8Oct04Conferenceresi mERZAN

GROWTH AND MIGRATION SCENARIOS TURKEY-EU: 

GROWTH AND MIGRATION SCENARIOS TURKEY-EU Refik Erzan, Umut Kuzubaş, Nilüfer Yıldız Boğaziçi University

Slide2: 

Potential migration from Turkey to the EU as a result of EU membership and removal of restrictions on labor mobility Turkish migration to the EU between 2004-2030 with different methods under various scenarios

Slide3: 

Highly Speculative Scenarios Up to 25% of Turkish population!!! EU Commission 2004 impact study Between 0.5 million to 4.4 million until year 2030

Slide4: 

Hacettepe-NIDI (1996) Based on surveys carried out by 1800 households Economic motives are the major determinants of migration Migrant profile: 25-30 years of age, need for employment, primary school education 85% of migrants have a network before departure Most non-migrants do not intend to migrate abroad Hubert Krieger (2004) Analyzes the potential to migrate and the heterogeneity of migrants Based on Eurobarometer survey, spring 2002 Multivariate, logistic regression of odds ratios Profile: 25-39 age group, student, unemployed, medium and higher education U-curve effect of income, opposite U-curve effect of deprivation Survey (Opinion Poll) Based Studies

Slide5: 

Anna Maria Mayda (2003) Investigates migration into 14 OECD countries between 1980 and 1996 Runs cross-country, time series and panel regressions Pull factors significantly increase emigration rates Push factors have a smaller effect Distance is the most important cost factor that affects emigration rates Alecke, Huber, Untiedt (2001) Estimates East-West migration in Germany and migration within the EU. Country specific fixed effects are included in the model Uses panel data analysis Results: Income differentials drive migration flow within German states Country specific effects are important in explaining differences in the level of migration between states. Econometric Estimations

Econometric Forecasts: 

Econometric Forecasts Hille, Straubhaar (2001) Pooled time series of bilateral migration from Greece, Portugal and Spain to seven EU member countries Predicts that 0.27- 0.34% of the CEEC population p.a. (270,000 - 340,000 migrants) will move to the EU This accrues to 0.07- 0.09% of the EU population Michael Fertig (2000) Estimates migration from CEEC-10 to Germany Uses maximum likelihood method Data includes net migration rates from 17 source countries to Germany between 1960-1994 In the baseline scenario with free movement, the total change in migrant stock is 1.4 million by 2015 Free movement effect turns out to be small

Econometric Forecasts: 

Econometric Forecasts Boeri, Brücker (2001) Estimates potential migration from CEEC to the EU-15 Applies SUR to an error correction model Data comprises of foreign population figures in Germany of 19 source countries between 1967-2001 Predicts 3.8 million CEEC nationals in the EU-15 by 2030, assuming immediate free movement by an out-of-sample forecast A net increase of CEEC migrants by 2.7 million in 28 years Brücker, Alvarez-Plata, Siliverstovs (2003) (European Commission Report) Update of Boeri-Brücker (2001) Estimates potential migration from CEEC to the EU-15 Two different samples are used Applies and compares various estimation methods on a dynamic panel. SUR estimator outperforms in the sample with large time dimension Predicts a net increase of 2.5 million in the migrant stock, implying 3.7 million migrants in 2030

Econometric Forecasts and Simulations: 

Econometric Forecasts and Simulations Harry Flam (2003) Estimates potential migration from Turkey Uses Boeri Brücker (2000) error correction model Predicts 3.5 million Turkish population in Germany by 2030, assuming no restrictions on migration A net increase of Turkish migrants by 1.2 million in 30 years Lejour, de Mooij, Capel (2004) Based on de Mooij and Tang (2003) migration elasticities Estimates long term migration of 2.7 million from Turkey to EU-15 This equals 4% of current Turkish population and 0.7% of the EU’s

Slide9: 

Following the European Commission Report Brücker, Alvarez-Plata, Siliverstovs (2003) mfht = αh + β1mfh,t-1 + β2mfh,t-2 + β3 ln(wft/wht) + β4 ln(wht) + β5ln(eft) + β6 ln(eht) + ufht (1) mfht: The share of migrants from country h residing in country f (Germany) as per cent of home population w: Wage (proxied by GDP-PPP per capita) e: Employment rate (1-unemployment rate) h, f,t: Home, foreign countries and year, respectively. Methodology Adopted in this Study

Data: 

Home income captures the pecuniary costs of migrating Employment rate captures the probability of finding a job Income differential captures gains from migrating Data

Data Sources 1967-2001: 

Population : World Development Indicators (2003) Migrant Stock : German Federal Statistical Office* GDP per capita**: GDP-PPP per capita (1990 international Dollars) from Maddison (2002) Employment Rate: OECD Economic Outlook in OECD databases * 1987 1988 1989 figures are adjusted by net immigration to Germany ** Luxembourg, Switzerland from Groningen Growth and Development Data Sources 1967-2001

Slide12: 

Stock of foreign population of 19 source countries (Austria, Belgium, Denmark, Finland, France, Greece, Iceland, Ireland, Italy, Luxembourg, the Netherlands, Norway, Portugal, Spain, Sweden, Switzerland, Turkey, United Kingdom, former Yugoslavia) in Germany Germany is the host (foreign) country Long time-series available Germany has the largest share of migrants in the EU

Naturalization of Turkish Migrants in Germany: 

Naturalization of Turkish Migrants in Germany

Slide14: 

Estimation method : Seemingly Unrelated Regression (SUR) or Equations (SURE) The European Commission Report (2003) has shown that for data sets having large time dimension and small cros-section SUR estimator outperforms the others (ie GMM, OLS, WITHIN) Two lags of the dependent variable is included Slope Coefficients are restricted to be equal for all countries Intercepts (constant terms) are estimated seperately for each country to cover the country specific factors GUEST and FREE dummies are also included to capture the periods of guest worker agreements and free labor mobility WAR (former Yugoslavia), INTERVENTION and INSURGENCY (Turkey) dummies are also included

Slide15: 

Regression Results

Test Statistics: 

Test Statistics LM test statistic 950.78 (H0:Homoskedasticity vs. H1: Heteroskedasticity) LR test statistic 773.82 (H0:Heteroskedasticity vs. H1: Heteroskedasticity + Countrywise correlation)

Actual vs. Fitted (SUR estimation) Turkish Migrant Stock in Germany (in millions): 

Actual vs. Fitted (SUR estimation) Turkish Migrant Stock in Germany (in millions)

Forecast Scenarios Based on SUR: 

Forecast Scenarios Based on SUR Assumptions for Germany : GDP per capita : Assumed to grow %2 per an Employment rate : 1991-2001 average is taken as constant throughout the projection interval Assumptions for Turkey : Population : Natural rate of growth projections based on World Bank (2002) (ie zero net migration is assumed) Then, migration adjusted population calculated iteratively throughout the projections GDP per capita and Employment rate : Based on naive growth and employment scenarios (own calculations)

International Migration Scenarios Sensitive to Growth?: 

International Migration Scenarios Sensitive to Growth?

Slide20: 

Estimation of internal migration m =  +  (rural-urban income difference) +  (rural-urban unemployment difference) m : rural-urban migration flow as a percentage of rural population Data: 1981-2000 (Source: SIS) Estimation Method: Ordinary Least Squares Simulation Period: 2003-2030 Growth and Unemployment Scenarios for Turkey

Slide21: 

Assumptions of the Growth Simulation Urban production function  Urban GDP=AF(L) Technology is exogenous Capital is slack variable Constant returns to scale Wages in the urban sector are endogenous and equal to marginal productivity of labor in the urban sector

Slide22: 

Assumptions of the Growth Simulation Growth of rural GDP is exogenous Employment creation in the urban sector is based on Harris-Todaro assumption The rate of urban GDP growth in excess of the productivity increase in the urban sector creates employment

Slide23: 

Assumptions of the Simulation UNDP medium variant population projection is used Working age fraction is assumed to grow exponentialy starting from 63% in 2003 and reaching 68% in 2030 (based on UNDP projection and other studies on population) Migrants are assumed to adopt urban labor force participation rate

Slide24: 

Assumptions of the Simulation Urban labor participation rate grows exponentialy starting from 43% in 2003 and reaching 60% in 2030 (converges to EU average) Full employment assumption in the rural sector

Slide25: 

GDP (NA) GROWTH 0.065 GDP (A) GROWTH 0.02 PRODUCTIVITY GROWTH 0.03 YEAR UNEMPLOYMENT RATE 2005 0.14 2015 0.13 2030 0.05 High Growth Scenario

Slide26: 

GDP (NA) GROWTH 0.04 GDP (A) GROWTH 0 PRODUCTIVITY GROWTH 0.015 YEAR UNEMPLOYMENT RATE 2005 0.15 2015 0.23 2030 0.27 Low Growth Scenario

Intermediate Scenario : 

GDP (NA) GROWTH 0.05 GDP (A) GROWTH 0.02 PRODUCTIVITY GROWTH 0.02 YEAR UNEMPLOYMENT RATE 2005 0.14 2015 0.18 2030 0.15 Intermediate Scenario

Slide28: 

Two scenarios are simulated Common Assumptions Free Movement between EU and Turkey will come into force in 2015 Growth rate in the urban sector % 6.5 Growth rate in the rural sector % 2 Productivity growth in the urban sector % 3 FREE Dummy taking over GUEST Dummy taking over Note: From the growth scenarios, total unemployment rate is calculated as a weighted average of urban and rural unemployment rates according to their shares in total population. International Migration Scenarios Based on Large Country Sample

Extrapolation of Results with Germany to the EU-15 : 

Extrapolation of Results with Germany to the EU-15 Based on the distribution of migrants in the EU-15 (OECD- SOPEMI 2000) the simulation results are extrapolated to the EU-15 The number of Turkish migrants in Germany multiplied with 1.32 (Germany has 76% of the Turkish migrants in EU-15) To make growth and international migration simulations compatible, outflow of Turkish migrants deducted from the population projections iteratively

“FREE” Dummy Taking Over (large country sample): 

“FREE” Dummy Taking Over (large country sample)

“GUEST” Dummy Taking Over (large country sample): 

“GUEST” Dummy Taking Over (large country sample)

Comparison of the Two Scenarios: 

Comparison of the Two Scenarios

Comparison of the Two Scenarios: 

Comparison of the Two Scenarios Net Change in Turkish Migrant Stock in the EU Turkish Migrant Stock in the EU

How to Inflate the Estimates Upwards?: 

How to Inflate the Estimates Upwards?

Migration Experience of Cohesion Countries: 

Migration Experience of Cohesion Countries

Model Specification with the Cohesion Countries and Turkey: 

Model Specification with the Cohesion Countries and Turkey mfht = αh + β1mfh,t-1 + β2 [ln(eft) - ln(eht)] + ufht mfht: The share of migrants from country h residing in country f in per cent of home population w: Wage (proxied by GDP-PPP per capita) e: Employment rate (1-unemployment rate) h, f,t: Home, foreign countries and year, respectively. Estimation Method: SUR

Estimation Results with the Cohesion Countries and Turkey: 

Estimation Results with the Cohesion Countries and Turkey

Actual vs. Fitted Turkish Migrant Stock (Estimation with Cohesion Countries and Turkey): 

Actual vs. Fitted Turkish Migrant Stock (Estimation with Cohesion Countries and Turkey)

Scenarios Using the Estimates Obtained from the Cohesion Countries and Turkey: 

Scenarios Using the Estimates Obtained from the Cohesion Countries and Turkey Two scenarios are simulated Common Assumptions Free Movement between EU and Turkey will get get into force in 2015 Growth rate in the urban sector % 6.5 Growth rate in the rural sector % 2 Productivity growth in the urban sector % 3 FREE Dummy taking over GUEST Dummy taking over

Slide40: 

Scenario FREE Dummy Taking Over Using the Estimates Obtained from the Cohesion Countries and Turkey

Scenario GUEST Dummy Taking Over Using the Estimates Obtained from the Cohesion Countries and Turkey: 

Scenario GUEST Dummy Taking Over Using the Estimates Obtained from the Cohesion Countries and Turkey

Comparison of the Two ScenariosUsing the Estimates Obtained from the Cohesion Countries and Turkey: 

Comparison of the Two Scenarios Using the Estimates Obtained from the Cohesion Countries and Turkey

Comparison of the FREE and GUEST ScenariosUsing the Estimates Obtained from the Cohesion Countries and Turkey: 

Comparison of the FREE and GUEST Scenarios Using the Estimates Obtained from the Cohesion Countries and Turkey Net Change in Turkish Migrant Stock in the EU Turkish Migrant Stock in the EU

Higher Numbers?: 

Higher Numbers?

Model Specification with Turkey only: 

Model Specification with Turkey only mft = α + β1 mf,t-1 + β2 ln(wft-1/wft-2)+ β3 ln(eft / et) + uft mft: The share of migrants from country Turkey residing in country Germany in per cent of Turkish population w: Wage (proxied by GDP-PPP per capita) e: Employment rate (1-unemployment rate) f,t: Foreign country (Germany) and year respectively Method: OLS

Estimation Results with Turkey only: 

Estimation Results with Turkey only

Actual vs. Fitted Migrant Stock (Turkey only): 

Actual vs. Fitted Migrant Stock (Turkey only)

Scenario Turkey Member – Free Movement of LaborHigh Growth Scenario: 

Scenario Turkey Member – Free Movement of Labor High Growth Scenario

Scenario No Membership – No Free MovementLow Growth Scenario: 

Scenario No Membership – No Free Movement Low Growth Scenario

Comparison of the Two Scenarios(Turkey only sample): 

Comparison of the Two Scenarios (Turkey only sample)

Comparison of the Two Scenarios: 

Comparison of the Two Scenarios Net Change in Turkish Migrant Stock in the EU Turkish Migrant Stock in the EU

A Summary of Simulations:Migration from TurkeyNet Change in Migrant Stock in EU from Turkey 2004-2030: 

A Summary of Simulations: Migration from Turkey Net Change in Migrant Stock in EU from Turkey 2004-2030

Slide53: 

Large Country Sample (SUR) – High Growth – Membership – Free Movement Cohesion Countries and Turkey Sample (SUR) – High Growth – Membership – Free Movement Turkey Only (OLS Estimation)

Low Growth – No Membership No Free Movement 2.734.000+Add an Intervention and/or Insurgency Dummy?: 

Low Growth – No Membership No Free Movement 2.734.000 + Add an Intervention and/or Insurgency Dummy?

Downward Adjustments: 

Downward Adjustments

Demographic Transition of Turkey: 

Demographic Transition of Turkey Propensity to Migrate according to age groups Based on reponses to NIDI-Hacettepe Survey (TIMS)

Slide57: 

Demographic Transition of Turkey

Slide58: 

Demographic Transition of Turkey Difference in year 2030 296.000

Slide59: 

Demographic Transition of Turkey Difference in year 2030 335.000