Presence of language-learning opportunities abroad and migration to Germany

Presence of language-learning opportunities abroad and migration to Germany Matthias Huber Silke Uebelmesser University of Jena, Germany International Forum on Migration Statistics OECD, Paris, January 218 Funded by the German Science Foundation (DFG, grant number UE 124/2-1) see also www.wi.uni-jena.de/dfg

Motivation Benets of migrants' prociency in the language of the destination country Labour market outcomes: earnings (Dustmann and Soest 21; Chiswick and Miller 1995) and employment probability (Dustmann and Fabbri 23) increase Social integration: probability of intermarriage increases and the likelihood of living in an ethnic enclave decreases (Bleakley and Chin 21) Language prociency as determinant of migration ows But linguistic distance has been found to have a negative impact due to higher costs of language acquisition (Adsera and Pytlikova 215; Belot and Hatton 212) 2 / 41

Language learning and migration decisions Focusing on linguistic distance neglects actual language acquisition despite higher costs. Children-age language learning Decision determined by factors outside the learner's direct control: language skills migration decision Compulsory foreign language learning positively related to migration ows within the EU (Fenoll and Kuehn 216) Adult-age language learning Decision determined by learner's migration (intention): migration decision language learning Migration as determinant of adult-age language learning (Uebelmesser and Weingarten 217) 3 / 41

Research Question: What is the eect of the presence of language learning opportunities for adults abroad on migration to Germany? Procedure: Basic panel regressions Robustness checks (especially to address reverse causality concerns) Results in a nutshell: Language learning opportunities (here: Goethe Institutes) are positively correlated with migration to Germany. There is evidence that this is a causal eect from language learning opportunities to migration. 4 / 41

Data Goethe-Institut (GI) Map Main actor in promoting German culture and language worldwide. Institutes worldwide oer Language services: courses and standardized exams. Information on the German culture and society: (cultural) events and libraries. Funded by the German government; language services by fees. Annual reports of the GI provide information about institutes. 5 / 41

Data From its annual reports, we constructed three datasets. 1. Dataset about the regional distribution of the GI from 1965-214, the opening and closing years of all institutes and whether they provide language services. 214: 137 institutes in 86 countries (plus 12 in Germany) 2. Dataset about language learning in GI placed in countries all over the world. We report numbers on course registrations (199-214), sold course units (1972-1989 and 1997-214) and exam participation (1986-214). 214: 229,72 registrations and 17,113,4 sold course units, 287,63 exams. 3. Dataset about information on language course participation at GI in Germany (1966-215). 214: 13,459 European registrations and 2,397 non-european registrations from about 2 countries. In 214, almost 1.5m people migrated to Germany. 6 / 41

Data For this study here, we use Dataset 1 about the number of institutes (including openings and closings). Our sample is a balanced panel of 77 countries from 1968 214. In 214, 51 countries had at least one GI with the number of institutes in these countries amounting to 86. 152,6 registrations took place. Almost 55, migrants from these countries with a GI came to Germany (and 621, from all countries in our sample). 7 / 41

Data Never a GI Figure 1: In some years a GI Always a GI Not in Sample The presence of GI (our sample) 8 / 41

Data 6 4 Countries 2 1965 197 1975 198 1985 199 1995 2 25 21 Year Countries with any GI Countries with GI with language services Countries without any GI Figure 2: Number of countries with GI (our sample) 9 / 41

Data Institutes 16 14 12 1 8 6 4 2 1965 197 1975 198 1985 199 1995 2 25 21 Year Total number of institutes Total number of institutes with language services Total number of institutes without language services Figure 3: Number of institutes (our sample) 1 / 41

Data Closings Openings 1 5 1969 197 1971 1972 1973 1974 1975 1976 1977 1978 1979 198 1981 1982 1983 1984 1985 1986 1987 1988 1989 199 1991 1992 1993 1994 1995 1996 1997 1998 1999 2 21 22 23 24 25 26 27 28 29 21 211 212 213 214 Number of language institutes Year Figure 4: Openings and closings All, Language Institutes, Language Institutes (jr), Unbalanced Panel 11 / 41

Estimation Strategy Fixed-eects model y jt = α GI jt + β x jt + φ td t + φ j d j + φ j d jt + η jt (1) y jt GI jt x jt d t, d j, d jt log of migration rate (annual migr. inows (Destatis)/ population size of origin country (PWT)) number of (language) institutes vector of control variables: log GDP/capita (PWT), EU, log population (PWT), conicts (UCDP), log bilateral trade ows (Destatis) log migrant stock time, origin-country and origin-country-1-year FE Summary Statistics Balanced panel dataset with 77 countries from 1968 214. Regressions weighted by population size. 12 / 41

Results: basic specications DV: log migration rate (1) (2) (3) (4) Number of language institutes.427**.537***.558***.685*** (.175) (.179) (.184) (.184) log GDP per capita -.343*** -.36*** -.154* (.866) (.931) (.792) EU member.454***.441***.443*** (.12) (.123) (.13) log population.68*.537*.38 (.317) (.314) (.289) Conict.62**.525** (.257) (.248) log (Exports+Imports) -.355 -.852* (.58) (.435) log (Migrant Stock / Population), lag=1.646*** (.611) Constant -1.98*** -18.5*** -16.91*** -7.2 (.91) (5.46) (5.592) (5.112) Observations 3,619 3,619 3,619 3,619 Adjusted R-squared.967.968.969.973 Year FE Yes Yes Yes Yes Country FE Yes Yes Yes Yes Country*1-year FE Yes Yes Yes Yes Countries 77 77 77 77 Years 1968-214 1968-214 1968-214 1968-214 Robust standard errors in parentheses; *** p<.1, ** p<.5, * p<.1. Observations are weighted by population size. 13 / 41

Results: robustness checks I - GI institutes DV: log migration rate (1) (2) (3) (4) (5) Number of institutes.282** (.124) Number of institutes without language services -.272 (.114) Number of language institutes.796***.594***.576** (.198) (.223) (.23) Number of language institutes, lag=1.469**.439* (.214) (.225) Number of language institutes, lag=2.794 (.194) log (Number of GI with language services.299** per 1m inhabitants) (.143) Observations 3,619 3,619 3,542 3,465 3,619 Adjusted R-squared.973.974.973.973.973 Year FE Yes Yes Yes Yes Yes Country FE Yes Yes Yes Yes Yes Country*1-year FE Yes Yes Yes Yes Yes Other controls included Yes Yes Yes Yes Yes Countries 77 77 77 77 77 Years 1968-214 1968-214 1968-214 1968-214 1968-214 Robust standard errors in parentheses *** p<.1, ** p<.5, * p<.1; observations are weighted by population size of the origin. 14 / 41

Results: robustness checks II interaction eects (1) (2) (3) (4) (5) (6) DV: log migration rate economic geographic linguistic EU conict distance distance distance Language institutes.658***.383**.117***.24.613***.679*** (.181) (.151) (.357) (.214) (.195) (.182) Language institutes *....177*** -.615.969***.432 -.313** (.529) (.49) (.343) (.419) (.147) Observations 3,619 3,619 3,619 3,619 3,619 3,619 Adjusted R-squared.973.974.973.973.973.973 Year FE Yes Yes Yes Yes Yes Yes Country*1year FE Yes Yes Yes Yes Yes Yes Other controls Yes Yes Yes Yes Yes Yes Countries 77 77 77 77 77 77 Years 1968-214 1968-214 1968-214 1968-214 1968-214 1968-214 Robust standard errors in parentheses *** p<.1, ** p<.5, * p<.1; observations are weighted by population size of the origin. 15 / 41

Possible issues (a) Reverse causality (b) Multi-lateral resistance (MLR) (c) Omitted variable bias 16 / 41

(a) Reverse Causality: Switzerland Decision by Federal Foreign Oce and GI to open (and close) institutes. Biased estimation possible: decision potentially not exogenous to migration to Germany. But: decision exogenous to migration ows to Switzerland. Analysis with DV: log migration rate to Switzerland... to assess the relevance of reverse causality to study the language eect vs. information eect of the GI by using variation of languages within Switzerland. Correlation, Rel. Size 17 / 41

Reverse Causality: Switzerland (1) (2) (3) DV: log migration rate Germany Switzerland Switzerland (non-german-speak.) (German-speak.) Number of language institutes.35*.268.72*** (.167) (.24) (.265) Observations 1,771 1,771 1,771 Adjusted R-squared.978.983.969 Year FE Yes Yes Yes Country Fe Yes Yes Yes Country*1-year FE Yes Yes Yes Other controls Yes Yes Yes Countries 77 77 77 Years 1992-214 1992-214 1992-214 *** p<.1, ** p<.5, * p<.1. Shorter time-period Robust, Conclusion 18 / 41

(b) MLR: common correlated eects estimator Migration decisions not only inuenced by the chosen destination country's attractiveness, but also by the attractiveness of other (alternative) destinations. CCE estimator (Pesaran 26) controls for multilateral resistance of migration (Bertoli, Fernandez-Huertas Moraga 213). y jt = α GI jt + β x jt + φ td t + φ j d j + φ j d jt + λ j z t + η jt (2) y jt GI jt x jt d t, d j and d jt log of migration rate number of (language) institutes vector of independent variables time, origin-country and origin-country-1-year FE and the cross-sectional averages of independent and dependent variables weighted with ω jt (population) interacted with country dummies λ j 1 z t = j ω ω jt y jt, ω jt x jt jt j j 19 / 41

Robustness checks: CCE estimator DV: log migration rate (1) (2) (3) (4) Number of language institutes.68***.776***.723***.683*** (.121) (.118) (.116) (.116) Observations 3,619 3,619 3,619 3,619 Adjusted R-squared.974.981.983.984 Year FE Yes Yes Yes Yes Country FE Yes Yes Yes Yes Country*1-year FE Yes Yes Yes Yes Other controls Yes Yes Yes Yes Countries 77 77 77 77 Years 1968-214 1968-214 1968-214 1968-214 CCE-test (p-value) Observations are weighted by population size of the origin country; results are estimated with the CCEestimator (Pesaran 26); the CCE-test is a F-test on the joint signicance of the cross-sectional averages of all dependent and independent variables interacted with country dummies. Standard errors in parentheses. 2 / 41

(c) Omitted variable bias As next steps, we want to include information about possible other inuences on migration that might aect the number of institutes as well: Language learning at schools: Compulsory language learning at schools in Europe (data used by Fenoll and Kuehn 216) German schools abroad (Auslandsschulen) International Association of German Teachers (IDV) Percentage of pupils learning German (Eurostat) Language learning at universities: Institutes of German studies abroad (Germanistik) German language courses Other inuences: Branches of chambers of commerce (AHK) German Academic Exchange Service (DAAD) 21 / 41

Conclusion The number of institutes provided by the GI is positively correlated with migration rates to Germany. GI also aect migration ows to the German-speaking part of Switzerland, but not to the French- and Italian-speaking part. Causal eect from language learning opportunities to migration ows. The relationship is due to language learning and not due to other eects (like information eect) coming with the GI. So: Language learning shapes international migration ows beyond linguistic properties, like linguistic distance. Contrary to children language learning, adult language learning is within reach of the policy-makers in the destination country. 22 / 41

Thank you for your attention! For more information, see www.wi.uni-jena.de/dfg 23 / 41

Theoretical considerations Individuals choose the destination country which maximizes expected utility. Cost-benet analysis on the basis of origin and destination country characteristics. Benets: monetary (e.g. wage, social security,...) and non-monetary (safety, partner, culture) Costs: monetary (e.g. transportation, visa, temporary unemployment) and non-monetary (e.g. leaving family and friends, social integration) Language skills are costly, but increase expected income in destination country. On an aggregate level, language learning opportunities can be expected to play a role in the migration decision. 24 / 41

Data Figure 5: Go back Countries with GI in 213 25 / 41

Data Fraction.2.4.6.8.1 1 2 3 4 5 6 7 Registrations Figure 6: Distribution of institutes (our sample), 213 26 / 41

Data Summary statistics Variable Obs Mean Std. Dev. Min Max Migration rate (emigration to Germany/pop.) 3619 Overall.2.61 5.48e-7.184 Emigration to Germany 3619 Overall 4622.816 15894.45 3 25152 Number of institutes 3619 Overall 1.5898 1.99367 14 Between 1.87117 7.93617 Within.71969-4.166 9.8345 Number of language institutes 3619 Overall 1.2685 1.5847 9 Between 1.48791 6.59575 Within.5756-3.6545 5.47361 Number of language institutes per 1m inhabit. 3619 Overall.753.11899.78797 GDP per capita 3619 Overall 162.91 11494.58 142.3924 6514.98 EU member 3619 Overall.11246.31598 1 Population in 1m 3619 Overall 41.31132 111.2261.2153 1295.292 Conict 3619 Overall.23985.52199 2 Migrant stock/population 3619 Overall.175.511 2.86e-6.4581 Exports + Imports 3619 Overall 6.81e+9 1.94e+1 1818 1.67e+11 Variation (at least one GI), Go back 27 / 41

Data and estimation strategy gi_migrants.3.25.2.15.1.5..3.25.2.15.1.5..3.25.2.15.1.5..3.25.2.15.1.5..3.25.2.15.1.5..3.25.2.15.1.5. ARG AUS BGD BOL BRA CAN CHL CHN CIV CMR COL EGY ESP ETH FIN FRA GBR GHA GRC HKG HUN IDN IND IRL ISR ITA JOR JPN KEN KOR LKA MAR MEX NGA NLD NOR NZL PER PHL PRT ROU SGP TGO THA TUN TUR URY USA VEN.3.25.2.15.1.5. 199 1995 2 25 21 199 1995 2 25 21 199 1995 2 25 21 199 1995 2 25 21 199 1995 2 25 21 199 1995 2 25 21 199 1995 2 25 21 year Figure 7: Share of course participants with migration intention on total migration (proxied), 213 28 / 41

Data and estimation strategy.3.25.2 gi_migrants.15.1.5. 199 1995 2 25 21 year Figure 8: Share of course participants with migration intention on total migration (proxied), France 29 / 41

Results: robustness checks - unweighted DV: log migration rate (1) (2) (3) (4) Number of language institutes.589***.642***.75***.788*** (.28) (.24) (.197) (.185) log GDP per capita -.534*** -.454*** -.236*** (.751) (.781) (.619) EU member.294*.312**.373*** (.152) (.151) (.113) log population -.723** -.733** -.368 (.31) (.296) (.258) Conict.883***.91*** (.229) (.22) log (Exports+Imports) -.86** -.739** (.354) (.316) log (Migrant Stock / Population), lag=1.673*** (.448) Constant -1.95*** 5.669 6.772 5.37 (.835) (5.355) (5.352) (4.667) Observations 3,619 3,619 3,619 3,619 Adjusted R-squared.954.956.956.964 Year FE Yes Yes Yes Yes Country FE Yes Yes Yes Yes Country*1-year FE Yes Yes Yes Yes Countries 77 77 77 77 Years 1968-214 1968-214 1968-214 1968-214 Robust standard errors in parentheses; *** p<.1, ** p<.5, * p<.1 3 / 41

TZA URY USA VEN ZAF ARG AUS BDI BEN BFA BOL BRA CAN CHL 1 5 CIV CMR COG COL CRI DNK DOM DZA ECU 1 5 EGY ESP ETH FIN FRA GBR GHA GIN GRC 1 5 GTM HND HTI IDN IND IRL IRN ISL ITA Number of institutes 1 5 1 5 1 5 JOR JPN KEN KOR LBR MAR MDG MEX MLI MYS NER NGA NIC NLD NOR NPL NZL PAK PAN PER PHL PRT PRY ROM RWA SEN SLE 1 5 SLV SWE SYR TCD TGO THA TTO TUN TUR 1 5 197 199 21 197 199 21 197 199 21 197 199 21 1 5 197 199 21 197 199 21 197 199 21 197 199 21 197 199 21 Year Figure 1: Numbers of all institutes, by origin countries Go back 32 / 41

TZA URY USA VEN ZAF ARG AUS BDI BEN BFA BOL BRA CAN CHL 7.5 5. 2.5. CIV CMR COG COL CRI DNK DOM DZA ECU 7.5 5. 2.5. EGY ESP ETH FIN FRA GBR GHA GIN GRC 7.5 5. 2.5. GTM HND HTI IDN IND IRL IRN ISL ITA Number of language institutes 7.5 5. 2.5. 7.5 5. 2.5. 7.5 5. 2.5. JOR JPN KEN KOR LBR MAR MDG MEX MLI MYS NER NGA NIC NLD NOR NPL NZL PAK PAN PER PHL PRT PRY ROM RWA SEN SLE 7.5 5. 2.5. SLV SWE SYR TCD TGO THA TTO TUN TUR 7.5 5. 2.5. 197 199 21 197 199 21 197 199 21 197 199 21 7.5 5. 2.5. 197 199 21 197 199 21 197 199 21 197 199 21 197 199 21 Year Figure 11: Numbers of language institutes, by origin countries Go back 33 / 41

TZA URY USA VEN ZAF ARG AUS BDI BEN BFA BOL BRA CAN CHL 1 5 CIV CMR COG COL CRI DNK DOM DZA ECU 1 5 EGY ESP ETH FIN FRA GBR GHA GIN GRC Number of language institutes (incl. joint reporting) 1 5 1 5 1 5 1 5 1 5 GTM HND HTI IDN IND IRL IRN ISL ITA JOR JPN KEN KOR LBR MAR MDG MEX MLI MYS NER NGA NIC NLD NOR NPL NZL PAK PAN PER PHL PRT PRY ROM RWA SEN SLE SLV SWE SYR TCD TGO THA TTO TUN TUR 1 5 197 199 21 197 199 21 197 199 21 197 199 21 1 5 197 199 21 197 199 21 197 199 21 197 199 21 197 199 21 Figure 12: Numbers of language institutes, by origin countries language institutes (courses assumed to take place in case of joint reporting) Year Go back 34 / 41

Correlation between migration to Switzerland and to Germany log migration rate to Germany 6 8 1 12 6 8 1 12 6 8 1 12 6 8 1 12 6 8 1 12 6 8 1 12 6 8 1 12 GI ARG * GI AUS * GI BOL * GI BRA GI CAN GI CHL GI CIV GI CMR GI COL GI CRI * GI DNK * GI DZA GI EGY GI ESP * GI ETH * GI FIN * GI FRA GI GBR * GI GHA * GI GRC GI IDN GI IND * GI IRL GI IRN GI ITA * GI JOR GI JPN * GI KEN GI KOR GI MAR GI MEX * GI MYS GI NGA * GI NLD GI NOR * GI NPL GI NZL * GI PAK GI PER * GI PHL GI PRT * GI SEN GI SWE GI SYR GI TGO GI THA * GI TUN GI TUR * GI TZA GI URY GI USA * GI VEN GI ZAF No GI BDI * No GI BEN No GI BFA No GI COG * No GI DOM * No GI ECU * No GI GIN * No GI GTM * No GI HND No GI HTI * Goethe Institut No Yes 6 8 1 12 6 8 1 12 No GI ISL No GI LBR * No GI MDG No GI MLI No GI NER No GI NIC No GI PAN No GI PRY * No GI RWA * No GI SLE * No GI SLV * No GI TCD No GI TTO 16 12 8 16 12 8 16 12 8 16 12 8 16 12 8 16 12 8 16 12 8 16 12 8 16 12 8 log migration rate to Switzerland Figure 13: Correlation, within countries with and without GI Go back 35 / 41

. 1. Goethe Institut density No Yes.5.5..5 1. Correlation: log migration rates to Germany and Switzerland Figure 14: Correlation (density), countries with and without GI Go back 36 / 41

y. 1..75 Goethe Institut.5 No Yes.25.5..5 1. Correlation: log migration rates to Germany and Switzerland Figure 15: Correlation (cdf), countries with and without GI Go back 37 / 41

Flows to Switzerland relative to Germany 6 4 density 2..3.6.9 Immigration to CHE/ Immigration to DEU Figure 16: Swiss/German inows Go back 38 / 41

Flows to Switzerland relative to Germany 6 4 density 2..2.4 Mean immigration to CHE from country j/ Mean immigration to DEU from country j Figure 17: Swiss/German inows, by origin countries (1992-214) Go Back 39 / 41

Data Summary statistics (countries with at least one GI) Variable Obs Mean Std. Dev. Min Max Number of institutes 2773 (59 countries) Overall 1.96935 2.691 14 Between 1.91458.1721 7.93617 Within.82221-3.54129 9.54382 Number of language institutes 2632 (56 countries) Overall 1.73366 1.62277 9 Between 1.4915.1721 6.59575 Within.6697-3.18123 5.94643 Go back 4 / 41

Reverse Causality: Switzerland (1) (2) (3) (4) DV: log migration rate Germany Switzerland Switzerland Switzerland (non-german-speak.) (German-speak.) (German-speak.) Number of language institutes.35*.268.72***.671* (.167) (.24) (.265) (.389) Observations 1,771 1,771 1,771 1,472 Adjusted R-squared.978.983.969.959 Year FE Yes Yes Yes Yes Country Fe Yes Yes Yes Yes Country*1-year FE Yes Yes Yes Yes Other controls Yes Yes Yes Yes Countries 77 77 77 64 Years 1992-214 1992-214 1992-214 1992-214 *** p<.1, ** p<.5, * p<.1. In (4), countries are excluded with signicantly (at least on the 1 %-level) related migration ows to Germany and to the German-speaking part of Switzerland that have a variation in the number of language institutes in the period 1992 214. Shorter time-period Go back 41 / 41