GRAVITY MODEL APPLICATION TO ESTIMATE ROMANIAN MIGRATION SIZE

Transcription

1 Knowledge horzons Volume 5, no. /013 GRAVITY MODEL APPLICATION TO ESTIMATE ROMANIAN MIGRATION SIZE Professor Marana BĂLAN, PhD. Insttute for Economc Forecastng-NIER, anan Academy, Emal: Professor Valentna VASILE, PhD. Insttute of Natonal Economy -NIER, anan Academy, Emal: Smona Mara STĂNESCU, PhD. Research Insttute for Qualty of Lfe, anan Academy, Emal: Abstract An mpressve number of labour force mgraton/moblty models, both at natonal and nternatonal level, are presented n the lterature. A large number of those ether descrbe nternatonal mgraton versus nternal mgraton or nterregonal mgraton n Europe or n other parts of the world. Some of the many methods and technques n use for the analyss and study of mgraton/labour force moblty are: the Markov chans method, the Harrs-Todaro model wth ts varants and extensons, models for the study of the economc mpact of labour moblty n enlarged Europe, models for the study of researchers moblty, models for the study of labour moblty through wage flexblty, models to characterze the mpact of labour moblty on macroeconomc ndcators. Ths paper, presents anan emgraton analyzed usng the gravty model. Keywords: mgraton, gravty model, mgraton statstcs JEL Classfcaton: J6 Introducton European Unon enlargement and labour movement lberalzaton towards EU countres substantally ncrease the possblty of mgraton to developed countres. Ths phenomenon, however, nfluences economc growth both n the short and long term. Accordngly, there s a need for both measures of mgraton control and the use of forecastng technques and models to estmate the economc mpact of ths phenomenon. Internatonal mgraton s a complex, multdmensonal phenomenon requrng for modelng and forecast to use analyss/methods developed by other dscplnes. A group of methods used n forecastng nternatonal mgraton s based on determnstc mathematcal models or on other approaches and technques, whch do not nvolve the uncertanty assocated wth the mgraton phenomenon. Unlke determnstc methods and models, stochastc (probablstc) nstruments for mgraton 74 Knowledge horzons Volume 5, no. /013

2 analyss and forecastng evolved from probablty theory. Although some of the models mentoned n the lterature relate to partcular theores of mgraton, such as, for example, econometrc models n economc theory, they do not try to buld, from a probablty pont of vew, a general theoretcal framework of populaton flows. Snce they present greater ablty to tackle uncertanty related to mgraton and forecastng, stochastc models are used more frequently than determnstc models. Based on an analogy wth physcs Newton's dea of gravty the gravty model used n economcs was developed by Lowry (1966) and then t was extended to nclude other economc varables n order to represent the effects of push and pull of mgrants (persons). Sometmes, the gravty model s used not only to descrbe geographcal moblty based on physcal dstance, but also to measure the smlartes between the pars of occupants n a matrx of ob changes. Ths paper presents, based on the gravtatonal model, estmates of mgraton flows out of ana. The results allow the forecast of the sze of anan mgraton. 1. Theoretcal background The soco-economc, demographc and cultural characterstcs of the mgrant populaton are the determnng factors of mgraton. Some of them operate also as ncentves or barrers, or change ther nfluence type and ntensty dependng on the tme horzon or geographcal area consdered. EU enlargement and, n the same tme, the free crculaton of labour to the EU have substantally ncreased the possblty of mgraton from ana to developed countres. The analyss of ths phenomenon n ana compared wth flows from other countres reveals the peculartes of anan mgrants. The analyss of statstcal data from the anan Natonal Statstcs Insttute shows the total number of anan emgrants durng for some destnaton countres: Germany wth people, Italy wth 1357, Canada wth 11016, USA wth and Span wth 566, followed by Austra, Hungary, Israel and France wth about 3500 people. The analyss of the sze and structure of mgraton flows usng varous methods allows at least ther short-term forecast. The measurement, analyss and forecast of anan mgraton flows should become a permanent task, manly because ther severe effects. The classcal mgraton analyss gravty model consders a flow from locaton to locaton to be proportonal wth the populaton of the orgn and destnaton countres and nversely proportonal to the dstance between the two locatons: M ( d ) = a P P f (1) where: a s a constant assocated wth overall moblty P and P represent the populaton n the countres of orgn, respectve of destnaton; f nverse functon of the dstance between ( ) d the two locatons. Snce on the ground d = d t results that net mgraton between the two locatons s zero. Therefore, the gravty model, n ths formulaton can be nterpreted as descrbng random flows that occur where there s no net mgraton. For the emprcal analyss case of geographc moblty, when net mgraton occurs, the gravty model s typcally extended wth a generalzed attracton factor so that regonal economc varables can be ncorporated nto the mgraton equaton as determnants (Molho, 1986): ( d ) A B M = a P P f () where: A s the "pushng" factor of the home country and B s the "attracton" factor of the host country. Both A and B are postve and are functons of explanatory varables for each locaton. To mantan consstency wth the orgnal equaton (1), both varables should be calbrated. In ther work "The Deregulaton of People Flows n Chna: Dd the Structure of Mgraton Change?", Bao Shumng, Orn B. Bodvarsson, Jack W. Hou and Zhao Yaohua developed several versons of the tradtonal gravty model for nterregonal mgraton appled for the study of nternatonal or nter-provncal mgraton. The unqueness of ths verson resdes n the ncluson of varables related to the economc, poltcal and socal aspects of the country/ provnce. As dependent varable s consdered the Knowledge horzons Volume 5, no. /

3 logarthm of gross emgraton rate ( log ( M )), calculated as the volume of emgraton from country to country dvded by total mgraton for country : logm = α α logd α lognetwork α logfdi α logfai 0 α (logfdi )(logfai ) α logy 5 α logminority α logwarm α logu α logmanemp α logurban α loge α loge α logu 7 ε (3) where: D = ral dstance (n km) between the captal of country to that of country ; NETWORK = sze of the mgrant communty resdng n country, who mgrated from country, measured as the rate of past mgraton flows; FDI = rato of real FDI per capta n country and actual foregn drect nvestment per capta n country ; FAI = rato of real fxed assets domestc nvestment per capta n country and real fxed assets domestc nvestment per capta n country ; Y = rato between real per capta ncome n country and real per capta ncome n country ; E, E = level of educaton n country and, respectvely ; U, U = unemployment rates durng the week precedng mplementaton of the census n country and respectvely ; MANEMP = rato between employment n the manufacturng sector share for country, and employment n the manufacturng sector share n country ; URBAN = rato between urban populaton share for country, and urban populaton share for country ; MINORITY = rato between the mnorty populaton share n country and the mnorty populaton share n country ; WARM = rato of the annual average temperature n the captal of country and the annual average temperature n the captal of country ; ε = error term Model assumptons: ral dstance and mgraton rate are assumed to be nversely proportonal; the rate of mgraton from country to country s postve n relaton to NETWORK varable; mgraton rate may be postvely or negatvely correlated wth any type of nvestment expendture n the country of destnaton compared to that n the orgn country; the rato between employment n ndustry share for the destnaton country and at home s ncluded n the model as an ndustry control mx; destnaton countres that are relatvely more urbanzed, have dfferent obs avalable, dfferent standards of lvng, etc. all wth nfluence on mgraton flows;. Study of anan emgraton development usng the gravty model For the study of anan emgraton, we used a smplfed gravtatonal model, because the data - provded by Eurostat statstcs, Natonal Statstcal Insttute of ana (Tempo database on-lne), or Statstcal Yearbooks of some EU countres where anan emgraton s sgnfcant are not complete for all exogenous varables used n the gravty model. As a soluton, the analyss of anan mgrant flows for was reported only for some EU countres and also for EU-7. Exogenous varables consdered n the model were: - GDP for country ; - Employment rate for country - Unemployment rate for country ; - Foregn drect nvestment for country ; - Number of anan emgrants n each country studed and for EU-7; - Dstance (on road) between ana and the varous destnatons of anan emgrants. The general form of the model used to analyze anan emgraton to some EU countres, USA and Canada wth the consdered varables s: Emgr = α α PIB α PIB 0 α Rocup 6 1 α dst 7 α Rsoma α Rsoma 3 α (log 8 FDI 4 ) ε α Rocup 5 (4) 76 Knowledge horzons Volume 5, no. /013

4 Estmates of anan emgraton to Span, the UK, Italy, Germany, USA, Canada, France, Austra, and Hungary have revealed a determnaton report ( R ) wth values above 75%, ndcatng that the exogenous varables consdered have a sgnfcant mpact on the dmenson of anan mgraton phenomena. Thus, for Span, the dependence degree of the endogenous varable of all the ndependent varables s 98.75%, for the UK 94.5%, for Italy 78%, for Germany 96%, for USA 93.5%, for Canada 89%, for France 86.3%, for Austra 97.5%, and for Hungary 9.4%. The lack of adequate statstcal databases lmt mgraton analyss and therefore, for some countres, the estmates were made wth fewer exogenous varables leadng to lower values of the coeffcent of determnaton (the case of Italy). Because the data sample has less than 30 observatons, we used the Student varable t ν, α, whose values were taken from the Student dstrbuton table (accordng to the value determned for α and the number of degrees of freedom, ν ). For the cases consdered, ts value s (wth a sgnfcance level of 5%). The analyss of the calculated values for each of the determned estmators ndcates that for all t calc > as they are sgnfcantly dfferent from zero. For the consdered model, for each estmator ts probablty ndcates that t can be accepted as unbased, consstent and effcent. Durbn-Watson statstc, whch s one of the most common procedures used to dentfy frstorder errors autocorrelaton n lnear regresson models, has values between.1 and.80, whch ndcates no autocorrelaton of errors. F test, measurng how well the ndependent varables explan the evoluton of the dependent varable, for each consdered regresson, has values below the F dstrbuton (Fsher-Snedecor) correspondng to the 9 degrees of freedom (39.55) and the sgnfcance level of 5%. Therefore, emgraton evoluton s stable over tme, and the model, wth all ts drawbacks, can be used for forecastng. Eurostat statstcs have enabled an estmate of the emgraton from ana to European Unon countres. The gravty model led to the estmators presented n table 1: Exogenous varables nfluence 91.5% the evoluton of the number of emgrants n the EU-7 for a sgnfcance level of 5% (Table 1). Comparson of Student varable t ν, α, whose values were taken from the Student dstrbuton table (accordng to the value determned for α and the number of degrees of freedom, ν for the consdered case, ts value s at a sgnfcance level of 5%), wth those calculated shows, for all estmators, that calculated values are hgher than Ths allows us to say that the model estmators are sgnfcantly dfferent from zero. Table 1 Dependent Varable: EMIGR Method: Least Squares Sample(adusted): Included observatons: 16 after adustng endponts EMIGR=C(1)C()*GDPEUC(3)*GDPROC(4)*RUNEMPUE C(5)* RUNEMPRO C(6)*REMPEU(-3) C(7)*REMPRO Coeffcent Std. Error t-statstc Prob. C(1) C() C(3) C(4) C(5) C(6) C(7) R-squared Mean dependent var Adusted R-squared S.D. dependent var S.E. of regresson Akake nfo crteron Sum squared resd Schwarz crteron Log lkelhood F-statstc Durbn-Watson stat.5607 Prob(F-statstc) Knowledge horzons Volume 5, no. /

5 Durbn-Watson statstc has the value.56 whch ndcates no autocorrelaton of errors (table 1). F test, that measures how well the ndependent varables explan the evoluton of the dependent varable, has a value of 6. (table 1), lower than the F dstrbuton (Fsher-Snedecor) correspondng to 7 degrees of freedom and the 5% sgnfcance level ( F 7 ;0,95 = ). Therefore, emgraton evoluton s stable over tme, and the model, wth all ts drawbacks, can be used for forecastng. Concluson Usng gravty model to estmate the number of emgrants n general and anans n partcular, can lead to sgnfcant results from an econometrc pont of vew. The varables consdered for these estmates descrbe wth a good percentage the varaton of the number of anan emgrants n each of the countres surveyed. However, these estmates can be substantally mproved f: ) the data would be avalable for a longer perod of tme; ) the nformaton about the number of anan emgrants would be more detaled; ) other exogenous varables, such as R&D&I costs would be ntroduced; v) qualtatve varables, whch very often have great mportance n the decson to mgrate, would be ntroduced; v) "educaton and tranng, etc. were consdered as exogenous varables BIBLIOGRAPHY 1. Barry R. Chswck, Edtor, (010), Hgh- Sklled Immgraton n a Global Labor Market, Lbrary of Congress Catalogng-n-Publcaton Data;. Coleman D., (004), Mgraton n the 1 st century: a thrd demographc transton n the makng? Plenary Address to the Brtsh Socety for Populaton Studes Annual Conference, Lecester, 13 September 004; 3. Koser K., (009), The Impact of Fnancal Crses on Internatonal Mgraton: Lessons Learned, IOM Mgraton Research Seres, No. 37, Geneva, 4. Molho, (1986), Theores of Mgraton: A Revew. Scottsh Journal of Poltcal Economy Volume 33, Issue 4, pages , November Ratha D., Mohapatra S., Slwal A., (009), Mgraton and Development Bref, Development Prospects Group, World Bank, www. steresources.worldbank.org; 6. Shumng B., Örn B. B., Hou J.W. ş Zhao Y., (0000), The Deregulaton of People Flows n Chna: Dd the Structure of Mgraton Change?, Presented n an AEA/CES sesson called Gender and Labor Moblty n the Chnese Labor Market at the ASSA conference n San Francsco on January 3, 009; 7. Srkec I.; Cohen J.H.; Ratha D., (01), Mgraton and Remttances durng the Global Fnancal Crss and Beyond, World Bank, p. 18, 8. Wdmaer S.; Dumont J.-C., (011), Are recent mmgrants dfferent? A new profle of mmgrants n the OECD based on DIOC 005/06, OECD Socal, Employment and Mgraton Workng Papers No. 16, Drectorate for Employment, Labour and Socal Affars, OECD Publshng, 9. *** Anuarul Statstc al âne , Insttutul Naţonal de Statstcă; 10. *** Internatonal Mgraton Outlook 01 - OECD 01, *** Internatonal Mgraton Outlook, SOPEMI 011, OECD, 1. *** Mgraton Statstcs 010, Annual Report, Offce for Natonal Statstcs, UK Statstcs Authorty, *** Statstcs Eurostat, *** World Mgraton Report Communcatng Effectvely about Mgraton, www. publcatons.om.nt 78 Knowledge horzons Volume 5, no. /013