Ethnic minorities in the UK: burden or benefit?

Transcription

1 Ethnc mnortes n the UK: burden or beneft? Aslan ZORLU Department of Economcs UNIVERSITY OF AMSTERDAM Roetersstraat 11, 1018 WB Amsterdam - the Netherlands Tel: +31 (0) Fax: +31 (0) zorlu@fee.uva.nl REPLACED 13/07/2001 Key Words: mmgrants, ethnc mnortes, labour market JEL CATEGORY: J15, J21, J31, J61 Abstract Ths paper nvestgates the degree of the labour market competton between Whte and ethnc mnorty labour usng four waves of the Brtsh Quarterly Labour Force Survey (1997). The extent of the labour market competton s examned n terms of wage elastctes and elastctes of complementarty usng translog producton technology. The estmatons suggest that ethnc mnorty labour s complementary to most of Whte workers. The substtuton relatonshp s especally evdent between ethnc mnorty labour on the one hand and low sklled, medum sklled non-manual and low sklled manual Whte labour on the other hand. Ethnc mnortes do not seem to be substtute for each other. Helpful comments from Rchard Berthoud, Stephen Drnkwater, Joop Hartog, Stephen Wheatley Prce and partcpants n the Monday afternoon semnar at Insttute for Socal and Economc Research, the Unversty of Essex, s gratefully acknowledged Ths paper s based on work carred out durng a vst to the European Centre for Analyss n the Socal Scences (ECASS) at the Unversty of Essex, supported by the European Commsson Tranng and Moblty of Researchers-Access to Large Scale Facltes Programme. Materal from the Labour Force Survey s Crown Copyrght; has been made avalable by the Offce for Natonal Statstcs through the Data Archve and has been used by permsson. Nether the ONS nor The Data Archve bears any responsblty for the analyss or nterpretaton of the data reported here

2 A. Zorlu/Ethnc mnortes n the UK (2000), 2 1 Introducton Whle Western Europe has been one of the magnet felds for nternatonal mgraton flows n the post-war perod, many European countres have swtched from beng an emgraton to an mmgraton country. Untl 1973, mmgraton flows were domnated by guest workers who were recruted to meet wth unsklled labour shortages generated by the post-war glory of many Western European economes. Snce the Ol Crss n 1973, mmgraton flows from developng countres towards West European countres have contnued n the form of famly reunfcaton 1. In the eghtes and nnetes, mmgraton of asylum seekers from poltcal unstable areas has ncreasngly ganed mportance. Unlke other West European countres, the Unted Kngdom has drectly recruted only a small mnorty of mmgrants. Immgraton polces have been desgned to restrct mmgraton flows from former Brtsh colones and domnons. However, these restrctve polces have not been always successful n reachng ther objectves. Immgraton from Carbbean countres and South Asa has ncreased consderably. Later mmgraton of non-whte people has gone on n the form of famly reunfcaton. Addtonally, a large number of professonal workers, who typcally dd not settle permanently, have been allowed entry, and the mmgraton of low sklled workers from Southern Europe ncreased after free moblty of persons wthn the European Communty was allowed (Wheatley Prce 2000). Integraton of mmgrants n host country labour markets does not proceed smoothly. As a reacton, the mmgraton polcy of the EU-member countres becomes more and more restrctve. In partcular, an attempt s made to lmt mmgrant flows from developng countres by new legslaton or ndrectly by tghter applcaton of exstng polcy nstruments. These restrctve mmgraton polces are based on an assumpton that newly enterng mmgrants would have an adverse effect on wages and employment of natves and would also cause a further deteroraton of the labour market poston of settled mmgrants n these countres. The dsadvantaged labour market poston of ethnc mnortes from developng countres s a man cause of rsng restrctve mmgraton polces n many European countres, whch suffer hgh structural unemployment. However, t s unclear what the reacton of host country labour markets s to the enterng mmgrant labour force n terms of the capablty of the labour markets to absorb new mmgrants and the adjustment of these mmgrants to host country labour markets. In the last two decades, a large body of studes has been conducted snce the semnal paper of Chswck (1978). Most of these studes concern tradtonal mmgraton countres lke the US, Canada, Australa (see for a survey Fredberg and Hunt 1994 and Borjas 1994; Schultz 1998). In Europe, the number of studes s stll lmted but ncreasng n the last few years (De New and Zmmermann 1994; Gang and Rvera-Batz 1994; Venturn 1999; Zorlu and Hartog 2000). Ths may be due to the lack of approprate data sets to make n-depth analyss. More and more questons about the ethnc orgn are ncluded n data sets. The emergence of data sets ncludng a reasonable number of observatons about ethnc mnorty workers stmulates research on ths queston. In the UK, the labour market poston of ethnc mnortes s nvestgated n several studes. The research on ethnc mnortes has focussed on the documentaton of the dsadvantaged poston of ethnc mnortes n the labour market. The man concern has been on dscrmnaton, dfferences n earnngs and unemployment rates of natve and ethnc mnorty labour forces. It has been shown that members of ethnc mnorty groups are more frequently unemployed and they earn less than natves wth smlar observed soco-economc and ndvdual characterstcs (Blackaby et al. 1998, 1999a, 1999b; Bell 1997; Berthoud 1998; Shelds and Wheatley Prce 1998). The gap between unemployment and the wages of natves 1 Increasng mgraton among developed countres as a result of economc globalsaton s not seen as a problem. Therefore, t has not been a research topc for labour economsts 2

3 A. Zorlu/Ethnc mnortes n the UK (2000), 3 and ethnc mnortes s explaned by the lack of relevant schoolng and experence for the Brtsh labour market and unexplanable factors, whch are partly referred to dscrmnaton practces (Moodod et al 1997). The dsadvantaged poston of ethnc mnortes can be attrbuted to factors on both the demand and supply sdes. Dscrmnatng behavour of employers and a decrease n demand for a certan type of labour may dsturb labour market partcpaton of ethnc mnortes on the demand sde whle skll dstrbuton, a lack of relevant educaton/tranng for the host country may play a crucal role on the supply sde. The lack of educaton and experence can be overcome by relevant polces. For nstance, provdng educaton to those who need t could be a useful tool. However, the unexplaned part of the dsadvantage s often hard to dentfy and remove. Ths last aspect may be correlated wth the absorpton capacty of the labour market and wth dscrmnaton. The sze and ntensty of dscrmnaton may, n turn, be, at least partly, related to busness cycles. It mght be expected that n perods of hgh unemployment, natve workers would have bgger chance to be employed than workers from ethnc mnorty groups. When a economy stagnates, members of ethnc mnorty groups wll be the frst to lose ther job. In the UK, the labour market poston of ethnc mnortes s studed only from the supply sde. No research has been conducted to study the mpact of mmgraton on the labour market and reacton of the Brtsh labour market to ethnc mnortes. The poor labour market poston of ethnc mnortes and the ncreasng fear of the natve populaton of beng flooded by mmgrants generate a strong bass to nvestgate the nteracton between the natve and the ethnc mnorty labour force n the UK. Ths study nvestgates the nteracton between Whte and ethnc mnorty labour n a broader context. The objectve of ths paper s to study the extent of labour market competton between ethnc mnortes and Whtes usng data from the four waves of the Quarterly Labour Force Survey 1997, whch are aggregated nto a sngle data set. The Whte labour force s dsaggregated nto sx categores: low, medum and hgh sklled manual workers, and low, medum and hgh sklled non-manual workers. Ethnc mnortes are dsaggregated frst nto three skll levels and then nto four man groups for estmaton of earnngs: Black, Indan, Pakstan and Mxed. For the estmaton of producton technology, Black and Pakstan are aggregated nto a sngle group, as well as Indan and Mxed n order to keep the number of observatons reasonable. In secton 2, a short hstory of mgraton n the UK s dscussed. In secton 3, an overvew of the data s gven and descrptve statstcs of selectve varables are dscussed. Secton 4 presents a theoretcal concept, whch deals wth the drect effect of ethnc mnorty labour on wages of Whte workers on a county bass. The hypotheses of ths model are tested by the estmaton of earnngs functons n whch percentages of ethnc mnorty labour n countes are ncluded n secton 5. Then, nteractons between Whte and ethnc mnorty labour are examned n a general equlbrum framework n secton 6. The absorpton capacty of the Brtsh labour market s approxmated by studyng complementarty and substtuton elastctes among varous types of labour. Conclusons are presented n secton 7. 2 UK's mmgraton experence The Unted Kngdom has tradtonally been an mportant source of mgraton flows. The vast majorty of Brtsh emgrants have gone to Englsh-speakng countres n the New World. In recent years has the Unted Kngdom turned out to be a country of net mmgraton 2. Immgraton flows, orgnatng n non-englsh speakng countres, have played a crucal role n that process. In partcular, mmgraton from Commonwealth countres and Pakstan, and most recently refugees from Eastern Europe and other poltcally unstable countres has ncreased n the nnetes. Table 1 shows net mmgraton and emgraton flows by orgn and 2 Hatton and Wheatley Prce (1998) dscuss mgraton from and to the Unted Kngdom extensvely. 3

4 A. Zorlu/Ethnc mnortes n the UK (2000), 4 host country. The New World seems to always be a destnaton for Brtsh emgrants whle South East Asan, Afrcan and Carbbean Commonwealth countres have tradtonally been a source of mmgraton flows. The EU countres, New Zealand and South Afrca have swtched from beng destnaton to beng source countres. Non-EU European countres have retaned ther earler poston as mmgraton countres between 1985 and Ths may be due to enlargng EU borders snce Table 1. UK Net Mgraton by country Commonwealth and Pakstan Australa Canada New Zealand South Afrca Other Afrcan Commonwealth Bangladesh, Inda, Sr Lanka Pakstan Carbbean Commonwealth +12.8* Other Commonwealth Foregn European Unon ** Rest of Europe ** Unted States Rest of Amerca Mddle East # Other Foregn Source: Hatton and Wheatley Prce (1998) Notes: Average annual net mgraton n thousands,.e. sgn + represents net mmgraton and sgn - denotes net emgraton * West Indes only ** EU countres before enlargement n 1981 # only In 1991 the total stock of non-brtsh people numbered nearly 4 mllon whch made up 7.4% of the UK populaton. Nearly 3 mllon people from ths populaton belonged to the ethnc mnortes (5.5 % of the total populaton) of whom the majorty was born outsde the UK (Shelds and Wheatley Prce 1999). Although Irsh mmgrants have formed the largest non-brtsh group n the UK (Wheatley Prce 2000 and Halpn 1997), both mmgraton polcy and research on mmgraton and mmgrants n the UK have been concerned n partcular wth non-whte mmgrants and ther descendants. Ths s partly related to the fact that Whte mmgrants have a relatvely better labour market poston than the non-whte mmgrants. However, the dsadvantaged labour market poston does not explan the focus on non-whte mmgrants. Chnese and Indan people, for example, do very well n the labour market but they nonetheless get attenton from polcy makers and researchers. Ths may be explaned by UK mmgraton polces. Hatton and Wheatley Prce (1998) state that UK mmgraton polces have been drven by poltcal concerns. Behnd the development of mmgraton polces, Brtsh governments have been concerned about domestc racal relatons and the Brtsh perceptons concernng the ablty and wllngness of mmgrants from dfferent ethnc orgn to assmlate n Brtsh socety. Economc consderatons have played lttle role n determnng restrctve mmgraton polces. 4

5 A. Zorlu/Ethnc mnortes n the UK (2000), 5 Prevous research on mmgrants n the UK has focussed manly on the explanaton of the dsadvantaged poston of ethnc mnortes wth respect to earnngs and (un-)employment prospect 3. Blackaby et al. (1997) study male and female unemployment dfferences across ethnc mnortes. They argue that the dfferences are not smply a result of dfferences n characterstcs or dscrmnaton. They emphasses characterstc dfferences among foregn born and UK born ethnc mnortes and conclude that UK born non-whte ethnc mnortes are not dong worst. Blackaby et al. (1999) suggest that the dfferences n unemployment across ethnc mnortes are not smply a result of dscrmnaton. They take Indan and Pakstan/Bangladesh groups as reference group assumng that a common level of dscrmnaton exsts aganst these groups because Indans are predomnantly non-muslm and the Pakstan/Bangladesh groups are predomnantly Muslm. They conclude that the relatvely hgh unemployment among the second group s due to less favourable characterstcs of ths group. Wheatley Prce (2000) fnds a large ntal employment dsadvantage for recent male mmgrants. However, the employment rate for Whte mmgrants rapdly ncreases over the frst fve years n the UK. On the other hand, the employment rate of non-whte mmgrants ncreases over 20 years but t never does attan the employment level of natve born men. The fndngs of Hatton and Wheatley Prce (1998) confrm ths result. Bell (1997) fnds a relatve dsadvantaged earnngs profle for blacks. However, ths dsadvantage s reduced as duraton of stay n UK ncreases. In other words, locaton specfc human captal ncreases wth the duraton of resdence n the UK. Lower earnngs refer to the lack of relevant human captal (educaton, experence, etc). Human captal obtaned abroad s apprecated less than the human captal n the UK although that strongly vares per country. Shelds and Wheatley Prce (1998) report a lower return to schoolng obtaned n the UK for most mmgrant groups. Only the beneft of UK born non-whtes and Whte mmgrants from ths educaton s comparable wth natve born Whtes f, at least, Englsh s ther frst language. Addtonally, experence obtaned n the UK s more benefcal than experence abroad. However, Irsh and non-whte mmgrants do not receve any reward from experence obtaned n ther country of orgn whle foregn experence of Whte mmgrants s apprecated more than ther potental UK experence. Language s a typcal locaton specfc human captal that has a large mpact on labour market performance of mmgrants. Indeed, Shelds and Wheatley Prce (1999) fnd, not surprsngly, that Englsh language fluency consderably mproves the labour market performance of mmgrants and ncreases mean hourly wages by about 20 per cent. As reported n the next secton, ethnc mnortes are concentrated n certan urban areas. Lvng n ethnc enclaves adversely affects human captal accumulaton lke obtanng Englsh profcency (Shelds and Wheatley Prce 1999) and consequently reduces labour market performance. Clark and Drnkwater (1999) fnd that n areas wth a hgh degree of ethnc concentraton, self-employment among ethnc mnortes s low and unemployment s hgh. They argue that the poor economc performance of ethnc mnortes s related to concentraton of ethnc mnortes n certan parts of urban areas. It becomes clear from prevous studes that some part of labour market dsadvantage of ethnc mnortes s not explanable from observable ndvdual characterstcs. Addtonally, perceptons about the transferablty of foregn human captal are more lkely subjectve n practce. Enterng relevant nsttutonal bodes, lke schools, employment agences, whch allow and stmulate the accumulaton of locaton specfc human captal, s a dynamc process. It may be dscouraged or ressted. Ths mples a certan level of resstance n the labour market aganst ethnc mnortes. That resstance s substantated from the vewpont of lack of nformaton and possble competton between natve and ethnc mnorty labour. However, there s no nformaton about the degree of the labour market competton. Ths 3 Hatton and Wheatley Prce (1998) gve a detaled survey of mmgraton polcy and prevous studes. 5

6 A. Zorlu/Ethnc mnortes n the UK (2000), 6 study examnes the extent of labour market competton among Whte and ethnc mnorty workers for the frst tme. In the next secton, a drect effect of ethnc mnortes on the wages of natve labour force s examned. In secton 6, the effect s researched n a general equlbrum model. 3 The Data The four waves of the Quarterly Labour Force Survey (LFS), Sprng 1997, Summer 1997, Autumn 1997, and Wnter 1998, are pooled for ths study because separate waves do not provde suffcent observatons for the analyss conducted n ths paper. Takng nto account the longdtunal aspect of the LFS, each respondent s allowed to enter once n the data. Among all respondents, the populaton aged s used for the descrptve analyss. Only employees are selected for further estmatons. People, who are self-employed, workng on government tranng programme and dong unpad famly work, are excluded from the survey. Employees are dvded nto fve man categores on the bass of ther ethnc background: frstly, Black, Indan, Pakstan/Bangladesh and Mxed/Other orgns are selected usng the defnton used n the Census of Populaton and then the rest of populaton s defned as Whte. Table 2 presents the descrptve statstcs of some selected varables concernng four ethnc mnorty groups and the Whte populaton. It can easly be seen that ethnc mnorty populaton, especally Pakstan and Bangladesh, s younger than the Whte populaton. More than 50 percent of the Black and 43 percent of the Mxed/Other are sngle whle less than 30 percent of Indan and Pakstan/Bangladesh s sngle and more than 60 percent of these groups are marred. The partcpaton rate among ethnc mnorty groups both for women and men s hgher than that among Whtes. Especally Pakstan/Bangladesh, Indan and Mxed/Other partcpate n the labour market less frequently. The unemployment rate among Black and Pakstan/Bangladesh s, on average, the hghest. However, the unemployment gap between Whte and the last three ethnc groups s consderably larger for men than for women, except Black women. A relatvely large share of Indan people s selfemployed. The average weekly earnngs of the employed labour force vary wdely among the fve samples. Indan employees earn the hghest net wages n a week, about 231. A large standard devaton for the mean net weekly earnngs ndcates a large varaton n the earnngs of Indan workers. In hgh level of earnngs, Indan workers are followed by Mxed/Other ( 212), Whte ( 206) and Black ( 193) respectvely. Pakstan/Bangladesh have the lowest average weekly earnngs ( 154). The same order of successon can be observed n the number of mean workng-hours. However, the pcture s dfferent when experence and tenure are concerned. On average, Whtes are much more experenced and they work much longer for a same employer than ethnc mnortes. Black and Indan workers have smlar dsperson n experence and tenure, as well as Pakstan/ Bangladesh and Mxed/Other. Skll dstrbuton across the groups dffers strongly as well. Almost threequarters of Pakstan/Bangladesh workers are low sklled. Indan workers posses, on average, the hghest skll levels, followed by Mxed/Other, Black, Whte and Pakstan/Bangladesh. Temporary jobs seem to be more popular for ethnc mnortes than for Whtes. Indan and Pakstan/Bangladesh workers are more often employed n the publc sector than ether other ethnc mnortes or Whtes. A relatvely large proporton of Indan and Mxed/Other s nonmanual workers and professonals. The dstrbuton of ethnc mnortes over regons dffers fundamentally from Whtes. Black people lve n the West Mdlands, Inner and Outer London. Indan people are concentrated manly n the East Mdlands, West Mdland and Outer London. Pakstan people are concentrated n Greater Manchester, West York, the West Mdlands, Inner and Outer London. People from category Mxed/Other are concentrated around London and the South East. In addton, the dfferences n employment by ndustres among the samples are notable: ethnc mnortes are more concentrated n manufacturng, the wholesale ndustres, hotels, transport, real estate and health. Although 6

7 A. Zorlu/Ethnc mnortes n the UK (2000), 7 these summary statstcs are only suggestve, they provde a useful background for the emprcal results presented n the followng part of ths paper. 7

8 A. Zorlu/Ethnc mnortes n the UK (2000), 8 Table 2. Descrptve Statstcs column percentages Age Whte Black Indan Pakst/Bangla Mxed/other Total Number of observatons Martal status Sngle, Marred lvng togeth Marred separated Dvorced Wdowed Total Number of observatons Economc Statute Female In employment ILO unemployed Inactve Total Number of observatons Male In employment ILO unemployed Inactve Total Number of observatons Economc actvty Employee Self-employed Goverm. emp/tran prog Unpad famly worker Total Number of observatons Net weekly earnngs 4 Mean net weekly pay Standard devatons Number of observatons Skll dstrbuton Low sklled Medum sklled Hgh sklled Total Number of observatons Type of job Permanent Not permanent Total Number of observatons Sector Prvate Publc Total Number of observatons Occupaton Manual Non-manual Armed force Total Number of observatons contnued 4 Net weekly earnngs are calculated only for employees. 8

9 A. Zorlu/Ethnc mnortes n the UK (2000), 9 Table 2. Descrptve statstcs (contnued) Whte Black Indan Paks/Bangla Mxed/other Means Experence (Std. Dev.) (13.66) (12.35) (12.64 ) (13.14) (11.64) Number of observatons Tenure (Std. Dev.) 8.08 (8.05) 6.43 (6.94) 6.94 (6.66) 4.18 (4.87) 5.88 (6.43) Number of observatons Work. Hours (Std. Dev.) (15.87) (15.03) (16.58) (16.04) (15.87) Number of observatons column percentages Socal class Professonal occs Intermedate occs Sklled occs (non-manu) Sklled occs (manual) Partly sklled occs Unsklled occs Armed forces Total Number of observatons Government offce regons Tyne & Wear Rest of Northern regon Greater Manchester Merseysde Rest of Northwest South Yorkshre West Yorkshre Rest of Yorks/Humbers East Mdlands West Mdlands Rest of West Mdlands Eastern Inner London Outer London South East South West Wales Strathclyde Rest of Scotland Northern Ireland Total Number of observatons Industry sectons Agrculture Fshng Mnng Manufacture Electrc Constructon Wholesale Hotels Transport Fnance Real estate Publc Educaton Health Other Prvate Extra Workplace outsde UK Total Number of observatons

10 A. Zorlu/Ethnc mnortes n the UK (2000), 10 4 THEORY: PARTIAL EFFECT The summary statstcs show that workers from ethnc mnorty groups dffer from ther counterpart Whtes wth respect to ther human captal accumulaton and soco-economc poston. On the other hand, ethnc mnorty groups dffer strongly from each other. Ths s also wdely documented n prevous studes (Berthoud 1998; Berthoud 1999; Blackaby 1997 and 1999; Moodod et al. 1997; Jones 1996). Ths rases the queston of how dfferent types of labour n the Brtsh labour market nteract. In the lterature on mmgraton, the effect of mmgraton on labour market outcomes s examned on the bass of substtutablty between natves and mmgrants n producton. Emprcal analyses are manly based on neo-classcal multfactor models. In these models, changes n the employment of a certan producton factor affect the demand curves of other factors through elastctes of substtuton and complementarty. The predctons of these theores are straghtforward: f mmgrants are close substtutes for natves, n general terms, the theory predcts that mmgraton wll have an adverse effect on wages and employment of natves (Borjas et. al. 1997). Frst, we analyse solated partal nteractons between dfferent types of labour by drect modellng n ths secton and then smultaneous nteractons wll be taken nto account n secton 5. We assume that aggregate producton n a gven labour market s a functon of N types of natve labour force and mmgrant labour, L M. The producton functon s lnearly homogenous and s charactersed by constant returns to scale. 1. Y = f L,..., L N, L ) ( 1 M The total labour force consst of natve and mmgrant workers, L=L 1 + +L N +L M, whch mples L 2. 1 = M L The frst order condtons requres that wage of each labour nput s equal to ts margnal product n terms of output. 3. w f = ( L L, L ) 1,..., L N M Ths demand system represents the wage determnaton process n a gven labour market, gven perfectly nelastc relatve labour supples. If we stress only an ncrease n the mmgrant labour force, the effect of a proportonal ncrease n mmgrant labour on wages of dfferent types of labour may be gven as follows. 4. dw w = ( S C m dl ) L m m where the prce of output s normalsed to 1. S m = wmlm Y s the factor share of mmgrants n producton and C m s the Hcks partal elastcty of complementarty between factors m and, defned as C m = Yf m f m f (Sato and Kozum 1973). f m s the frst dervatve of producton functon wth respect to factor m. Identcally f s the frst dervatve of the producton functon wth respect to factor. f = f L and f m s the second 10

11 A. Zorlu/Ethnc mnortes n the UK (2000), 11 dervatve, f m = 2 f Lm L. The elastcty of complementarty regsters the effect of a relatve change n the quantty of one factor on the prce of another factor. If an ncrease n the quantty of one factor has a postve (negatve) effect on the prce of another factor, these two factors are complements (substtutes). In terms of equaton 4, mmgrants are complementary to labour type 2 f C m2 >0. In opposte, mmgrants and factor 2 are substtutable f C m2 <0. Equaton 4 reveals that a proportonal ncrease n the mmgrant labour force wll affect the wages of dfferent types of labour through elastctes of substtuton or comlementarty among types of labour 5. Ths effect s related not only to relatve share of mmgrants, but also the relatve shares of other nputs n producton, even n the absence of labour market dscrmnaton. Ths analyss mples that f a number of new mmgrant workers wth low skll enters a certan county, the wages of workers who are closer substtute for mmgrants, n ths case low sklled workers, wll decrease, and consequently the wages of workers who are complments wth mmgrants wll ncrease. Straghtforwardly, f skll level of mmgraton flow s hgh, wages of hgh sklled workers wll decrease n a certan county. These theoretcal predctons may be tested by estmaton of earnngs functons. We prefer to choose local labour market,.e. county, rather than ndustres, as the producton unt where dfferent types of labour are employed n the producton process. Ths choce reles on an assumpton that labour force s less moble across countes than across ndustres. Ethnc mnortes are hghly concentrated n certan regons, as shown n the precedng secton. That mples a lower moblty. Low sklled labour s generally supposed to be less moble across regons. Dfferently, t s a well-known phenomenon that hgh sklled labour s much more moble across regons. However, the concentraton of ndustres n certan regons weakens for moblty. Ths, together wth some expermentaton, brngs us to the concluson that labour market competton between ethnc mnortes and Whte labour f at all present, may occur wthn countes rather than wthn ndustres. In the next secton, we assume that no smultaneous nteractons occur among the dfferent categores of labour n producton process. Only drect partal nteractons are allowed. Therefore, earnngs functons of dfferent categores of labour are estmated separately. Estmatons For the analyss of earnngs, employees are selected accordng to ther economc statute based on the defnton used by the Internatonal Labour Organsaton. We assume that there s no dfference between the physcal quanttes of natve and mmgrant labour but these two knds of labour dffer n ther human captal accumulaton expressed, for nstance, as educaton and experence. In other words, ndvdual wages are determned by observable ndvdual characterstcs and other soco-economc varables for both natves and mmgrants. After many experments, the entre labour force s dsaggregated nto 10 categores of workers, whch are approxmated by ther educaton level and ethnc background. Whte employees are dvded nto two major categores: manual and non-manual snce these two knds of jobs are dfferent n the producton process. After ths both groups are dsaggregated nto three skll levels usng the hghest educaton level: low, medum and hgh sklled workers. Subsequently, the dsaggregaton of the Whte labour force nto sx categores wll create more nsght wth regard to competton n the labour market snce the theoretcal effect of mmgraton s typcally based on the skll dstrbuton of mmgraton flows. L 5 Snce 1 = s assumed, the ncrease n the relatve supply of mmgrants must be compensated by L a decrease n share of natve labour force. 11

12 A. Zorlu/Ethnc mnortes n the UK (2000), 12 Ethnc mnorty groups are sub-sampled on bass of ther country of orgn accordng to the Census of Populaton defnton. A more detaled dsaggregaton of ethnc mnortes s desrable because mmgrants are hghly heterogeneous n the UK. However, the lmted number of workers from ethnc mnorty groups n the data makes further dsaggregaton dffcult. The sub-samples are then as follows: 1- low sklled manual workers 4- low sklled non-manual workers 2- medum sklled manual workers 5- medum sklled non-manual workers 3- hgh sklled manual workers 6- hgh sklled non-manual workers 7- Black workers (a pool of Carbbean, Afrcan and Mxed Black workers) 8- Indan workers 9- Pakstan (a pool of Pakstan and Bangladesh workers 10- Mxed (a pool of Chnese, Asan (non-mxed), other (non-mxed) and other (mxed) workers) In the rest of the paper, the sample of Pakstan/Bangladesh s shortly called Pakstan and the sample of Mxed/other s called Mxed. The effect of the share of ethnc mnorty workers on wages of Whtes s estmated n the form of conventonal earnngs functons 2 w = α + βx + γp + ε where ε ~ N(0, ) 1. xj xj jc xj σ ε w xj s the logarthmc weekly net earnngs from frst and second job of ndvdual x from the group j. X s a vector of explanatory varables. P s a vector of the percentage of the ethnc mnorty groups n 68 countes. Weekly net ncome s the sum of weekly net ncome from the frst and second job. Because the data consst of four waves of the Quarterly Labour Force Survey, seasonal dummes are ncluded n earnngs functons to capture possble seasonal effects. Earnngs functons are estmated for the three skll levels of manual and non-manual Whte workers as well as for four non-whte ethnc mnorty groups 6 separately. In all equatons, explanatory varables of a Mncer type earnngs functon are ncluded. Experence s defned as age mnus the age when the full-tme educaton was fnshed. Tenure s the years that the ndvdual started workng for the current employer. Workng hours refers to total actual hours worked n the man and second job. As the theoretcal concept n Secton 4 suggests, shares of dfferent ethnc groups n employment n 66 countes are ncluded n the earnngs functons 7. That allows us to measure the effect of ethnc concentraton on the wages of other ethnc groups. To control for dfferences between both ndustres and regons, we nclude dummes for ndustres and regons where ethnc mnorty workers are concentrated. Earler studes on ethnc mnortes/mmgrants n the UK show that the labour market dsadvantage of ethnc mnortes occurs n the form of relatve hgh unemployment among ethnc mnortes (Blackaby et al and 1999; Wheatley Prce 2000). Ths fndng suggests that ethnc mnortes are more vulnerable to unemployment than natves. Therefore unemployment rates for 66 countes are calculated and ncluded n the earnngs functons. In order to capture the scale effect n workng places, a dummy varable for frms wth less than 20 employees s used. Because the data s an aggregaton of four quarters of LFS 1997, dummes are constructed to elmnate seasonal effects. Further, a number of dummy varables 6 The four ethnc mnorty groups are dsaggregated accordng to ther ethnc orgn by Census of Populaton. The name of varable s ETHNICA n the Labour Force Survey. ETHNICA s a summary of ETHCEN. 7 These eght ethnc mnorty groups are dentfed accordng to ther ethnc orgn, whch s a defnton of Census of Populaton. The name of ths varable s ETHCEN n the Labour Force Survey. The varable ETHNICA n whch ethnc mnortes are dvded nto four man groups s a summary of ETHCEN. Snce ths more detaled dsaggregaton of ethnc mnortes has no consequences for the number of observatons n the estmatons, we prefer ths varable rather than ETHNICA. 12

13 A. Zorlu/Ethnc mnortes n the UK (2000), 13 are used to control dfferences n gender, martal status, type of job, health problems, unon membershp, commutng and publc/prvate sector. Addtonally, for the earnngs equatons of ethnc mnortes, we construct dummy varables to capture cohort effects. So, dummes are ncluded for the fve perods when mmgrants arrved n the UK: between , , , and after The sx categores of the Whte labour force and the four categores of ethnc mnortes are estmated separately. The coeffcents of P ndcate the degree of labour market competton between Whtes from dfferent skll levels and ethnc mnortes as well as ethnc mnorty groups themselves, gven that the ethnc mnorty groups possess dfferent skll levels. The results of the estmated earnng functons wth robust standard errors are presented n Table A1 and A2. Usng the coeffcents (γ) and mean values of the percentages of eght ethnc mnorty groups (P), partal wage elastctes are calculated for Whte workers between varous types of labour 8. A negatve sgn for elastcty ndcates the substtuton relatonshp between types of labour concerned. Straghtforwardly, a postve sgn ndcates the complementarty relatonshp. Table 3 shows the wage elastctes for Whte workers by skll categores wth respect to ethnc mnorty groups. The elastctes for Carbbean Black labour share are negatve for low sklled non-manual, and low and medum sklled manual Whte workers but the elastcty s almost statstcally sgnfcant for only the low sklled nonmanual workers. Black Afrcan workers have a small postve sgnfcant effect on the earnngs of hgh sklled non-manual Whte workers. The other elastctes are not sgnfcant and have varyng sgns. The elastctes for mxed Black workers are negatve for all categores but sgnfcant only for medum sklled non-manual and hgh sklled manual Whte workers. The elastcty of Indan workers s negatve only for low and medum sklled workers and s postve for the rest of the categores but t s sgnfcant only for the hgh sklled manual. The elastctes for Pakstan workers have, n many cases, the nverse sgns of Indan workers. Table 3. Estmated wage elastctes for Whtes Wth respect to quantty of Changes n the wage of NON-MANUAL MANUAL LOW MED HIGH LOW MED HIGH Elast. t Elast. t Elast. t Elast. t Elast. t Elast. t Per. black (Carbb) Per. black (Afrc.) Per. black (Mxed) Percentage Indan Percent. Pakstan Perc. Bangladesh Percent. Chnese Percentage Mxed Elastctes are calculated on the bass of the estmated coeffcents of earnngs functons presented n Appendx. Elastcty can be smply calculated as (the estmated coeffcent)*(the mean of the explanatory varable). t-statstcs are the estmated coeffcents of earnngs functons presented n Table A1 n Appendx. The percentage of Bangladesh workers has a negatve effect on all the other categores, except for low and medum sklled manual workers. The spatal concentraton of Chnese workers affects the manual workers' wage negatvely and the non-manual workers' wage 8 The wage elastcty s the product of the estmated coeffcent and ts mean. 13

14 A. Zorlu/Ethnc mnortes n the UK (2000), 14 postvely. The coeffcents of Mxed workers are negatve only for hgh sklled and postve for the rest of the workers. In other words, n terms of sgnfcant coeffcents, Carbbean Black labour may be a substtute for low sklled non-manual Whte labour and Afrcan Black s complementary to hgh sklled non-manual workers. Mxed Black seems to be a substtute for medum sklled non-manual and hgh sklled manual workers. Indan workers may be complementary to nonmanual hgh sklled manual workers. Fnally, Mxed workers are complementary to low sklled non-manual Whte workers. In Table 4, the estmatons of earnngs functons of ethnc mnortes shows the labour market competton among ethnc mnortes. Indan workers are complementary to Pakstan workers. The rest of the underlyng coeffcents are not statstcally sgnfcant. Regardless of the sgnfcance of the coeffcents, Black workers may compete wth Indan workers and may be complementary to Pakstan and Mxed workers. Indan and Pakstan workers may be complementary to other ethnc mnorty workers. Mxed workers may be substtutes for Black, and complementary to Mxed workers. Mxed workers may be substtute for Black and complmentary to Indan workers. Own-wages elastctes are negatve for Black and Indan, as expected but those are postve for Pakstan and Mxed. It s notable that the coeffcents of the shares are larger among ethnc mnorty groups than among Whte and ethnc mnortes. Ths outcome s n lne wth earler studes. Table 4. Estmated wage elastctes for ethnc mnorty groups Wth respect to Changes n the wage of quantty of BLACK INDIAN PAKISTANI MIXED Elast. t Elast. t Elast. t Elast. T Percentage Black Percentage Indan Percentage Pakstan Percentage Mxed t statstcs are the estmated coeffcents of earnngs functons presented n Table A2 n Appendx Concernng the coeffcents of the varables n the earnngs functons presented n Table A1 n appendx, the unemployment rate has a small negatve effect on the wages of all Whte workers. The sgnfcance level of ths effect s hgher for lower skll levels, e.g. lower t- statstcs for hgh sklled manual and non-manual workers. In the earnngs functons of ethnc mnortes n Table A2, the coeffcents for the unemployment rate are nsgnfcant. It s even notable that the coeffcent for Indan workers s postve whch mples that the wages of Indan and Mxed workers are not affected by unemployment. As expected, experence, tenure and the number of workng hours have a hghly sgnfcant effect on earnngs, as presented n Table A1 and A2. However, the coeffcents are not sgnfcant for Pakstan/Bangladesh workers. A possble reason could be the relatvely young age structure of the Pakstan/Bangladesh workers. Female workers earn around 20 percent less than male and marred workers do. The dsadvantage s relatvely less for women wth a hgh educaton who are employed n non-manual jobs as well as for women from Black and Mxed groups. Temporary and, especally, part-tme jobs pay notably lower wages than permanent and full-tme. Wages for workers who are employed n frms wth less than 20 workers, are 8 to 31 percent lower than large scale companes wth more than 20 employees. The prvate sector offers 4 to 9 percent greater wages than the publc sector for manual workers and Black workers. Unon membershp has a postve mpact on earnngs but 14

15 A. Zorlu/Ethnc mnortes n the UK (2000), 15 ths effect s relatvely small for non-manual workers and negatve for hgh sklled manual workers. Workers who commute between home and work earn 4 to 13 percent more than non-commuters do. Workers wth health problems have lower earnngs than healthy workers do, except workers from the Indan, Pakstan and Mxed categores. Earnngs dsperson among ndustres and geographcal areas s substantal. In general, manufacturng offers slghtly hgher wages for all types of labour whle hotels, educaton and wholesale ndustres pay lower wages although not all coeffcents are sgnfcant. Only Pakstan workers n wholesale ndustres and educaton are better off. Employment n publc admnstraton and defence pays relatvely hgher wages for low sklled workers and relatvely lower wages for hgher sklled workers. All workers who are employed n London earn sgnfcantly hgher wages. However, Whtes n London have a bgger advantage than non-whtes. Only workers from the category Mxed workers who are employed n London earn wages whch are comparable wth earnngs of Whte workers. Other mnorty workers n the same area have relatvely low wage rates. The coeffcent of havng a health problem has a surprsngly postve sgn for Indan, Pakstan/Bangladesh and Mxed workers but t s not sgnfcant. Generally, the coeffcents of many explanatory varables for ethnc mnortes have more or less the same sgns as those n the earnngs functons for Whtes. However, there are some dfferences among earnngs functons for Whtes: the sgns of coeffcents for dummy varables for the prvate sector are, n general, postve and the coeffcents for the manufacturng-dummy s not always postve. In addton, the t-statstcs for these varables are relatvely low. 5 THEORY: WHEN PRODUCTION FACTORS ARE INTERRELATED The emprcal analyss of the competton between dfferent categores of labour, whch can be determned on the bass of skll, occupaton, ethnc orgn and so on s typcally based on neo-classcal nput demand theory (Johnson 1980; Grant and Hamermesh 1981; Grossman 1982; Borjas 1985; Borjas 1987; Altonj and Card 1991; Lalonde and Topel 1991). The man methodologcal nstrument of ths theory s to determne the extent of the competton s the estmaton of substtuton parameters among type of labour and between labour types and captal. A prelmnary procedure s the choce of a functonal form of producton functon and ts estmaton to determne technology coeffcents. Then, elastctes of substtuton and complementarty among producton factors can be calculated. On the bass of these elastctes, we can examne the degree of competton between labour nputs. In the lterature on estmaton methodologes of demand for types of producton factors, the dscusson has been focussed on the choce between estmatng a cost functon or a producton functon, the choce of functonal form and sample, and dsaggregaton of labour force. Hamermesh and Grant (1979) argue that emprcal methodology makes sense for outcomes of substtuton elastctes because estmatons are based on dfferent assumptons. They suggest that a choce of producton functon approach s a more approprate methodology when quantty of labour, rather than wages, s assumed to be exogenous. Most of the choces mentoned here are restrcted by avalablty of sutable data. So, captal s excluded n ths study because data about captal s not avalable n the UK. Snce ths study examnes substtuton relatonshps only among types of labour, a rsk exsts that estmatons may be based. Hamermesh and Garant (1979) argue that estmates of own-prce elastcty wll be based towards zero when a seperablty assumpton s not satsfed. By the excluson of nput captal, the effect of captal on the substtuton elastctes s gnored and a strong seperablty between labour and captal s mplctly assumed. Snce we do not have data about captal, we are unable to test the seperablty and must assume that the seperablty assumpton holds. In prevous studes, whch have been overwhelmngly conducted n the Unted States, there s some evdence of substtuton between natve and mmgrant labour n lne wth theoretcal 15

16 A. Zorlu/Ethnc mnortes n the UK (2000), 16 predctons. Grant and Hamermesh (1981) analyse the competton among youth, adult Whte male, Whte female and Black workers n manufacturng usng translog producton functons. They fnd a weak substtutablty between Black and Whte men, and a strong substtutablty between youth and Whte women workers. Usng a comparable method, Grossman (1982) studes complementarty among natve, foregn-born and second generaton workers, and reports that both foregn-born and second generaton workers are substtutes for natve workers but second generaton workers are stronger substtutable. Foregn-born workers are stronger substtutes for second generaton than for natve workers. Captal s complementary to all knds of labour but the complementarty s the strongest for mmgrants. Usng the generalsed Leontef producton functon, Borjas (1983) fnds no ndcaton for substtutablty between Black and Hspanc workers. In hs work, Hspanc and Whte labour are complements whle Black and Whte labour are nether substtutes nor complements. In hs later work, Borjas (1986) fnds competton between mmgrant and Whte men as well as between women and Whte men but also fnds complementarty between mmgrant and Black males. Borjas (1987) dsaggregates labour nputs further. Whte, Black, Hspanc, and Asan workers are all dvded nto two categores: natve-born and foregn-born. Immgrants as a group are substtutes for Whte natve-born workers. Black natve-born workers are weakly substtutes for Black and Hspanc mmgrants whle they are strongly complementary to Whte mmgrants. He fnds no evdence of substtutablty between Hspanc natve-born and other natve-born populaton groups. Grffn (1996) uses frm level data to detect substtutablty relatonshps among varous types of labour nput and fnds that Whte male and female workers are strong substtutes as well as mnorty male and Whte female workers. On the other hand, mnorty men and women seem to be closely complementary n producton. New and Zmmermann (1994) use wage functons to measure competton between blue/whte collar and mmgrant workers. They show a small postve effect of the shares of mmgrants n ndustry sectors on the wages of whte-collar German workers but a large nverse effect on the wages of blue-collar German workers. They conclude that mmgrants on the whole are substtutes for natve German workers. Ths study goes a step further. Ethnc mnortes are frstly dvded nto the three skll categores and then nto four ethnc groups. The dsaggregaton of Whtes nto 6 dfferent categores allows nvestgaton of labour market competton among sx types of Whte labour and 3 skll- /4 ethnc categores of mmgrant labour. The dsaggregaton of labour force n many categores may provde more nsght nto the labour market competton between varous types of labour nputs. Suppose that an economy produces Y unts of goods employng eght types of labour. The producton functon n a gven labour market s gven by: 1. Y = f L,..., L ), f > 0, f < 0, N = 1,2,,N ( 1 N We prefer the producton functon approach because the quantty of labour s more lkely exogenous than wages as long as the man focus here s on mmgrant labour, and nternatonal mgraton s hghly restrcted. It s assumed that producton functon exhbts standard neo-classcal propertes and s lnearly homogenous. In a perfectly compettve market, the frst order condton for proft-maxmsaton s gven. Y 2. = w, = 1,2,3,.,N L where w refers to wage rate of labour nput. Equaton 9 demonstrates that labour nput s pad the value of ts margnal productvty. The analyss of multfactor producton functon requres more specfcaton of producton technology. A transcendental logarthmc (translog) producton functon s a functonal form, 16