Introduction The migration model in EUROPOP24 Giampaolo LANZIERI Eurostat Unit F-1: Demographic and Migration Statistics Nowadays, migration is the most important component of population change. Migration is also the most volatile component. More scientific interest but frustrating results. Scenario approach: quantifying the stories Constancy is not a neutral assumption. Eventually constancy should be for all components. Past Eurostat models: 198 and 1985 round no migration, 199 constancy, 1995 short-term trends + target value. 1 2 Theoretical background The story - 1 Many partial theories developed so far: Neoclassical, World Systems, New Economics, Social Network, Dual Labour Market (classification from Howe and Jackson, 26). but few unifying approaches: Migration Systems (Kritz et al., ), or Massey (22). Difficulty to make them operational, especially for forecasting purposes very little/no use in official migration projections. Economy is a key driver for migration continued growth but not disappearance of disparities, especially in the short run between new MSs and EU15. Freedom of labour for EU8 new MSs in EU15: 24 26 27 29 211 BE, DK, LU, NL, FI EL, ES, FR, IT, PT DE, AT Assumed IE, SE, UK IE, SE, UK EL, ES, IT, PT, FI NL Opened 3 4 The story - 2 Only partial replacement migration. Migration is not the only answer to shortages of labour force: increase of labour participation rates, pro-fertility policies, delocalisation of production sites, etc. (see also UN 2). Increasing competitiveness on the migrants market : other economies of the world are expected to deal with labour force shortages. Moreover, development of new economies (e.g., China, India, etc.) may divert migration streams. Social cohesion: massive increases in immigration would also increase the challenge of integration, especially if combined with population decline. Third demographic transition? (Coleman, 26) 5 The story - 3 EU policies on migration and asylum: since the EC in Tampere in October 1999, many actions put in place. General orientation towards the management/control of the migratory flows ( economic/selective migration ). Reduction of the inflows of illegal migrants: set up of more efficient systems and regulations (e.g., new Schengen Information System). Difficult everyday life out of the legal channels. Demographic transition in the developing countries: reduction in fertility in the less developed countries, at different speeds. For instance, TFR Tunisia from 3.38 to 2.9 (below replacement level) in only 9 years. 6
NRR 2.5 2. 1.5 Demographic development Estimated and projected net reproduction rates (source: UNPD, World Population Prospects - 26 revision) Estimated Basic assumption Migration trends (not the yearly changes) are not supposed to radically change in the future: from one side economic and probably environmental factors will drive an increase of the flows towards EU; from the other side, social and (at a later stage) world demographic factors will play as constraints. 1. Projected.5. 195-1955- 196-1965- 197-1975- 198-1985- 199-1995- 2-25- 21-215- 22-225- 23-235- 24-245- 1955 196 1965 197 1975 198 1985 199 1995 2 25 21 215 22 225 23 235 24 245 25 Less developed regions Least developed countries Africa Quantification of the assumptions for a TREND scenario 7 8 The model for the new MSs The model for the EU15 Scarce availability of data. Major socio-economic changes happened in the recent past. Model: a) estimation of net migration in 23 based on trends -22; b) minimum value assumed in 212-13; c) target values for 25 for five clusters of countries: CZ/HU, PL/SK, EE/LV/LT, MT/CY/SI, BG/RO; d) Minimum and target values bridged by logistic curve. No availability of separate in- and out-migration flows unit of analysis: net migration calculated as difference between population and natural change. Model based on 3 main components: Time series analysis International consistency Expert opinion 9 1 Past trends EU migration trends Data analysis for each country to identify peculiarities: regularisations, special events, etc. Millions 2.5 Residual total net migration 196-26 in the EU27 and EU15 areas Post-war European Migration (Salt et al., 2): 2. 195s: liberal immigration 196s-1973: guest workers 1973-1989: consolidation and early asylum Mid-198s onwards: high asylum phase From 1989: opening of Central and Eastern Europe 199s: irregular migration 1.5 1..5. -.5 196 197 198 199 2 22 24 26 eu27 eu15 11 12
and EU15 national trends Residual total net migration 196-26 for the EU15 Member States Asylum applications New asylum applications in EU27 and EU15 countries, 1985-26 2 7 Crude rate of net migration (x1) 15 1 5-5 -1 196 197 198 199 2 22 24 26 absolute numbers, thousands 6 5 4 3 2 1 EU-27 EU-15-15 -2 BE DK DE IE GR ES FX IT LU NL AT PT FI SE UK 1985 1987 1989 199 1991 1993 1995 1997 years 1999 2 21 22 23 24 25 26 13 14 Time series analysis Migration trends ARIMA models. In most cases, AR(1) with dummies/transfer functions to model special events (e.g., regularisations). Previously tested for male-female crosscorrelation with temporal lag (VAR in state-space form). In general, no statistical significance. It is a stochastic projections for migratory flows intervals of confidence. Long-term time series: extrapolation over a long period based on an observed similar long period. 1 8 6 4 2-2 196 197 Corrected net migration Germany 198 199 2 22 24 26-4 15 16 Migration trends Migration trends Corrected net migration Germany Corrected net migration Germany (yellow line) and Spain (red line) 1 8 1 8 6 6 4 4 2 2 196 197 198 199 2 22 24 26 196 197 198 199 2 22 24 26-2 -2-4 -4 17 18
Migrants 1 9 8 7 6 5 4 3 2 1 Output from ARIMA Total Net Migration ARIMA prediction and 5% confidence interval The NETHERLANDS International consistency Important to distribute migratory flows across countries and to avoid that the same migrants inflows are taken into account by more countries. Ensure coherence between assumptions for EU15 and new MSs. Distinction in intra- and extra-eu15 flows. -1 197 1973 1979 1985 1991 1997 2 23 26 29 212 215 218 221 224 227 23 233 236 239 242 245 248 ACTUAL PREDICT UCL5 LCL5 19 2 Extra-EU15 flows - 1 1) From the census 21 data, identification of the most relevant sending countries: the by then Accession and Acceding Countries, all the North Africa, Balkans area and the former European Soviet Union countries have been included a priori (27 countries); Additionally, the top-five communities of migrants (other than from EU and neighbouring countries) have been identified for each MS (36 countries); 8 further communities were included, as they had relevant presence in EU, even if not between the top-5 in any country. Total: migrants from 71 sending countries Extra-EU15 flows - 2 2) Assumptions on the total emigration flows from these 71 countries: for the by then Accession and Acceding Countries, the migration flows as from the EUROPOP24 model; for all the rest, elaborations on data from UN assumptions available by then (World Population Prospect 22 Revision). 3) Assumptions for each sending country on the quota of the total emigration flow going to EU25 countries, based on historical links, language, geographical proximity, etc.. 21 22 Extra-EU15 flows - 3 4) Distribution of the flows going to EU25 between EU1 and EU15 area based on the stock of migrants in the two areas for each sending country (network effect). 5) Allocation to the single EU15 MSs on the basis of the stock of migrants (tested also considering the migratory flows in the past decade and the population quota). 6) A minimum threshold for outflow to EU15 has been set for the new MSs. This has corrected less plausible flows from selected countries (e.g., SK). Migrants 8 7 6 5 4 3 2 1-1 24 26 28 Extra-EU15 flows - 4 Total net migration from selected geographical areas (Base assumption for international consistency component) EU15 area (24-25) 21 212 214 216 218 22 222 224 226 228 23 232 234 236 238 24 EU1 AC CC Balkans Ex USSR Maghreb 242 244 246 248 25 23 24
Migrants 12 1 8 6 4 2-2 24 From new to old MSs Alternative assumptions for total net migration from EU1 and Accession Countries areas (International consistency component) EU15 area (24-25) 26 28 21 212 214 216 218 22 222 224 226 228 23 232 234 236 238 24 242 244 EU1 baseline AC baseline EU1 high AC high EU1 low AC1 low 246 248 25 Extra-EU15 flows - 5 7) A supplementary flow of migrants has been estimated from the latest trends on asylum applications. This flow was not attributed to specific sending countries ( non-attributable migration ). 8) The distribution across EU15 MSs has been based on the average quota of asylum applications and on the being or not a border country (taking into account their size). 9) High and low assumptions were formulated based also on stock of migrants, GDP quotas and population size. 25 26 Intra-EU15 flows 1) Intra-EU15 flows has been estimated for the most recent years. 2) For the base assumption, these intra-eu15 flows have been kept constant. 3) For the low assumption, the deviation from the average EU15 GDP has been taken into account. 4) For the high assumption, the EU15 economies has been assumed equally attractive in the long run and thus the intra-eu15 net migration flows converging to zero. 27 Migrants Flows to EU15 Composition of the migration flows (Base assumption for international consistency component) EU15 area (24-25) 1 9 8 7 6 5 4 3 2 1 24 26 28 21 212 214 216 218 22 222 Not attributable migration Others Maghreb Ex USSR Balkans CC AC EU1 224 226 228 23 232 234 236 238 24 242 244 246 248 28 25 Expert opinion Experts can provide views which are not caught by analysis of past trends. Dangers: often generic arguments with no quantification; sensitivity to the way information is proposed (UPE project); no assessment of the performances of the experts; who is an expert? Inclusion in the model of the latest national assumptions on migration. 29 Merging the components Need for combination of different inputs: trends extrapolation, international consistency and expert opinion. Transparent approach: weighted average. Simple mean best performing approach for combination of forecasts (Stock and Watson, 24). M w t t i i e e t i e y = y Mc, y+ y Mc, y+ y Mc, y y+ y+ y = Possibility of different weights over country, component and time. w t i e M y =.3 M y +.3 M y +. 3 M y 1 3
Example 1 Example 2 EUROPOP24: Weighted total net migration and its components GERMANY Base assumption (24-25) EUROPOP24: Weighted total net migration and its components The NETHERLANDS Base assumption (24-25) 3 25 8 6 2 4 Migrants 15 1 Migrants 2 195 1953 1956 1959 1965 1971 1977 198 1983 1989 1995 21 24 27 21 213 216 219 222 225 228 231 234 237 24 243 246 249-2 5-4 24 26 28 21 212 214 216 218 22 222 224 226 228 23 232 234 236 238 24 242 244 246 248 25-6 Extrapolated Assessed National Weighted Extrapolated Assessed National Weighted Observed 31 32 Validating the results - 1 Once estimated the total net migration, the outcomes have been validated using several methods: Pull models: country-specific regression models with pull socio-economic determinants of migration. The specification process often identified unemployment as main driver. Other sources: comparison with previous Eurostat assumptions, with values from other sources (UNPD, scientific literature), and with policy views (UN and OECD-SOPEMI reports). Projected quota of migrants: plausibility of the stock of migrants generated by these assumptions on migration. Importance of integration issues. Validating the results - 2 Special events: in case of forthcoming event (like it was the regularisation in Spain at end 24), figures may be appropriately adapted. Consultation on CIRCA: national representatives asked to agree/disagree with the proposed assumptions. High and low assumptions prepared with the same approach. Composed by: 5% confidence limits from ARIMA model; variations of the international consistent flows: variants of the national assumptions. 33 34 Overall view Catching the exact value? Trends extrapolation 1 8 6 Estimated and assumed corrected net migration Germany 4 Data analysis International consistency Weighted prediction Coherence analysis Calibrated prediction 2-2 195 1952 1954 1956 1958 196 197 198 199 2 22 24 26 28 21 212 214 National forecasts -4-6 -8-1 35 36
Volatility of immigration SPAIN: estimates of immigration flows (in thousand) 8 7 6 5 4 Regularisation in 24/25 3 Regularisation in 2/21 2 Change of source 1 1985 1987 1989 199 1991 1993 1995 1997 1999 2 21 22 23 24 25 LFS - persons residing abroad 1 year ago Immigrants from Municipal Registers 37 Migration in Spain Spain is experiencing an unexpected high immigration. However: The regularisations might have inflated the migration figures in specific years. The Padron Continuo (Population Register) might be affected by duplications (residents can also use different documents to register) and over-coverage (persons who have emigrated but not de-registered). According to a recent law (Resolution 28 April 25), all non-european foreigners without a permanent resident permit must renew their municipal registration every two years. The first application of this law resulted in a reduction of the population figures for 26 by 49.. Caution in formulating migration assumptions 38 Model performance: EU27 Estimated and projected residual net migration in the EU27 area Model performance: EU15 Estimated and projected residual net migration in the EU15 area Millions 2.5 Millions 2.5 2. 2. Net migrants 1.5 1. Net migrants 1.5 1..5.5. 23 24 25 26. 23 24 25 26 Base Low High Estimated Base Low High Estimated 39 4 Assessment Caveats: the figures considered as observed are in facts estimates or assumptions, and could be subject to relevant changes after next census. the unit of analysis is the residual net migration, thus including all the statistical adjustments. Results: all the variants of the model have caught the decreasing trend. For the EU27, the % differences for the High assumption are very small and decreasing: 24 25 26-5.6% -4.1% -3.% especially considering the volatility of migration: in 3 years, decrease by 25%, from 2 to 1.5 million. Larger differences for single MS. International models United Nations: future path of migration set on the basis of past migration estimates and assessment of policy stance of countries with regard to future migration flows. No alternative trends, only one (usually) constant assumption based on: information on net migration; data on labour migration flows; estimates of undocumented of irregular migrants; data on refugee movements in the recent past. UPE: expert opinion for point forecast; combination of stochastic model and expert opinion for the confidence intervals. 41 42
National models National model: Spain Italy: models for emigration and immigration; immigration extrapolated by an ARIMA model on the corrected time series for one year ahead and then kept constant. United Kingdom: average of a trend extrapolation (exponential smoothing) over the next ten years by 8 subgroups, adjusted by expert opinion for switchers and asylum seekers. Value kept constant over the projections period. Spain: from data 2-24, extrapolation to 21 and then assumptions. 7 6 5 4 3 2 1 21 23 25 27 29 211 INE assumptions on total net migration migration 24-26 213 215 217 219 221 223 225 227 229 231 233 235 237 239 241 243 245 247 249 251 253 255 257 259 Hypothesis 1 Hypothesis 2 Hypothesis 3 43 44 Some conclusions The Eurostat migration model can be considered: comprehensive, as it includes analysis of the time series, as well as of the determinant of migration and expert opinions; transparent, as the method clearly describe how the final result is obtained (no black boxes) and based also on consultation of national representatives; internationally consistent, as this element is a fundamental part of the model; innovative, as it includes stochastic models (thus confidence intervals for uncertainty) and robust combination of projected flows. Thanks for your attention! For any further information please contact: giampaolo.lanzieri@ec.europa.eu 45