Migration-Induced Women s Empowerment: The Case of Turkey

Transcription

1 RSCAS 2013/77 Rober Schuman Cenre for Advanced Sudies Migraion Policy Cenre Migraion-Induced Women s Empowermen: The Ca of Turkey Şule Akkoyunlu

2

3 European Universiy Insiue Rober Schuman Cenre for Advanced Sudies Migraion Policy Cenre Migraion-Induced Women s Empowermen: The Ca of Turkey Şule Akkoyunlu EUI Working Paper RSCAS 2013/77

4 This ex may be downloaded only for personal rearch purpos. Addiional reproducion for oher purpos, wheher in hard copies or elecronically, requires he conn of he auhor(s), edior(s). If cied or quoed, reference should be made o he full name of he auhor(s), edior(s), he ile, he working paper, or oher ries, he year and he publisher. ISSN Şule Akkoyunlu, 2013 Prined in Ialy, Ocober 2013 European Universiy Insiue Badia Fiesolana I San Domenico di Fiesole (FI) Ialy cadmus.eui.eu

5 Rober Schuman Cenre for Advanced Sudies The Rober Schuman Cenre for Advanced Sudies (RSCAS), creaed in 1992 and direced by Brigid Laffan since Sepember 2013, aims o develop iner-disciplinary and comparaive rearch and o promoe work on he major issues facing he process of inegraion and European sociey. The Cenre is home o a large pos-docoral programme and hoss major rearch programmes and projecs, and a range of working groups and ad hoc iniiaives. The rearch agenda is organid around a of core hemes and is coninuously evolving, reflecing he changing agenda of European inegraion and he expanding membership of he European Union. Deails of he rearch of he Cenre can be found on: hp:// Rearch publicaions ake he form of Working Papers, Policy Papers, Disinguished Lecures and books. Mos of he are also available on he RSCAS websie: hp:// The EUI and he RSCAS are no responsible for he opinion expresd by he auhor(s). Migraion Policy Cenre (MPC) The Migraion Policy Cenre (MPC) a he European Universiy Insiue, Florence, conducs advanced rearch on global migraion o rve migraion governance needs a European level, from developing, implemening and monioring migraion-relaed policies o asssing heir impac on he wider economy and sociey. The MPC carries ou field as well as archival rearch, boh of which are scienifically robus and policy-relevan, no only a European level, bu also globally, argeing policy-makers as well as poliicians. This rearch provides ools for addressing migraion challenges, by: 1) producing policyoriened rearch on aspecs of migraion, asylum and mobiliy in Europe and in counries locaed along migraion roues o Europe, ha are regarded as prioriies; 2) bridging rearch wih acion by providing policy-makers and oher sakeholders wih resuls required by evidence-bad policymaking, as well as necessary mehodologies ha address migraion governance needs; 3) pooling scholars, expers, policy makers, and influenial hinkers in order o idenify problems, rearch heir caus and conquences, and devi policy soluions. Our rearch includes a core programme and veral projecs, mos of hem co-financed by he European Union. The MPC working paper ries, published since April 2013, aims a disminaing high-qualiy rearch peraining o migraion and relaed issues. All EUI members, as well as oher exernal scholars and praciioners, are welcome o submi heir work o he ries. For furher queries, plea conac he Migraion Policy Cenre Secrearia a mpc@eui.eu More informaion can be found on: hp:// Disclaimer: The EUI, RSCAS and MPC are no responsible for he opinion expresd by he auhor(s). Furhermore, he views expresd in his publicaion canno in any circumsances be regarded as he official posiion of he European Union.

6

7 Absrac Migraion no only conribues o developmen hrough financial remiances, bu also hrough flows of knowledge and hrough he diffusion of social, culural and poliical norms and values. In fac, he more inangible conribuions are more appreciaed during economic and financial cris, as financial remiances become unsable or decrea in ho circumsances. This paper, herefore, address he effec of migraion on women s empowermen in Turkey. The number of women in parliamen in Turkey is chon as a gauge of women s empowermen and is explained by he emigraion rae, he relaive educaion of women o men, and a measure of democracy. Uilizaion of daa over six decades from 1960 unil 2011 gives he possibiliy ha he ries can be spuriously correlaed. Therefore, he paper address he issue of spurious correlaion in an analyical way. Spurious correlaion is he risk of linking he share of women in parliamen, for example, o he emigraion rae when in fac here is no associaion. This sudy adops he bounds esing procedure as a mehod o deermine and o avoid spurious correlaion. The resuls of bounds esing gives clear-cu evidence ha women s empowermen, he share of women in parliamen in he pren conex, is relaed o he emigraion rae, he relaive educaion of women and o a measure of democracy. The bounds-esing procedure is replicaed for emigraion flows by desinaion counry groups such as European and oher core OECD counries, Arab counries, and Russia and CIS (Commonwealh Independen Saes) counries. Again, i is found ha he share of women in parliamen is relaed o he counry groups wih he larges effec in European and core OECD counries. The resuls are robus for he inclusion of asylum ekers and refugees in he emigraion daa. The resuls have imporan policy implicaions for nding as well as for desinaion counries, implicaions which are discusd in he paper. Keywords Emigraion, Social Remiances, Women s Empowermen, Women share in parliamen, Turkey

8

9 1. Inroducion * The migraion-developmen nexus considers hree Rs recruimen, remiances and reurn. Remiances are financial remiances, and he majoriy of migraion and developmen rearch focus on he effecs of financial remiances on nding counries. However, migrans are expod o differen values, norms, narraives, and forms of behaviour in differen hos counries. The ransfer of he differen values, norms, narraives, and forms of behaviour, as well as he knowledge and skills ha are acquired in hos counries ermed social remiances can have a subsanial impac on he economic and human developmen of home counries. More imporanly, social remiances are more sable and robus even during he financial and economic crisis. Figure 1 below shows he growh of financial remiances o Turkey and i is eviden ha financial remiances are raher unsable and flucuae, since hey are nsiive o he economic and financial siuaion in hos counries and, herefore, hey can be unsable or can decrea during economic downurns. Figure 1: Growh of financial remiances o Turkey Therefore, his paper sudies he impac of social remiances on a ypical emigran counry Turkey. Firs, his sudy conribues o he lieraure on social remiances, and hence o he lieraure on he migraion-developmen nexus. Second, we examine Turkey, which has been an imporan migran nding counry since he 1960s, desinaion counries for Turkish ciizens having varied over he years. Hence, he clo invesigaion of emigraion by desinaion counries will highligh how social norms and how he values of migrans in differen desinaions are formed differenly. Third, he paper invesigaes he effec of emigraion on women s empowermen he share of women in he Turkish parliamen along wih oher variables such as he relaive educaion of women and a measure of democracy. To he bes of our knowledge his is he firs sudy which invesigaes social remiances in Turkey as well as he deerminans of he share of women in parliamen. This is an imporan ask * I am graeful o Professors Alessandra Venurini and Philipp Fargues for heir encouragemen, suppor, suggesions, and commens; all remaining errors are, of cour, my own. I also hank o paricipans of he MPC Annual Conference (21 June 2013) and he Economics Deparmen of Vienna Universiy of Economics and Business (30 Sepember 2013) for heir commens. 1

10 Şule Akkoyunlu given ha one of he Millennium Developmen Goals is o promoe gender equaliy and o empower women. Fourh, a sudy ha is paricularly relevan finds ha migraion o counries wih higher woman poliical empowermen significanly increas he share of women in parliamen in he home counry (Lodigiani and Salomone, 2012). However, heir daa cover he period and migraion daa is available only by decade. Our daa are annual and cover he pos-2000 period when here was a major shif in he parliamenary paricipaion of women in Turkey. Fifh, he uilizaion of daa over six decades from 1960 unil 2011 gives he possibiliy ha he ries can be spuriously correlaed. Therefore, he paper address he issue of spurious correlaion in an analyical way. Spurious correlaion enails he risk of linking he parliamenary share of women, for example, o he emigraion rae when, in fac, here is no associaion. This sudy adops he bounds esing procedure sugged by Pesaran, Shin and Smih (2001) o deermine and hwar spurious correlaion. The bounds esing procedure as oulined in Secion 5 has more advanages han oher mehods and can deec he genuine relaionship beween he variables in quesion. This is, we sress again, he firs sudy ha employs he bounds esing procedure in he field of social remiances. Sixh, asylum ekers and refugees migh have sronger views abou heir home counries and heir insiuions and, herefore, migh ransfer more social values, norms and ideas han labour migrans. However, i is also possible ha hey migh break conac wih he home counries and sar a new life in a new counry: in his ca hey, naurally, do no have an impac on he home counry s insiuions. Therefore, he paerns of behaviour of asylum ekers and refugees migh differ from oher migrans and, in order o ake his ino accoun, we also include asylum ekers and refugees in he emigraion daa and replicae he exerci. Las, bu no leas, he paper discuss he poliical implicaions of he findings for boh nding and receiving counries, since he empirical resuls deliver imporan insighs. Migrans are, indeed, expod o differen social norms and values in differen desinaions, and, herefore, ransfer desinaion-specific norms and values ha can have differen and diver impacs on nding counries. The res of he paper is organized as follows: cion 2 describes he social remiances. Secion 3 discuss women in poliics in Turkey; cion 4 describes he daa and is sources; in Secion 5 he bounds esing procedure of Pesaran e al. (2001) is described in deail; in Secion 6 he es resuls as well as he final models are prened; and he final cion concludes by delivering imporan policy implicaions. 2. Social Remiances Social remiances are defined as ideas, know-how, norms, values, knowledge, behaviour, pracices and skills ha migrans bring wih hem or ha hey nd back. The promoe or deer developmen in home counries, (Levi, 1998, 2001, and Levi and Lamba-Nieves, 2011). Levi (2001) argues ha, in fac, social remiances are more imporan han financial remiances. She menions four ypes of social remiances ha are ransferred from migrans o heir home counries: norms, pracices, ideniies and social capial. Social remiances can be diffud by migrans as well as by refugees. They are ransferred when migrans reurn or visi heir home counries; when non-migrans visi heir family and friends in he desinaion counries; or when a leer (or phone-call, fax, , video...), is received (Levi and Lamba-Nieves, 2011). Social remiances do no only affec family relaions, economic and social well-being, gender roles, class and race ideniy. They also have a subsanial impac on poliical, social, culural, economic and religious paricipaion. They can challenge people s ideas, beliefs and views abou, among oher hings, democracy, poliics, insiuions, healh, culure, sociey, religion, echnology, science, business, economics, educaion, and gender issues. For example, Jimenez (2008) poins ou ha more migrans and non-migrans paricipaed in poliics as a resul of inernaional migraion and ha poliical power beween acors shifed. Kapur (2008) also confirms ha migrans change he 2

11 Migraion-Induced Women s Empowermen: The Ca of Turkey poliical power beween differen acors, ake policy iniiaives and affec poliical insiuions in heir home counries. Baisa and Vicene (2011) show ha curren as well as reurn migrans incread poliical paricipaion in Cape Verde. Docquier e al. (2011) find ha emigraion has a srong impac on democracy as well as economic freedom in home counries. Similarly, Spilimbergo (2009) demonsraes ha foreign sudens diffu democracy if heir educaion is acquired abroad in democraic counries. Comparaively, in a very recen sudy i has been shown ha higher voes for he communis pary, a he disric level in Moldova, are explained by migraion o Russia, while a negaive correlaion is associaed wih migraion o he EU (Omar Mahmoud e al., 2013). Beine and Sekka (2012) find ha emigraion has an impac on he qualiy of insiuions 1 in he nding counries, and his depends especially on wheher emigrans are locaed in counries wih beer qualiy insiuions. For he diffusion of educaion, Sark (2002) shows ha highly-skilled immigrans encourage he non-migran populaion in heir home counries o inves in educaion, which generaes more human capial han if no brain drain had aken place. 2 Likewi, migraion can affec gender imbalances in veral ways. Belanger and Rahman (2013) show ha, hough he inernaional migraion of Bangladeshi women liberaed women economically, i also reinforced a pariarchal ideology and incread women s vulnerabiliy and subordinaion. Lindsrom and Munoz-Franco (2005) find ha knowledge of conracepive mehods spread from migrans o non-migrans in rural Guaemala. Frank (2005) shows ha inernaional migraion has a posiive impac on infan healh (birh weigh) in Mexico, in spie of he loss of social suppor and high sress levels due o he migraion of he family members. Mexican women wih inernaional migran parners have, in fac, significanly lower raes of smoking, higher levels of exerci, and gain more weigh during pregnancy han heir counerpars in Mexico wihou migran parners. However, he ransfer of pracices does no always resul in uniformly posiive changes in behavioural pracices, as women wih inernaional parners were less likely o breasfeed heir infans. Fargues (2007, 2011) find ha birh raes in Morocco and Turkey wen down, becau migrans wen o Europe, where hey encounered smallsize families and individualisic values. In conras, in Egyp birh raes ro, becau mos Egypian migrans wen o he Gulf, where hey encounered larger families han ho lef behind and deeperrooed pariarchal values. Likewi, Beroli and Marchea (2013) show ha Egypian houholds wih a reurnee husband from anoher Arab counry have a significanly larger number of children. Similarly, Beine, Docquier and Schiff (2013) demonsrae ha he ransfer of norms from low-feriliy desinaion counries reduces feriliy in migrans home counries. The sudies demonsrae ha social remiances are, indeed, an imporan par of he migraion experience. 3. Women in Poliics in Turkey Turkish women achieved poliical righs o elec and o be eleced locally 3 April 1930 and naionwide 5 December 1934, well in advance of women in European counries such as France, Ialy and Belgium. Unil muli-pary democracy in 1946, here was an unofficial quoa of 4-5 percen for women in parliamen. However, afer he incread compeiion in poliics in he 1950s, women s share in parliamen dropped o 0.61 percen. The figure remained a 1-2 percen unil he 1990s wih he excepion of he 1983 elecions when he share ro o 3 percen. Since he mid-1990s, he share of women parliamenarians has incread. However, he figure remains low compared o oher European counries. Several facors can explain he increa in women s parliamenary share in Turkey over he pas decades. Firs, he Fifh Five Year Developmen Plan ( ) and he General Direcorae for he Saus and Problems of Women in 1990 recognid women s issues and aimed a srenghening he posiion of women in social, economic and poliical life, and a providing he equal 1 2 They employ four indicaors for he qualiy of insiuions, namely, voice and accounabiliy, governmen effeciveness, regulaory qualiy, and conrol of corrupion. See also Mounford (1997), Beine e al. (2001), Kaz and Rapopor (2005) and Mariani (2007) for migraion-induced human capial invesmen. 3

12 Şule Akkoyunlu uilizaion of righs and opporuniies. Second, criminal and civil law was reformed in The Turkish governmen gave, a his poin, equal righs o women and men during marriage, divorce, and propery righs. As a resul of his reform, programmes were inroduced o improve access o educaion for girls. Third, veral measures were aken o increa schooling for girls, by he Turkish governmen and NGOs, (Caha, 2010). Third, he KA.DER associaion (The Associaion for he Suppor and Training of Women Candidaes) was esablished in 1997 o suppor women candidaes boh wihin he poliical paries and also publicly a general elecions, and o increa he number of women in poliics and in decision making posiions. Fourh, he Republican People s Pary (RPR) inroduced a gender quoa of 25 percen in 1989 and incread i o 33 percen in Whereas, some paries such as he Jusice and Developmen Pary (JDP) had adoped sof gender quoas. Fifh, emigrans, by bringing back home social, culural and poliical norms, values, ideas, and narraives from he more democraic desinaion counries, migh have encouraged women s paricipaion in poliics in Turkey. I is widely recognid ha he paricipaion of women in poliics in Turkey was no limied by legal means: i was raher a quesion of culural, radiional and menal obsacles (Kasapoglu and Ozerkmen, 2011; Cansun, 2013). Cansun (2013) finds ha Turkish men do no wan o share houwork and child-care responsibiliies wih women, and ha; herefore, Turkish women do no have ime for poliics. Poliics is considered o be a male domain in Turkey, and i is believed ha Turkish men are more naural poliicians han Turkish women. More imporanly, Turkish men would be jealous of heir spous if hey worked in male-dominaed occupaions such as poliics (Cansun, 2013). The following (invisible) barriers conribue o he low paricipaion of women in poliics in Turkey. Pariarchal srucure: Turkish women follow heir husbands or parens in poliical decisionmaking (Caporal, 1982). Turkish women view poliics as an arena of lies, fraud, unfulfilled promis, and corrupion and so prefer oher occupaions (Keskin, 1997; Kasapoglu and Ozerkmen, 2011). Turkish women are financially consrained and do no have much capial for poliical campaigns (Keskin, 1997; Kasapoglu and Ozerkmen, 2011; Cansun, 2013). The low level of educaion among Turkish women hinders heir paricipaion in an eliis democracy (Koray, 1995; Kasapoglu and Ozerkmen, 2011). The Islamis view (fira) argues ha women should carry ou family roles while men deal wih oher issues (Ergil, 1975). However, he increa in women s parliamenary share in Turkey over he pas decades suggess ha he obsacles migh be fading away: he pariarchal srucure is weakening and Turkish women are enering ino a more independen decision-making process. This paper, herefore, quesions wheher Turkish emigraion has conribued o he change in menaliy/aiude owards he paricipaion of Turkish women in poliics, along wih oher imporan variables, such as he relaive educaion of women and a measure of democracy. Thus, we quesion wheher Turkish migrans, by ransferring culural, social and poliical ideas, norms and values from he hos counries, have conribued o he ransformaion of he poliical lives of women in Turkey. 4. Daa We will sudy he deerminans of women s parliamenary share in Turkey wih he emigraion rae as well as wih oher imporan poenial variables, such as he relaive educaion of women and a measure of democracy. The main variables in our analysis are: The share of women in he Turkish parliamen was obained from he Grand Naional Asmbly of Turkey. 4

13 Migraion-Induced Women s Empowermen: The Ca of Turkey The emigraion rae is defined as he number of Turkish migrans divided by he Turkish populaion. Emigraion daa were gahered from he General Direcorae for Securiy in Turkey, whereas he daa on he Turkish populaion were obained from he Turkish Saisical Insiue (TSI). The emigraion daa consis of Turkish naionals who are residen in Turkey as well as abroad. The emigraion of Turkish naionals who are residen abroad is especially relevan in our ca, as hey are he ones mos likely o ransfer social, culural, and poliical ideas, norms and values o Turkey, by having regular conacs wih Turkish ciizens lef behind. Alhough, he majoriy of emigraion daa covers labour migrans, he daa also include sudens as well as visiors for he following purpos: culure, holiday, relaives, spors, enerainmen, business, and medical reamen. The educaion variable for women is defined as he raio of girls o boys in primary and condary schools 3 and was gahered from he Educaion Saisics of he TSI. The Freedom Hou Index for Liberal Righs 4 is ud as a measure of democracy. The index covers he main aribues of he Universal Declaraion of Human Righs. The civil liberies quesions consis of four sub-caegories: Freedom of Expression and Belief (4 quesions), Associaional and Organizaional Righs (3 quesions), Rule of Law (4 quesions), and Personal Auonomy and Individual Righs (4 quesions). 5 Then, he index is consruced by summing up each counry s subcaegory. The index ranges from 1 o 7. Counries wih a raing of 1 enjoy liberies in freedom of expression, asmbly, associaion, educaion, and religion. Counries wih a raing of 7 have few or no civil liberies and, herefore, do no have freedom of expression or associaion. They do no proec he righs of deainees and prisoners, and ofen hey conrol or dominae economic aciviy. According o his index Turkey currenly proecs, a leas o a moderae exen, almos all civil liberies: Parly- Free. This is a fair asssmen. The daa on emigraion by couny of desinaion were gahered from he General Direcorae for Securiy in Turkey. The desinaion counries are grouped ino hree groups: EU and OECD Counries: Ausria, Germany, Belgium, Denmark, Finland, France, he Neherlands, he UK, Ireland, Spain, Sweden, Swizerland, Ialy, Iceland, Luxemburg, Norway, Porugal, Greece, Ausralia, USA, Japan, Canada, New Zealand, and Israel. Arab Counries: Iraq, Lebanon, Syria, Saudi Arabia, Oman, Jordan, Unied Arab Emiraes, Qaar, Kuwai, Yemen, Bahrain, Algeria, Morocco, Egyp, Tunisia, and Libya. Russia and CIS counries: Russia, Azerbaijan, Kazakhsan, Kyrgyzsan, Uzbekisan, Tajikisan, and Turkmenisan. We also included Asylum Seekers and Refugees in he emigraion daa and repeaed he exerci as a robusness check. The daa on Asylum Seekers and Refugees come from he UNHCR Populaion Saisics, and are available by counry of desinaion. Oher imporan poenial variables, such as per capia real GDP, were also considered in he esimaes, bu hey did no produce any significan resuls. The labour force paricipaion of women was anoher imporan poenial variable, bu his variable is only available from 1988, and i is srongly correlaed wih he relaive educaion of women We also ried he raio of women o men in eriary educaion in Turkey, bu his daa did no produce any significan resuls. We also ried he Freedom Hou Index for Poliical Righs, POLITY IV, and Economic Freedom of he World as a measure of democracy, bu he Freedom Hou Index for Liberal Righs produced beer resuls. hp:// In our sudy, we also considered he possible effecs of emigraion on he relaive educaion of women and also on he measure of democracy. Therefore, we ook ino accoun he indirec effecs of emigraion on women s parliamenary 5

14 Şule Akkoyunlu We were, herefore, able o analy he deerminans of he parliamenary share of women in Turkey wih he emigraion rae, he relaive educaion of women and a measure of democracy. Imporanly, figures 2, 5, 8, and 11 show he salien feaures of he daa ha he share of women in parliamen has cerain common feaures wih he emigraion rae, he relaive educaion of women and he democracy variable. This visual inspecion suggess ha he share of women in parliamen, he emigraion rae, he relaive educaion of women and he democracy variable end o move ogeher and leaves saisical mehods o verify his conclusion. For example, an increa in women s parliamenary share in he pos-1990 period maches cloly he increas in he emigraion rae and he relaive educaion of women in figures 2 and 5. 7 However, over he enire period, he share of women in parliamen and he relaive educaion of women em o be highly correlaed. Therefore, we expec o e he larges effec on women s parliamenary share from he relaive educaion of women. In addiion, democraizaion in he pos-2000 period ems also o conribue o women s empowermen. Likewi, he share of women in parliamen, emigraion share o he EU and OECD counries, and emigraion share o he Arab counries all ended o increa in he lae-1960s, 1970s and he early-1980s. Therefore, he salien feaures of he daa and heir relaionship over he enire period sugges a formal economeric analysis o deermine he relaionships. 5. Mehod The bounds esing procedure of Pesaran e al. (2001) which is ud o address he spurious correlaion problem as well as o deec a long-run relaionship beween he variables of ineres will be prened in his cion. The choice of his procedure can be jusified bad on he following reasons. Firs, he bounds esing procedure has a sraighforward applicabiliy, as he regressors can be eiher I(1), I(0), or muually coinegraed. This is of grea imporance, since he differen uni roo ess can deliver differen resuls wih respec o he order of he inegraion of he relevan variables, yielding inconclusive resuls as well as pre-esing bias. Second, he bounds esing procedure allows a join esimae of shor- and long-run effecs, since i is bad on he unresriced error correcion model. Banerjee e al. (1993, 1998) show ha join esimaes have beer saisical properies han he wop Engle Granger procedure, which shifs he shor-run dynamics ino he error erm. Third, he bounds esing procedure performs very well in small samples, (Narayan, 2005). In small samples, oher mehods such as he more popular Full Informaion Maximum Likelihood Mehod (Johann, 1995) perform poorly. The bounds esing procedure will, in he pren conex, be performed by relaing women s parliamenary share o he emigraion rae, he relaive educaion of women and a measure of democracy according o a vecor auoregressive (VAR) model of order p ha can be furher reduced o he following condiional error correcion model (ECM): WP WP p i0 1 Educ i 0 Emig i 1 p i0 1 FHI i Educ 2 i 1 D FHI 3 1 p i1 WP i i p i0 Emig i i (1) (Cond.) share in Turkey as well as he correlaion beween he independen variables. However, we found ha emigraion had no effec on he relaive educaion of women, and on he measure of democracy; herefore, he independen variables are no correlaed. 7 Figures 2 and 5 do no differ significanly in erms of emigraion daa: wihou and wih asylum ekers and refugees. Thus, he inclusion of asylum ekers and refugees in our emigraion daa does no change he main paern of he emigraion daa. Similarly, he inclusion of asylum ekers and refugees o he emigraion daa by desinaion counries in figures 8 and 11 does no make much of a difference. The obrvaions are also confirmed by similariies in he coefficien esimaes of emigraion raes wih and wihou asylum ekers and refugees. 6

15 Migraion-Induced Women s Empowermen: The Ca of Turkey WP, Emig, Educ and FHI denoe, respecively, women s parliamenary share, he raio of emigraion o Turkish populaion, girls o boys raio in primary and condary schools, and he Freedom Hou Index for liberal righs. The lagged values of WP, Emig, Educ, and FHI (are in logs) pren, respecively, a long-run relaionship in levels. and D are he deerminisic erms; consan and dummy variables. The shor-run dynamics is reprened by lagged values of WP and curren and lagged values of Emig, Educ and FHI. The condiional long-run elasiciies of women s parliamenary share wih respec o Emig, Educ and FHI are given by, respecively, 1 /0, 2 / 0 and 3 /0 (Banerjee e al., 1998). The bounds esing procedure uilizes he convenional F-es for esing he null hypohesis H 0 : Alhough Pesaran e al. (2001) provide a of asympoic criical values, generaed from 1,000 boosraps, he values are no appropriae in our ca due o our small sample size. We, herefore, employ he criical values simulaed in Narayan (2005) for he hypohesis esing for a small sample size. Two s of criical values are imporan for he hypohesis esing. The firs is he lower bound, applicable when all regressors are I(0). The cond is he upper bound, applicable when all regressors are I(1). If he calculaed F-saisic falls below he lower bound, he null hypohesis of no relaionship beween women s parliamenary share, and he emigraion rae, he relaive educaion of women and a measure of democracy canno be rejeced. However, if he calculaed F-saisic exceeds he upper bound, he null hypohesis of no long-run relaionship beween he variables of our ineres is rejeced. Boh cas hold irrespecive of he order of inegraion of he regressors. Finally, if he F-saisic falls wihin he lower and upper criical bounds, he order of inegraion of he variables mus be deermined in order o draw conclusive inference. 6. Tes Resuls and he Final Models Equaion (1) was esimaed for he four cas: (i) he emigraion rae is defined as emigraion divided by he Turkish populaion; (ii) daa on asylum ekers and refugees are included in he emigraion daa, and he emigraion rae is defined as emigraion including asylum ekers and refugees divided by he Turkish populaion; (iii) emigraion daa are grouped ino hree groups according o desinaion counries; and (iv) daa on asylum ekers and refugees are included in he hree desinaion groups. 6.1 Aggregae Emigraion wihou Asylum Seekers and Refugees As a saring poin, we define he lag order p for he Equaion (1). Table (1) displays he resuls of he lag order p lecion procedure. The informaion crieria (Akaike, AIC, and Schwarz, SIC) and he Lagrange Muliplier saisic esing for remaining auocorrelaion up o he firs, fourh, and sixh orders in regression residuals are ud for he lag order lecion. The daa cover he ime period from 1972 unil 2011, as he Freedom Hou Index is available only from Boh informaion crieria (AIC and SIC) lec p=5. In addiion, here is no evidence of any remaining auocorrelaion in he regression residuals, when p=5. Given he resuls from he informaion lecion crieria and he evidence of no residual auocorrelaion, he model wih p=5 is chon. The las column of Table 1 repors he corresponding F-es saisics for he join hypohesis H 0 : uilising he finie-sample criical values repored in Narayan (2005) for T=35, corresponding o ca III wih unresriced consan and no linear deerminisic rend. As shown in Table 1 he null hypohesis of no long-run relaionship beween women s parliamenary share, he emigraion rae, he relaive educaion of women, and a measure of democracy can be decisively rejeced for p=3 and p=5 a he 1% significance level. For p=1, p=2, and p=4 he es saisics falls below he lower bound a he 1% significance level, indicaing ha here is no long-run relaionship beween he variables of ineres. 7

16 Şule Akkoyunlu Table 1: Lag order lecion for he model wih aggregae emigraion daa p AIC SIC AR(1) AR(4) AR(6) III FH 0 0: *** *** Noes: p is he lag order of he underlying VAR model for he condiional ECM of Equaion (1). AIC and SIC are, respecively, he Akaike and Schwarz Informaion Crieria. AR(1), AR(4), and AR(6) are he p-values of he Lagrange Muliplier (LM) es saisics for esing, respecively, for residual auocorrelaion of orders up o one, four, and six. Bold enries repren he preferred lag order. III FH 0 is he F-es saisic for he null hypohesis H 0 using he 0: : finie-sample criical values simulaed in Narayan (2005, p. 1988) for T=35 corresponding o ca III wih unresriced consan and no linear deerminisic rends. *** indicaes ha he null hypohesis can be rejeced a he 1% significance level. ++ indicaes ha he es saisic falls below he lower bound a he 1% significance level, herefore, he null hypohesis of no relaionship in levels canno be rejeced. Afer esablishing he exince of a long-run relaionship beween women s parliamenary share, he emigraion rae, he relaive educaion of women, and a measure of democracy, we esimae he coefficiens of ineres as follows: we sar wih an error-correcion model corresponding o p=5; delee he insignifican augmenaion lags; and arrive a he parsimonious model in Equaion (2) (sandard errors in parenhes, and error probabiliies in brackes). In he model below (2), D2002 denoes an impul dummy ha is equal o 1 in 2002 and zero oherwi, since an oulier in 2002 was deeced as ho residuals exceeding regression sandard error by facor wo in he esimaed equaion (1). The dummy, in 2002, jusifies a decrea in women s parliamenary share in ha year ha canno be explained by he emigraion rae, he relaive educaion of women, and a measure of democracy. 8 8 In 2002, various poliical measures were aken o improve he social, economic and poliical saus of women in Turkey; however, we would expec he measures o have a posiive impac on women s parliamenary share. Therefore, his dummy canno repren poliical changes eiher. 8

17 Migraion-Induced Women s Empowermen: The Ca of Turkey WP WP 0.935Emig Educ 0.474FHI (0.345) (0.159) (0.169) (1.614) (0.151) 0.768WP 0.203WP 0.198WP 0.293WP (0.126) (0.114) (0.095) (0.082) 1.253Emig 2.467Emig 0.644Emig 0.650Emig (0.261) (0.322) (0.278) (0.372) 0.346Emig Educ 1.952Educ Educ (0.201) (2.726) (2.563) (2.362) (2) Educ Educ 06FHI 4 5 (2.778) (3.295) (0.223) 0.346FHI 2 (0.170) 0.381FHI 0.614FHI 0.412D (0.134) (0.139) (0.098) R , F(23,10) = 24.33[0.000], T = 34 AR(1 1) ARCH (1) F 4.689[0.058], F 0.466[14] (1,9) (1,8) 2Norm 0.965[0.617], (2) F RESET (1,9) 1.528[0.248] The model in (2) pass all he sandard specificaion ess such as ess for no residual auocorrelaion, for no residual ARCH effecs, for residual normaliy and he RESET es for funcional form misspecificaion. 9 This suggess ha his parsimonious model explains women s parliamenary share very well over he period under invesigaion, a period ha was characerized by veral naional poliical and economic cris, and major legislaive and social changes. This is furher suppored by he clo mach beween acual and fied values shown in he op panel of Figure 3; and he corresponding cross-plo is shown in he righ op panel. The esimaed regression residuals do no show any signs of misspecificaion in he lef boom panel of Figure 3. The auocorrelaion funcion up o fifh order which is displayed in he righ boom panel akes small values ha, moreover, change signs. Finally, Figure 4 shows he values of he one-p, breakpoin, and forecas Chow es saisics scaled by heir respecive 1% criical values (Doornik and Hendry, 2001). The ess show no signs of model insabiliy. According o Equaion (2), he long-run elasiciies of women s parliamenary share wih respec o he emigraion rae, he relaive educaion of women and a measure of democracy are, respecively, 0.49, 7.16 and The larges effec comes from he relaive educaion of women; 1% increas in he raio of girls o boys in primary and condary schools increas women s parliamenary share by 7%. This is a very large effec. Similarly, 1% increas in emigraion rae increas he share of women in parliamen by 0%. This is also a significan impac. Thus, emigraion conribues o women s empowermen in Turkey. In addiion, 1% increas in he measure of democracy increas women s parliamenary share by 0.25%, suggesing democraizaion encourages he empowermen of women. 9 Even hough he final model has 34 obrvaions and 24 parameers, herefore 10 degrees of freedom in esimaes, he sample size is only one of he facors ha deermine how much informaion appears in he sample, e Campos and Ericsson (1999). Figures 2, 5, 8 and 11 show ha he variabiliy of he daa, and hence he informaional conex of he daa are very high. In addiion, he large -raios sugges ha over-parameerizaion is no an issue in our ca. 9

18 Şule Akkoyunlu In he shor, changes in he relaive educaion of women and changes of a measure of democracy have a srong posiive impac. The larges effec comes, also, from he relaive educaion of women in he shor-run. However, changes in he emigraion raes in he curren period, and lagged one and hree periods have a negaive impac, implying ha in he shor-run he emigraion rae decreas he share of women in parliamen. Possibly, in he very shor-run poenial women candidaes for he parliamen are emigraing, bu as ime pass he migrans exer a posiive impac by ransferring he poliical ideas, norms and values from he hos counries, since changes in emigraion rae lagged four and five periods have a posiive effec. In addiion, i migh ake more han a few years o le in a foreign counry and o become acquained wih he social, culural and poliical norms and values, and hen o ransfer hem. Furhermore, he negaive effec of emigraion in he shor-run migh also occur, if he people who are emigraing are he mos influenial people and if hey are involved in he curren social, culural, and poliical arena and debae. However, overall resuls show ha Turkish migraion, indeed, expos he Turkish populaion o more democraic values and norms; and more imporanly migrans ransfer he values and norms o ho lef behind, somehing ha helps ransform Turkish sociey. 6.2 Aggregae Emigraion including Asylum Seekers and Refugees In his cion, we esimae he Equaion (1) by updaing he emigraion daa by adding he daa on asylum ekers and refugees from Turkey. On he one hand, asylum ekers and refugees migh no influence he social, culural and poliical dynamics of he counry lef behind, sar a new life in a foreign counry, and lo heir conacs wih heir home counry. On he oher hand, hey are he ones mos likely o make a change, as hey are he ones who speak abou insiuional failures. Therefore, we leave he empirical resuls o solve his dilemma. Again, we sar by defining he lag order p for he Equaion (1). Table (2) displays he resuls of he lag order p lecion procedure. The informaion crieria (Akaike, AIC, and Schwarz, SIC) and he Lagrange Muliplier saisic esing for remaining auocorrelaion up o he firs, fourh, and sixh orders in regression residuals are ud for he lag order lecion. The daa cover 1972 unil Boh informaion crieria (AIC and SIC) lec p=5. In addiion, here is no evidence of remaining auocorrelaion in he regression residuals, when p=5, as in he previous ca. Given he resuls from he informaion lecion crieria and he evidence of no residual auocorrelaion, he model wih p=5 is leced, as before. Therefore, he resuls are no affeced by he inclusion of asylum ekers and refugees in our daa. The las column of Table 2 repors he corresponding F-es saisics for he join hypohesis uilising he finie-sample criical values repored in Narayan (2005) for T=35, corresponding o ca III wih unresriced consan and no linear deerminisic rend. Table 2: Lag order lecion for he model wih aggregae emigraion daa including asylum ekers and refugees P AIC SIC AR(1) AR(4) AR(6) III FH 0 0: *** *** Noes: p is he lag order of he underlying VAR model for he condiional ECM of Equaion (1). AIC and SIC are, respecively, he Akaike and Schwarz Informaion Crieria. AR(1), AR(4), and AR(6) are he p-values of he Lagrange Muliplier (LM) es saisics for esing for, respecively, residual auocorrelaion of orders up o one, four, and six. Bold enries repren he preferred lag order. 10

19 Migraion-Induced Women s Empowermen: The Ca of Turkey III FH 0 0: is he F-es saisic for he null hypohesis H 0 : using he finie-sample criical values simulaed in Narayan (2005, p. 1988) for T=35 corresponding o ca III wih unresriced consan and no linear deerminisic rends. *** indicaes ha he null hypohesis can be rejeced a he 1% significance level. ++ indicaes ha he es saisic falls below he lower bound a he 1% significance level, herefore, he null hypohesis of no relaionship in levels canno be rejeced. As shown in Table 2 he null hypohesis of wih no long-run relaionship beween women s parliamenary share, emigraion rae, he relaive educaion of women, and a measure of democracy can be decisively rejeced for p=3 and p=5 a he 1% significance level, as before. For p=1, p=2, and p=4 he es saisic falls below he lower bound indicaing ha here is no long-run relaionship beween he variables of ineres a he 1% significance level. Hence, he resuls are no affeced by he inclusion of asylum ekers and refugees ino he emigraion daa. Afer esablishing he exince of a long-run relaionship beween women s parliamenary share, he emigraion rae, he relaive educaion of women, and a measure of democracy, we esimae he coefficiens of ineres as he same as before: we sar wih an error-correcion model corresponding o p=5; delee he insignifican augmenaion lags; and arrive a he parsimonious model in Equaion (3) (sandard errors in parenhes, and error probabiliies in brackes). In he model below (3), D2002 again denoes an impul dummy ha is equal o 1 in 2002 and zero oherwi for he oulier in WP WP 1.342Emig Educ 12FHI (0.378) (0.219) (0.231) (1.925) (0.163) 1.114WP 0.255WP 0.230WP 1.425Emig (0.158) (0.104) (0.088) (0.328) 3.712Emig 0.993Emig 1.466Emig Educ (0.395) (0.416) (0.411) (2.908) Educ Educ Educ 1.214FHI (3.010) (4.063) (4.346) (0.264) 0.293FHI 0.679FHI 0.415D (0.159) (0.172) (0.117) (3) R , F(19,14) = 15.14[0.000], T = 34 AR(1 1) ARCH (1) F 0.391[43], F 0.245[0.630] (1,13) (1,12) 2Norm 0.843[0.656], (2) F RESET (1,13) 0.086[0.774] The model in (3) pass all he sandard specificaion ess, including ess of no residual auocorrelaion, of no residual ARCH effecs, of residual normaliy and, again, he RESET es for funcional form misspecificaion. In addiion, acual and fied values follow each oher very cloly, as displayed in he op panel of Figure 6; and he corresponding cross-plo which is displayed in he righ op panel. The esimaed regression residuals do no show any signs of misspecificaion in he lef boom panel of Figure 6. The auocorrelaion funcion up o fifh order, which is displayed in he righ boom panel akes raher small values ha, moreover, change signs. Finally, Figure 7 displays 11

20 Şule Akkoyunlu he values of he one-p, breakpoin, and forecas Chow es saisics scaled by heir respecive 1% criical values (Doornik and Hendry, 2001). The ess again show no signs of model insabiliy. The long-run elasiciies of women s parliamenary share wih respec o he emigraion rae, he relaive educaion of women and a measure of democracy are, respecively, 0.60, 6.42 and % increas in he raio of girls o boys in primary and condary schools increas women s parliamenary share by 6.4%. This is again a very large effec. Similarly, 1% increas in emigraion rae increas he share of women in parliamen by 0.6%. This is also an imporan impac. Thus, emigraion including asylum ekers and refugees again has a posiive impac on women s empowermen in Turkey. In addiion, 1% increas in he measure of democracy increas women s parliamenary share by Thus our resuls are robus for he inclusion of asylum ekers and refugees in emigraion daa. In he shor-run changes in he relaive educaion of women and changes in he measure of democracy have had a srong posiive impac, especially changes in he relaive educaion of women. Changes in emigraion raes in he curren period, and lagged one, wo and hree periods have a negaive impac. Emigraion daa in his cion includes asylum ekers and refugees, implying ha Turkish migrans including asylum ekers and refugees were probably involved in poliics before hey lef Turkey and ha, herefore, heir abnce has a negaive impac in he shor-run. However, he posiive long-run impac of emigraion suggess ha emigrans including asylum ekers and refugees do no lo ouch wih heir fellow ciizens, and more imporanly ha hey ransfer ho social, culural and poliical norms and values accumulaed in he desinaion counries. 6.3 Emigraion by Desinaion Counry wihou Asylum Seekers and Refugees In his cion, we esimae he impac of emigraion by desinaion counry. Alhough he majoriy of Turkish labour migraion was direced owards European and he OECD counries: in he 1980s and 1990s Arab counries and also, in he 1990s, Russia and CIS counries played an imporan role in Turkish migraion, (Icduygu, 2009). Icduygu (2009) repors ha in fac 71% of labour migrans wen o Arab counries beween , and 98% beween and 60% beween , ye his figure sared o decline afer We need o ake ino accoun his imporan fac even if our daa include no only labour migrans, bu also sudens and visiors. Therefore, in his cion, we invesigae he desinaion-specific spillover effecs of emigraion on women s empowermen in Turkey. We expec emigraion o democraic counries, such as European and oher OECD counries, o have a greaer and a more posiive impac on he share of women in he Turkish parliamen han emigraion o he Arab and Russia and CIS counries. This is becau migrans in he Wern counries are expod o Wern poliical insiuions, Wern culure as well as Wern syms. This ransformaive experience can induce migrans o change heir poliical aiudes, beliefs and preferences and hence suppor women in poliics in Turkey. However, migraion o religious or auocraic or mi-auocraic counries such as Arab counries and Russia and CIS counries can have a weaker effec on women s parliamenary share in Turkey. Therefore, in his cion, we will accoun for he desinaion-specific effecs on women s empowermen in Turkey. The bounds esing procedure in he pren conex will be performed by relaing women s parliamenary share o emigraion o EU and OECD counries, emigraion o Arab counries, emigraion o Russia and he CIS counries and he relaive educaion of women 10 according o a vecor auoregressive (VAR) model of order p ha can be furher reduced o he following condiional error correcion model (ECM): 10 We could no include a measure of democracy in his ca, becau of he small sample size. 12

21 Migraion-Induced Women s Empowermen: The Ca of Turkey WP WP 0 1 EmigEUOECD 1 1 EmigARAB 2 1 EmigRussiaCIS 3 1 Educ 4 1 p i1 WP i i p i0 EmigEUOECD i i p i0 EmigARAB i i (4) p i0 EmigRussiaCIS i i p i0 Educ i i D WP, EmigEUOECD, EmigARAB, EmigRussiaCIS and Educ denoe, respecively: women s parliamenary share, he share of emigraion o he EU and OECD counries in oal emigraion, he share of emigraion o he Arab counries, he share of emigraion o Russia and he CIS counries in oal emigraion, and girls o boys raio in primary and condary schools. The lagged values of WP, EmigEUOECD, EmigARAB, EmigRussiaCIS and Educ (are in logs) pren a long-run relaionship in levels. and D are, respecively, he deerminisic erms; consan and dummy variables. The shorrun dynamics is reprened by lagged values of WP and curren and lagged values of EmigEUOECD, EmigARAB, EmigRussiaCIS and Educ. The condiional long-run elasiciies of women s parliamenary share wih respec o EmigEUOECD, EmigARAB, EmigRussiaCIS and Educ are given by 1 /0, 2 /0, 3 /0, and 4 /0 (Banerjee e al., 1998). The daa span 1960 o 1995, since he emigraion daa by desinaion counries are only available unil We sar by defining he lag order p for he Equaion (4). 11 Table (3) displays he resuls of he lag order p lecion procedure. The informaion crieria (Akaike, AIC, and Schwarz, SIC) and he Lagrange Muliplier saisic esing for remaining auocorrelaion up o he firs, fourh, and sixh orders in regression residuals are ud for he lag order lecion. AIC lecs p=3, whereas SIC lecs p=4. However, auocorrelaion ess deliver beer resuls for p=3, herefore p=3 is preferred. The las column of Table 3 repors he corresponding F-es saisics for he join hypohesis H 0 : using he finie-sample criical values repored in Narayan (2005) for T=30, corresponding o ca III wih unresriced consan and no linear deerminisic rend. As shown in Table 3 he null hypohesis of no long-run relaionship beween women s parliamenary share, emigraion share o he EU and oher OECD counries, emigraion share o he Arab counries, emigraion share o Russia and he CIS counries and he relaive educaion of women can be decisively rejeced for p=3 a he 1% significance level. For p=1, p=2, and p=4 he es saisic falls below he lower bound indicaing ha here is no long-run relaionship beween he variables of ineres a he 1% significance level. Table 3: Lag order lecion for he model wih emigraion daa by desinaion counries P AIC SIC AR(1) AR(4) AR(6) III FH 0 0: *** Noes: p is he lag order of he underlying VAR model for he condiional ECM of Equaion (4). AIC and SIC are, respecively, he Akaike and Schwarz Informaion Crieria. AR(1), AR(4), and AR(6) 11 Unforunaely, we could no es for higher orders of lag, due o he sample size. However, he lag order of 4 delivered very saisfacory resuls. 13

22 Şule Akkoyunlu are he p-values of he Lagrange Muliplier (LM) es saisics for esing for residual auocorrelaion of orders up o, respecively, one, four, and six. Bold enries repren he preferred lag order. III FH 0: is he F-es saisic for he null hypohesis H 0 : using he finie-sample criical values simulaed in Narayan (2005, p. 1988) for T=30 corresponding o ca III wih unresriced consan and no linear deerminisic rends. *** indicaes ha he null hypohesis can be rejeced a he 1% significance level. ++ indicaes ha he es saisic falls below he lower bound a he 1% significance level, herefore, he null hypohesis of no relaionship in levels canno be rejeced. Afer esablishing he exince of a long-run relaionship beween women s parliamenary share, emigraion share o he EU and OECD counries, he emigraion share o Arab counries, emigraion share o Russia and he CIS counries, and he relaive educaion of women, we esimae he coefficiens of ineres as follows: we sar wih an error-correcion model corresponding o p=3; and we delee he insignifican augmenaion lags; and arrive a he parsimonious model in Equaion (5) (sandard errors in parenhes, and error probabiliies in brackes). In he model below (5), D1993 denoes an impul dummy ha is equal o 1 in 1993 and zero oherwi for he oulier in WP WP 7.414EmigEUOECD 2.057EmigARAB (3.268) (0.308) (2.398) (76) 0.498EmigRussiaCIS 7.717Educ 0.260WP 44WP (0.121) (1.721) (0.199) (0.215) 0.279WP 5.647EmigEUOECD 3.593EmigEUOECD (0.181) (1.756) (1.322) 3.093EmigEUOECD 12EmigARAB 1.557EmigARAB 3 1 (1.284) (0.201) (0.402) 0.652EmigARAB 36EmigARAB 0.259EmigRussiaCIS 2 3 (0.287) (0.221) (0.143) 0.682EmigRussiaCIS 47EmigRussiaCIS 1 2 (0.239) (0.186) 0.477EmigRussiaCIS Educ Educ (0.186) (7.146) (8.157) Educ 0.422D (8.251) (0.232) R , F(23,8) = 4.559[0.016], T = 32 AR(1 1) ARCH (1) F 3.030[0.125], F 0.008[0.931] (1,7) (1,6) (5) 2Norm 2.112[0.348], (2) F RESET (1,7) 4.896[0.063] 12 An oulier in 1993 was deeced and he dummy in 1993 accouns for a decrea in women s parliamenary share in ha year ha canno be explained by he emigraion share o he EU and oher OECD counries, emigraion share o Arab counries, emigraion share o Russia and CIS counries, and he relaive educaion of women. 14