The Determinants of Migration Flows in Europe

Transcription

1 59 The Determnants of Mgraton Flows n Europe Vol II(2) The Determnants of Mgraton Flows n Europe Ulrke Stens * Introducton The Treaty of Rome, whch was sgned n 1957, lad the foundaton for a European common market. The member states,.e. France, Germany, Italy and the Benelux countres, set themselves the target to create an economc area wth free movement of goods, servces and workers across borders (European Unon, 2012). Ths unon of states was expanded n the followng years, so that n 2012 the European Unon comprses 27 member states. However, the current European soveregn debt crss has questoned the constructon of the European Unon. A publc debate s underway on whch structural reforms are needed to ensure the coheson of the EU n the future. One fundamental problem s the lack n geographc labor moblty. The smultaneous presence of hgh unemployment rates and labor shortages ndcates an neffcent allocaton of resources. Bonn et al. (2008) argue that the economc gan of hgher labor moblty would overweght negatve externaltes such as potental downward pressure on wages. Thus, by mprovng the adjustment of labor market mbalances, labor moblty wthn the European Unon would have a pos- * Ulrke Stens receved her degree n Economcs (M. Sc.) from the Unversty of Bonn n The present artcle refers to her master thess under the supervson of Prof. Dr. Klaus F. Zmmermann, whch was submtted n August 2012.

2 60 The Determnants of Mgraton Flows n Europe Vol II(2) tve welfare effect n total. Although mgraton flows are too small, there are already people who move wthn the EU. The determnaton of forces that have drven these mgraton flows may gve ndcaton whch polces could effcently ncrease labor moblty. Ths work analyzes blateral mgraton wthn the European area from a macroeconomc perspectve. Beneath economc and demographc factors, such as ncome dfferences, the nfluence of past poltcal events s also analyzed. Ths accounts for the fact that the dentfcaton of mgraton determnants may serve as a bass for polces ntended on hgher labor moblty. For nstance, f the Maastrcht Treaty, whch eased free movement wthn the EU, affected mgraton flows n a postve way, ths would suggest that a change n nsttutonal settngs could be a way to ncrease moblty. It s focused on the mgraton flows between Germany and a panel of the remanng frst eleven member states (EU-11) for the followng reasons. Frst, defntons of mgraton flows dffer wthn the EU. By lmtng the analyss to Germany, data from one source can be used so that comparablty ssues are avoded. Second, Bonn et al. (2008) fnd that there s a great dsparty among EU countres regardng mgraton patterns so that poolng countres for the analyss on mgraton n Europe mght not be reasonable. The queston whch factors drve mgraton flows has been dscussed n varous emprcal studes. Those studes that use macro panel data can be broadly dstngushed nto two groups. On the one hand, there s a group of emprcal papers that study blateral mgraton flows wth underlyng data of both varous orgns and destnaton countres (Pedersen, Pytlkova, and Smth, 2008; Ortega and Per, 2009; Mayda, 2010). On the other hand, there are studes that examne the mgraton flows from and to one specfc country (Karemera, Oguledo, and Davs, 2000; Bertol and Moraga, 2011; Brücker, Slverstovs, and Trübswetter, 2003). One ssue that arses when blateral mgraton flows are analyzed s that potental

3 61 The Determnants of Mgraton Flows n Europe Vol II(2) mgrants have to choose between varous destnaton countres. Ths means the attractveness of other countres nfluences the mgraton decson. Several studes do not consder ths aspect n the emprcal specfcaton, whch may bas results. Based on Bertol and Moraga (2011), t s controlled for the presence of alternatve destnatons by applyng the so-called Augmented Mean Group estmator (Eberhardt and Bond, 2009). It ncludes an addtonal regressor durng the estmaton procedure whch measures the average dynamc evoluton of explanatory varables across countres. Ths estmaton method s also an approprate choce as t s robust to cross-sectonal dependence across countres and non-statonary varables, whch s lkely to be an ssue n ths sample (Brücker, Slverstovs, and Trübswetter, 2003). Determnants of Mgraton Most of the emprcal studes refer to a model of mgraton n whch an ndvdual maxmzes ts utlty by comparng the outcome when stayng n the source country wth the one n potental destnaton countres. Thereby, the net utlty of movng takes nto accounts the costs and rsks of mgraton. If the utlty of movng s greater than the one of stayng, the ndvdual decdes to mgrate. The sum of these ndvdual decsons causes mgraton flows from a sendng country to a destnaton country j (Clark, Hatton, and Wllamson, 2007). The emprcal model s based on Pedersen, Pytlkova, and Smth (2008). A mgraton flow from country to country j at tme t s modeled n the followng

4 62 The Determnants of Mgraton Flows n Europe Vol II(2) way: ln(m jt ) = β 1j ln(s jt 1 ) + β 2j ln(y t 1 ) + β 3j ln(u t 1 ) + β 4j ln(p t 1 ) }{{} P ushfactors + β 5j ln(y jt 1 ) + β 6j ln(u jt 1 ) +β 7j T jt + c j + u jt (1) }{{} P ullfactors where m j represents the mgraton flow from to j and s jt 1 stands for the stock of country s mmgrants already resdng n country j, whch accounts for network effects. In order to acheve comparablty across countres, m j and s jt 1 are expressed n relatve terms. Push and pull factors of the source and destnaton country are accounted for by real GDP per capta as a proxy for ncome (Y t 1 and Y jt 1 ) and the unemployment rate of the sendng and destnaton country (U t 1 and U jt 1 ). Besdes, the share of young people n the source country s populaton (P t 1 ) s taken as a demographc push factor. The explanatory varables are lagged one perod, whch accounts for the fact that potental mgrants need tme to gather nformaton before they decde to move to another country (Pedersen, Pytlkova, and Smth, 2008). In addton, ssues of reverse causalty are avoded. c j captures country fxed effects that are constant over tme and β 1j,..., β 4j are the country-specfc slope parameters. T jt s a lnear trend varable and u jt s an error term. Except for the lnear trend, all varables that vary over tme are taken n logs. By dong so, the coeffcents represent the partal elastctes of mgraton flows wth respect to the correspondng explanatory varable (see Wooldrdge, 2002, pp.15 18). The followng hypotheses regardng the explanatory varables are stated (see Table 1): β 1j, the coeffcent of s jt 1 n (1), should have a postve sgn. The more mmgrants already lve n a destnaton country, the hgher the mgraton nflows. The hgher the ncome n the destnaton country or the lower the n-

5 63 The Determnants of Mgraton Flows n Europe Vol II(2) come n the source country, the more lkely does an ndvdual decde to move. Hence, β 2j, the coeffcent of Y t 1 n (1), s assumed to have a negatve sgn, whereas β 5j, the coeffcent of Y jt 1, s expected to have a postve sgn. A hgh unemployment rate decreases the probablty of fndng a job, so that earnng opportuntes are lowered. As a result, the push factor U t 1 should have a postve coeffcent, whereas the correspondng pull factor U jt 1 should have a negatve one,.e. β 3j > 0 and β 6j < 0 (see Table 1). Clark, Hatton, and Wllamson (2007) argue that the ndvdual s utlty can be understood as the future ncome stream. They conclude that f an ndvdual expects a hgher ncome n the destnaton country, the utlty gan s the greatest for young ndvduals due to ther longer workng lfe. To put t dfferently, the hgher the share of young people n the populaton s, the more people decde to move. Hence, the effect of the push factor P t 1 should be postve,.e. β 4j > 0. As mentoned before, the emprcal analyss focuses on the mgraton flows from and to Germany. Thus, regardng mgraton nflows, the sample conssts of one sngle destnaton country, Germany, and multple source countres, the member states of the EU. On the contrary, there s a sngle country of orgn and multple destnaton countres when outflows from Germany to the EU member states are analyzed. Applyng ths to equaton (1), nduces the followng two equatons for nflows to and outflows from Germany, respectvely: ln(m t ) = β 1 ln(s t 1 ) + β 2 ln(y t 1 ) + β 3 ln(u t 1 ) + β 4 ln(p t 1 ) }{{} P ushfactors + β 5 ln(y Gt 1 ) + β 6 ln(u Gt 1 ) +β 7 T t + c + u t (2) }{{} P ullfactors

6 64 The Determnants of Mgraton Flows n Europe Vol II(2) and ln(m jt ) = β 1j ln(s jt 1 ) + β 2j ln(y Gt 1 ) + β 3j ln(u Gt 1 ) + β 4j ln(p Gt 1 ) }{{} P ushfactors + β 5j ln(y jt 1 ) + β 6j ln(u jt 1 ) +β 7j T jt + c j + u jt (3) }{{} P ullfactors In the followng, equatons (2) and (3) wll be refered to as the basc versons wth respect to nflows and outflows, respectvely. Beneath economc and demographc factors, t s also accounted for poltcal events whch mght have restraned or eased blateral mgraton between Germany and the European Unon. Ths s done by ncludng dummy varables, denoted as d, whch are zero untl ncludng the year of a poltcal event and one afterwards. Ths mples two assumptons: frst, a poltcal event wll mpact mgraton flows wth a one year delay as ndvduals need tme to gather nformaton before they decde to mgrate. Second, the event s effects wll be lastng. Beneath the drect mpact of poltcal events ther nteracton wth push and pull factors s measured. In partcular, t s analyzed whether a poltcal occason has strengthened or reduced the effects of these factors on mgraton. Followng Mayda (2010), ths nteracton s measured by multplyng the dummy varables wth push and pull factors, respectvely. Four mportant poltcal changes, concernng Germany and the EU-15, are consdered: The German Reunfcaton n 1990, the Maastrcht Treaty comng nto force n 1993, the Schengen Agreement n 1995 and 2000, respectvely, and the Introducton of the Euro n It s expected that the reunfcaton had negatve effects both on mgraton nflows drectly and ndrectly through a reducton of push and pull effects. The same effect s assumed to be true for the outflows. The Maastrcht Treaty, the Schengen Agreement and the Introducton of the Euro are

7 65 The Determnants of Mgraton Flows n Europe Vol II(2) expected to have postvely nfluenced mgraton flows between Germany and the EU-11. The correspondng dummy varables that account for a specfc event are denoted as d Reunfcaton, d Maastrcht, d Schengen and d Euro, respectvely. Table 2 lsts the dfferent poltcal varables that are added to the basc verson. For the sake of clarty, only the extensons regardng the basc verson of nflows are lsted. The basc verson of outflows s extended analogously. In total there are four modfcatons of the basc verson: frst, dummes of the poltcal events are added solely. In modfcatons two, three and four, the respectve push and pull nteracton terms of the GDP per capta, the unemployment rates and the share of the young populaton are added. Estmaton Strategy DESTATIS (2012a) and DESTATIS (2012b), the German Federal Statstcal Offce, provdes tmes seres on mgraton flows between Germany and the EU-11 over the perod from 1986 untl 2011 so that there are 286 observatons n total. The economc explanatory varables,.e. real GDP per capta and the unemployment rate, are taken from OECD (2012a) and OECD (2012b). The real GDP per capta s GDP per capta n US-dollars adjusted for purchasng power partes and at constant prces wth base year 2005 so that t s comparable both across countres and across tme. Up to 1990, all used data refers to West Germany. Afterwards, the unfed Germany s consdered. One problem wth the data s that the natonaltes of mgrants are not dstngushed. Everyone who moves to and from Germany s accounted as an mmgrant and emgrant, rrespectvely of the ctzenshp. In 2011, 20% of the nflows to Germany from the EU-11 were people of German natonalty. At the same tme, only 31% of the outflows from Germany to a state of the EU-11 were German ctzens. These numbers suggest

8 66 The Determnants of Mgraton Flows n Europe Vol II(2) that the data on mgraton flows ncludes a substantal amount of people wth a natonalty dfferent to that of the sendng country. Ths mght nfluence the coeffcent estmates of the stock of mmgrants resdng n the destnaton country. The data structure s lkely to be both autorcorrelated due to the long tme perod and cross-sectonal dependent. The latter ssue may arse due to the small sample of EU countres whch have close economc tes. In addton, non-statonarty of the macroeconomc varables mght be a problem. Usng Monte Carlo smulatons,?? fnd that so called Common Correlated Effects mean group estmator (CCE) developed by Pesaran (2006) s robust to non-statonarty and spatal correlatons. Besdes, Bertol and Moraga (2011) show that the estmaton procedure of the CCE can account for multlateral resstance to mgraton. The CCE assumes heterogeneous slope parameters across unts. In addton, both the explanatory varables and the error term of the unt specfc regresson equatons are assumed to be nfluenced by unobservable common factors, whereby the mpact of these factors dffers across unts. To control for the unobservable common factors,.e. for the spatal dependence, panel averages of the explanatory and dependent varables are ncluded as addtonal ndependent varables n the regresson equaton. Ths equaton s estmated for each unt separately. Afterwards, the ndvdual coeffcent estmates are averaged over all panel unts. However, one ssue of the CCE s that the number of estmated parameters s rather hgh whch means that the results may become mprecse n small samples. Eberhardt and Bond (2009) developed an estmator, the Augmented Mean Group estmator (AMG), whch assumes the same econometrc model as the CCE. Ther Monte Carlo smulatons ndcate that the AMG performs equally well as the CCE n the presence of cross-sectonal dependence and non-statonarty. The estmaton procedure dffers, though. Frst, the regresson equatons are set up n frst dfferences and year dummes are ncluded. At that stage, the observatons are pooled,.e. slope

9 67 The Determnants of Mgraton Flows n Europe Vol II(2) parameters are assumed to be homogeneous across unts. The estmated coeffcents of the year dummes represent a common dynamc process of the varables. In a second step, ths common dynamc process s ncluded as an explct varable n the orgnally regresson equaton. Agan, the regresson equaton s estmated for each unt ndvdually and the coeffcents averaged across unts. The AMG estmator has the advantage that just one varable, the common dynamc process, s added to the regresson equaton. Hence, for the estmaton of mgraton determnants ths latter estmator s employed nstead of the CCE. In order to analyze f and how the "common dynamc process" of the AMG nfluences estmaton results, the estmates of the AMG are compared to those of OLS wth Newey-West standard errors. Results Table 3 n the appendx dsplays results of both the averaged and ndvdual AMG coeffcents plus the estmates of the OLS regresson wth Newey-West standard errors. The common dynamc process that s ncluded n the regresson equaton durng the AMG estmaton procedure seems to mpact mgraton flows substantally: ts parameter estmates are sgnfcant on the ndvdual level for the majorty of countres. Comparng the AMG and OLS outcomes shows a consderable dfference wth regard to coeffcent estmates. Ths s partcularly true for those countres whose coeffcent of the common dynamc process s sgnfcant. The consderable nfluence of the common dynamc process on estmaton results underpns the arguments that estmaton results become ndeed based f an estmator s used that s not robust to ssues of cross-sectonal dependence and non-statonarty. Hence, the followng analyss concentrates on the estmaton results of the AMG. The stock of mmgrants, whch accounts for network effects, s ether nsgnf-

10 68 The Determnants of Mgraton Flows n Europe Vol II(2) cant or even seems to have a negatve effect on nflows to Germany regardng the majorty of countres. These results may be due to the data structure, as the natonaltes of mgrants are not dstngushed. The push effect of ncome,.e. GDP per capta, s sgnfcant for the majorty of countres. As predcted by theory the coeffcents are negatve. The pull effect of the German GDP per capta only has a sgnfcant nfluence on nflows from a few countres. However, the respectve parameter estmates are all postve. Ths suggests that a hgh ncome n the destnaton country ndeed pulls mgrants. When the ndvdual parameter estmates are averaged, the push and pull effect of sendng and destnaton country s GDP per capta, respectvely, are stll sgnfcant. Ths ndcates that nflows from the EU-11 react smlarly to changes n ncome. The unemployment rate does not seem to nfluence mgraton flows to Germany. Ths result supports the pont dscussed above that labor s not effcently adjusted wthn the European Area. The share of young people affects mgraton nflows only wth respect to a few countres. Beneath economc and demographc factors, the nfluence of poltcal varables s analyzed. Thereby, both the drect effects of poltcal events and the ndrect effects va a change n push and pull effects are evaluated (see Table 4 and 5). However, the results show that none of the poltcal events seem to have clearly changed mgraton patterns regardng nflows to Germany from the EU-11. The drect effect of poltcal events measured va the ncluson of dummy varables s close to zero. In addton, the ndrect effect measured by the nteracton varables does also not allow for an unambguous nterpretaton, ether. Tables 6 to 8 dsplay estmates regardng outflows. The comparson between estmaton results of outflows wth those of nflows reveals consderable dfferences regardng the forces that drve these flows. Frst, the push and pull effects of GDP per capta that are predcted by theory are clearly dentfed for nflows to Germany.

11 69 The Determnants of Mgraton Flows n Europe Vol II(2) However, ths does not apply to outflows. The coeffcents of GDP have dfferent sgns across countres. Estmaton results wth respect to unemployment rates and the share of young people are smlar for n- and outflows. The effects of unemployment rates are rather low. The parameters wth respect to the share of young people are sgnfcant for several countres, ther drectons dffer though. The man dfference wth respect to poltcal events s the nfluence of GDP per capta nteracton varables. There was no clear effect for the nflows, whle estmaton results of outflows suggest that the Maastrcht Treaty and, partcularly, the Schengen Agreement strengthened push and pull effects assumed by theory. Agan, the nterpretatons regardng the nfluence of poltcal events have to be seen wth cauton, as t s not controlled for events n the EU-11 countres. Concluson Wthn the European Unon labor moblty s too low, whch nduces an neffcent allocaton of resources. Partcularly wth regard to the demographc change and the resultng shortage of labor and sklls, the mprovement of labor adjustment s an mportant ssue. Ths thess amed at the determnaton of forces of prevous mgraton flows n order to dentfy potentals to ncrease moblty of European ctzens. To obtan consstent results, the AMG estmator was appled to the emprcal specfcaton of mgraton flows. The comparson between OLS and AMG estmaton results ndcated substantal dfferences among estmaton results, whch confrmed that estmaton results are ndeed based f the ssues dscussed above are not consdered. In addton, the results affrmed that the adjustment of labor does not work well wthn the European area. Both mgraton nflows to Germany and outflows from Germany were not drven by the evoluton of unemployment rates. In contrast, results of nflows to Germany suggest that ncomes of the sendng and destnaton country do seem to nfluence mgraton

12 70 The Determnants of Mgraton Flows n Europe Vol II(2) flows. Ths s n lne wth the fndngs of Bonn et al. (2008, p. 8) that the prospect of a hgher ncome s a key factor of the mgraton decson. There s no clear evdence that ether the Maastrcht Treaty or the Schengen Agreement of the Introducton of the Euro has strengthened moblty. One aspect that has been left out n the analyss s the nfluence of educaton on mgraton flows. The country s level of educaton could not be ncluded as there was no consstent data source. However, partcularly wth regard to the mmnent skll shortage n the European Unon t would be nterestng to see whether a hgh level of educaton nduces hgh moblty of people and consequently a more effectve allocaton of labor. A further ntrgung topc would be the queston f and how the European debt crss has changed labor moblty wthn the EU. Several member states have to make massve cuts whch also concern to socal welfare systems. Ths mght ncrease the pressure to mgrate to another member state of the EU f t offers better employment opportuntes.

13 71 The Determnants of Mgraton Flows n Europe Vol II(2) References Bertol, S., and J. F.-H. Moraga (2011): Multlateral Resstance to Mgraton, IZA Dscusson Papers 5958, Insttute for the Study of Labor (IZA). Bonn, H., W. Echhorst, C. Florman, M. O. Hansen, L. Sköld, J. Stuhler, K. Tatsramos, H. Thomasen, and K. F. Zmmermann (2008): Report No. 19: Geographc Moblty n the European Unon: Optmsng ts Economc and Socal Benefts, IZA Research Reports 19, Insttute for the Study of Labor (IZA). Brücker, H., B. Slverstovs, and P. Trübswetter (2003): Internatonal Mgraton to Germany: Estmaton of a Tme-Seres Model and Inference n Panel Contegraton, Dscusson Papers of DIW Berln 391, DIW Berln, German Insttute for Economc Research. Clark, X., T. J. Hatton, and J. G. Wllamson (2007): Explanng U.S. Immgraton, , The Revew of Economcs and Statstcs, 89(2), Coakley, J., A.-M. Fuertes, and R. P. Smth (2006): Unobserved heterogenety n panel tme seres models, Computatonal Statstcs & Data Analyss, 50(9), DESTATIS (2012a): Bevölkerung und Erwerbstätgket. Vorläufge Wanderungsergebnsse. 2011, de/de/publkatonen/thematsch/bevoelkerung/wanderungen/ vorlaeufgewanderungen pdf? blob=publcatonfle, Accessed: August 21, (2012b): Wanderungen zwschen Deutschland und dem Ausland: Jahre, Staaten der Europäschen Unon, de/geness/onlne/logon, Accessed: August 21, Eberhardt, M., and S. Bond (2009): Cross-secton dependence n nonstatonary panel models: a novel estmator, MPRA Paper 17692, Unversty Lbrary of Munch, Germany. Karemera, D., V. I. Oguledo, and B. Davs (2000): A gravty model analyss of nternatonal mgraton to North Amerca, Appled Economcs, 32(13), Mayda, A. (2010): Internatonal mgraton: a panel data analyss of the determnants of blateral flows, Journal of Populaton Economcs, 23(4), OECD (2012a): Gross Domestc Product, aspx?queryd=556#, Accessed: August 21, 2012.

14 72 The Determnants of Mgraton Flows n Europe Vol II(2) (2012b): Labour Force Statstcs by Sex and Age, oecd.org/index.aspx?datasetcode=lfs_sexage_i_r, Accessed: August 21, Ortega, F., and G. Per (2009): The Causes and Effects of Internatonal Mgratons: Evdence from OECD Countres , Workng Paper 14833, Natonal Bureau of Economc Research. Pedersen, P. J., M. Pytlkova, and N. Smth (2008): Selecton and network effects Mgraton flows nto OECD countres , European Economc Revew, 52(7), Pesaran, M. H. (2006): Estmaton and Inference n Large Heterogeneous Panels wth a Multfactor Error Structure, Econometrca, 74(4), Wooldrdge, J. (2002): Data. Mt Press. Econometrc Analyss of Cross Secton and Panel

15 73 The Determnants of Mgraton Flows n Europe Vol II(2) Appendx Explanatory varables s j Push Pull Y U P Y j U j Expected effect Table 1: Expected effects of explanatory varables on mgraton flows Modfcaton Dummy Push factors Pull factors d Event d Event represents d Reunfcaton d Event ln(y t 1 ) d Event ln(u t 1 ) d Event ln(p t 1 ) d Event ln(y Gt 1 ) d Event ln(y Gt 1 ), d Maastrcht, d Schengen and d Euro, respectvely Table 2: Poltcal varables added to the basc verson of nflows

16 74 The Determnants of Mgraton Flows n Europe Vol II(2) Countres Independent varables ln(s t 1) ln(y t 1) ln(u t 1) ln(p t 1) ln(y Gt 1) ln(u Gt 1) Common dynamc process EU-11 AMG 0.17 (0.19) 2.14 (0.87) 0.15 (0.12) 0.34 (0.86) 1.34 (0.76) 0.02 (0.14) 0.85 (0.28) BE OLS (0.28) (0.51) (0.05) (0.48) (0.57) (0.06) AMG 0.44 (0.29) 1.24 (0.58) 0.15 (0.09) 0.55 (0.68) 0.12 (0.79) 0.23 (0.08) 0.05 (0.10) DK OLS (1.81) (0.21) (6.07) (0.59) (1.55) (0.33) AMG 0.37 (0.43) 8.88 (3.87) 0.52 (0.35) 2.13 (1.43) 7.64 (2.5) 0.65 (0.32) 0.73 (0.43) FR OLS 0.5 (1.16) (1.47) (0.17) (2.33) (1.94) (0.25) AMG 0.63 (1.00) 3.64 (2.04) 0.13 (0.63) 2.77 (3.86) 0.81 (3.32) 0.33 (0.32) 0.84 (0.64) GR OLS (1.08) (0.84) (0.49) (1.87) (1.96) (0.21) AMG 2.69 (0.79) 3.03 (1.19) 0.49 (0.35) 3.66 (1.34) 1.64 (2.38) 1.13 (0.19) 0.94 (0.44) IR OLS (0.17) (1.67) (0.30) (1.88) (1.37) (0.35) AMG 0.15 (0.19) 0.4 (1.10) 0.30 (0.23) 0.52 (2.04) 3.26 (1.43) 0.04 (0.27) IT OLS (0.78) (1.38) (0.34) (1.48) (4.34) (0.17) AMG 0.74 (0.47) 5.95 (0.87) 0.37 (0.24) 3.62 (1.28) 0.16 (1.65) 0.31 (0.12) 2.06 (0.33) LU OLS (0.28) (0.49) (0.06) (0.56) (1.07) (0.17) AMG 0.01 (0.28) 0.56 (0.49) 0.03 (0.09) 3.62 (0.62) 2.52 (1.29) 0.42 (0.17) 0.08 (0.17) NL OLS (0.32) (0.82) (0.10) (0.53) (0.72) (0.09) AMG 0.76 (0.37) 2.05 (1.24) 0.06 (0.12) 0.73 (0.62) 4.04 (1.25) 0.30 (0.18) 0.05 (0.20) PT OLS (1.24) (4.27) (0.57) (6.54) (3.28) (0.37) AMG 0.16 (1.54) 2.13 (2.71) 0.76 (0.49) (8.15) 1.61 (3.94) 0.21 (0.57) 2.94 (1.01) SP OLS (0.25) (2.12) (0.30) (1.26) (0.77) (0.16) AMG 1.04 (0.35) 1.51 (1.83) 0.39 (0.26) 1.2 (1.39) 0.55 (1.14) 0.07 (0.16) 0.48 (0.28) UK OLS (0.22) (0.69) (0.17) (0.31) (0.63) (0.09) AMG 0.02 (0.29) 2.41 (0.63) 0.15 (0.13) 0.42 (0.53) 0.94 (0.71) 0.03 (0.08) 0.71 (0.14) Dependent varable: ln(m t ). Addtonal controls: T t, c. Common dynamc process ncluded as addtonal ndependent varable durng AMG estmaton procedure. Estmates of EU-11 are the averages of country specfc coeffcents. Numbers refer to the perod Standard errors are n brackets. For the OLS regresson Newey-West standard errors are computed. * sgnfcant at 10%, ** sgnfcant at 5%, *** sgnfcant at 1%. Table 3: Influence of economc and demographc factors on nflows to Germany (results of OLS and AMG estmaton)

17 75 The Determnants of Mgraton Flows n Europe Vol II(2) Countres d Reunfcaton EU (0.07) BE 0.02 (0.05) FR 0.54 (0.21) GR 0.08 (0.17) IR 0.28 (0.12) IT 0.14 (0.09) LU 0.08 (0.09) NL 0.09 (0.11) PT 0.31 (0.26) SP 0.14 (0.07) Independent varables d Maastrcht 0.06 (0.07) 0.04 (0.05) 0.17 (0.31) 0.29 (0.18) 0.05 (0.14) 0.34 (0.13) 0.07 (0.09) 0.17 (0.11) 0.36 (0.32) 0.12 (0.08) d Schengen d Euro 0.05 (0.03) 0.00 (0.04) (0.04) (0.04) (0.18) (0.23) (0.12) (0.16) (0.13) (0.07) (0.09) (0.07) (0.08) (0.10) (0.08) (0.23) (0.21) (0.05) (0.06) Dependent varable: ln(m t ). Addtonal controls: ln(s t 1 ), ln(y t 1 ), ln(u t 1 ), ln(p t 1 ), ln(y Gt 1 ), ln(u Gt 1 ), T t, c. Estmates of EU-11 are the averages of country specfc coeffcents. Numbers refer to the perod Standard errors are n brackets. * sgnfcant at 10%, ** sgnfcant at 5%, *** sgnfcant at 1%. Table 4: Influence of poltcal events on nflows to Germany (results of AMG estmaton)

18 76 The Determnants of Mgraton Flows n Europe Vol II(2) Countres ln(y t 1 ) Independent varables ln(y t 1 ) d Reunfcaton EU (2.51) 2.62 (4.42) DK (3.82) (6.79) GR (2.88) 5.95 IT (5.47) (6.02) NL (12.74) (13.63) PT (2.35) (40.18) SP (2.55) (5.22) UK (1.20) (1.42) d Maastrcht d Schengen d Euro 2.23 (1.71) 0.80 (2.79) 0.46 (1.2) (8.27) (4.24) (18.19) (20.51) (3.29) (11.33) (8.65) (47.02) (45.3) (3.24) (50.17) (15.83) (4.92) 0.00 (-) (4.66) (1.37) (1.48) Countres ln(y Gt 1 ) EU (1.72) DK 1.85 (3.61) GR (3.64) IT (5.41) NL 0.34 (12.93) PT 6.67 (3.69) SP 2.29 (2.37) UK 2.34 (1.11) Independent varables ln(y Gt 1 ) d Reunfcaton d Maastrcht 2.49 (4.4) 2.24 (1.7) (6.78) (8.23) (5.72) (17.38) (5.98) (0.01) (13.66) (47.06) (38.31) (47.88) (5.08) (0.01) (1.4) (1.44) d Schengen d Euro 0.72 (2.87) 0.45 (1.18) (4.26) (19.58) (3.16) (11.22) (8.57) (45.35) (3.27) (15.08) (4.75) (4.52) (1.34) - - Dependent varable: ln(m t ). Addtonal controls: ln(s t 1 ), ln(u t 1 ), ln(p t 1 ), ln(u Gt 1 ), T t, c. Estmates of EU-11 are the averages of country specfc coeffcents. Numbers refer to the perod Standard errors are n brackets. * sgnfcant at 10%, ** sgnfcant at 5%, *** sgnfcant at 1%. Table 5: Influence of poltcal events on nflows push and pull effects of GDP per capta (results of AMG estmaton)

19 77 The Determnants of Mgraton Flows n Europe Vol II(2) Countres Independent varables ln(y Gt 1 ) ln(u Gt 1 ) ln(p Gt 1 ) ln(y t 1 ) ln(u t 1 ) Common dynamc process EU (0.87) 0.08 (0.06) 0.13 (0.51) 0.20 (0.61) 0.01 (0.09) 0.62 (0.15) BE 0.23 (0.97) 0.00 (0.09) 0.47 (0.21) 0.12 (0.69) 0.15 (0.12) 0.68 (0.12) DK 5.20 (2.21) 0.88 (0.34) 2.83 (0.63) 3.67 (2.81) 0.37 (0.30) 1.94 (0.33) FR 0.16 (0.64) 0.12 (0.06) 0.56 (0.26) 0.39 (0.52) 0.21 (0.12) 0.62 (0.09) GR 0.25 (1.10) 0.11 (0.16) 1.35 (0.71) 2.49 (0.70) 0.34 (0.23) 0.70 (0.23) IR 2.38 (2.17) 0.38 (0.30) 0.15 (0.91) 1.55 (1.75) 0.45 (0.37) 1.81 (0.39) IT 3.53 (1.88) 0.17 (0.22) 0.67 (1.19) 1.96 (0.78) 0.05 (0.41) 0.23 (0.30) LU 2.30 (1.18) 0.02 (0.13) 1.78 (0.39) 0.95 (0.43) 0.02 (0.09) 0.56 (0.17) NL 2.80 (0.83) 0.22 (0.10) 0.55 (0.29) 2.22 (0.78) 0.03 (0.08) 0.77 (0.14) PT 3.90 (1.83) 0.16 (0.16) 5.34 (0.69) 0.23 (1.23) 0.22 (0.24) 1.87 (0.33) SP (1.48) UK 0.54 (0.47) (0.19) 0.23 (0.06) (0.86) 0.62 (0.18) (1.51) 0.63 (0.51) (0.26) 0.01 (0.10) (0.19) 0.90 (0.09) Dependent varable: ln(m jt ). Addtonal controls: T jt, c j. Common dynamc process ncluded as addtonal ndependent varable durng AMG estmaton procedure. Estmates of EU-11 are the averages of country specfc coeffcents. Numbers refer to the perod Standard errors are n brackets. * sgnfcant at 10%, ** sgnfcant at 5%, *** sgnfcant at 1%. Table 6: Influence of economc and demographc factors on outflows from Germany (results AMG estmaton)

20 78 The Determnants of Mgraton Flows n Europe Vol II(2) Countres d Reunfcaton EU (0.04) GR 0.08 (0.08) IR 0.36 (0.19) IT 0.00 (0.10) NL 0.04 (0.06) PT 0.21 (0.11) SP 0.14 (0.08) Independent varables d Maastrcht 0.02 (0.03) 0.02 (0.12) 0.03 (0.28) 0.14 (0.13) 0.18 (0.08) 0.37 (0.14) 0.07 (0.10) d Schengen d Euro 0.00 (0.04) 0.04 (0.04) (0.06) (0.09) (0.24) (0.07) (0.10) (0.06) (0.06) (0.09) (0.12) (0.07) (0.08) Dependent varable: ln(m jt ). Addtonal controls: ln(y Gt 1 ), ln(u Gt 1 ), ln(p Gt 1 ), ln(y jt 1 ), ln(u jt 1 ), T jt, c j. Estmates of EU-11 are the averages of country specfc coeffcents. Numbers refer to the perod Standard errors are n brackets. * sgnfcant at 10%, ** sgnfcant at 5%, *** sgnfcant at 1%. Table 7: Influence of poltcal events on outflows from Germany (results of AMG estmaton)

21 79 The Determnants of Mgraton Flows n Europe Vol II(2) Countres ln(y Gt 1 ) EU (1.42) FR 1.25 (3.21) IR 4.98 (5.16) IT 8.43 (3.27) NL (5.82) PT 2.71 (2.08) SP 0.05 (1.56) UK 0.96 (0.95) Countres ln(y t 1 ) EU (0.57) FR 1.01 (3.09) IR 1.48 (3.11) IT 6.97 (2.84) NL (5.72) PT 0.16 (1.26) SP 0.06 (1.08) UK 0.61 (0.74) Independent varables ln(y Gt 1 ) d Reunfcaton 3.49 (2.36) 1.02 (3.99) (9.08) 5.56 (4.82) (5.80) (23.22) 5.77 (3.87) 3.44 (1.27) d Maastrcht 3.09 (2.27) 1.61 (12.50) (7.70) (9.74) (24.12) 5.20 (28.19) 0.02 (0.01) 2.43 (1.08) d Schengen Independent varables ln(y t 1 ) d Reunfcaton 3.55 (2.38) 0.98 (4.01) (9.38) 5.63 (4.85) (5.79) (24.36) 5.91 (3.97) 3.43 (1.28) d Maastrcht d Euro 6.28 (2.00) 0.10 (0.99) (13.59) (2.36) (1.97) (0.02) (7.87) (23.70) (1.70) (7.49) (2.64) (3.68) (1.30) - - d Schengen d Euro 2.95 (2.39) 6.45 (2.08) 0.11 (1.00) (12.62) (13.72) (2.38) (7.96) (1.96) (9.86) (7.94) (24.10) (23.67) (1.68) (29.54) (7.86) (2.74) 0 (-) (3.79) (1.33) (1.11) Dependent varable: ln(m jt ). Addtonal controls: ln(u Gt 1 ), ln(p Gt 1 ), ln(u jt 1 ), T jt, c j. Estmates of EU-11 are the averages of country specfc coeffcents. Numbers refer to the perod Standard errors are n brackets. * sgnfcant at 10%, ** sgnfcant at 5%, *** sgnfcant at 1%. Table 8: Influence of poltcal events on outflows push and pull effects of GDP per capta (results of AMG estmaton)