) have found evdence supportng assmlaton. The matter s doubtless of suffcent nterest partcularly n lght of the large-scale nflux of mmgrants from Eastern Europe n recent years to conduct a more n-depth analyss n future. Among other ssues, future research should examne the occupatonal progress of mmgrants and analyse the factors affectng t. A fnal result to hghlght s that beng able to work legally has a postve wage mpact for mmgrants from all geographc areas. The wage dfferental for mmgrants 12 The hgh value of the coeffcent (0.035), whch s clearly greater than the 0.012 estmated for natves n the EES, tends to confrm the possblty that wage assmlaton does exst for Latn Amercan mmgrants. 21