Supplementary of the IAB-BAMF-SOEP survey of refugees in Germany (M5) 2017

Jacobsen, Jannes e al. Resea ch Repo Supplemen a y o he IAB-BAMF-SOEP su ey o e ugees in Ge many (M5) 2017 SOEP Su ey Pape s, No. 605 P o ided in Coope a ion wi h: Ge man Ins i u e o Economic Resea ch (DIW Be lin) Sugges ed Ci a ion: Jacobsen, Jannes e al. (2019) : Supplemen a y o he IAB-BAMF-SOEP su ey o e ugees in Ge many (M5) 2017, SOEP Su ey Pape s, No. 605, Deu sches Ins i u ü Wi scha s o schung (DIW), Be lin This Ve sion is a ailable a : h ps://hdl.handle.ne /10419/194098 S anda d-Nu zungsbedingungen: Die Dokumen e au EconS o dü en zu eigenen wissenscha lichen Zwecken und zum P i a geb auch gespeiche und kopie we den. Sie dü en die Dokumen e nich ü ö en liche ode komme zielle Zwecke e iel äl igen, ö en lich auss ellen, ö en lich zugänglich machen, e eiben ode ande wei ig nu zen. So e n die Ve asse die Dokumen e un e Open-Con en -Lizenzen (insbesonde e CC-Lizenzen) zu Ve ügung ges ell haben soll en, gel en abweichend on diesen Nu zungsbedingungen die in de do genann en Lizenz gewäh en Nu zungs ech e. Te ms o use: Documen s in EconS o may be sa ed and copied o you pe sonal and schola ly pu poses. You a e no o copy documen s o public o comme cial pu poses, o exhibi he documen s publicly, o make hem publicly a ailable on he in e ne , o o dis ibu e o o he wise use he documen s in public. I he documen s ha e been made a ailable unde an Open Con en Licence (especially C ea i e Commons Licences), you may exe cise u he usage igh s as speci ied in he indica ed licence. h ps://c ea i ecommons.o g/licenses/by-sa/4.0/ SOEP Su ey Pape s Se ies C – Da a Documen a ion The Ge man Socio-Economic Panel s udy Supplemen a y o he IAB-BAMF-SOEP Su ey o Re ugees in Ge many (M5) 2017 605 SOEP — The Ge man Socio-Economic Panel S udy a DIW Be lin 2019 Jannes Jacobsen, Ma in K oh, Simon Kühne, Jana A. Scheible, Raine Siege s, and Manuel Siege Jannes Jacobsen, Ma in K oh, Simon Kühne, Jana A. Scheible, Raine Siege s, and Manuel Siege Supplemen a y o he IAB-BAMF-SOEP Su ey o Re ugees in Ge many (M5) 2017 Jannes Jacobsen1Ma in K oh1 2 Simon Kühne2 Jana A. Scheible3Raine Siege s1 Manuel Siege 3 Feb ua y 28, 2019 1DIW Be lin – Deu sches Ins i u ü Wi scha s o schung e.V. 1Sozio-oekonomisches Panel (SOEP) 1Moh ens aße 58 110117 Be lin 2Uni e si ä Biele eld 2Fakul ä ü Soziologie - Me hoden de empi ischen Sozial o schung 2Uni e si ä ss aße 25 233615 Biele eld 3Bundesam ü Mig a ion und Flüch linge (BAMF-FZ) 2Fo schungszen um Mig a ion, In eg a ion und Asyl 2F ankens . 210 290461 Nü nbe g Con en s Lis o Tables II Lis o Figu es II 1 In oduc ion 5 2 Ta ge Popula ion and Sampling F ame 7 3 Sampling Design 7 3.1 Clus e ing o Add esses ............................. 8 3.2 Sampling o P ima y Sampling Uni s ..................... 8 3.3 Sampling o Seconda y Sampling Uni s .................... 9 4 Fieldwo k Resul s and Response Ra es 11 4.1 Add ess A ailabili y and Quali y ........................ 11 4.2 Fieldwo k .................................... 11 4.2.1 Ve sions o he Ques ionnai e in Di e en Languages ........ 12 5 C oss-Sec ional Weigh ing 15 5.1 Non-Response Weigh ing ............................ 16 5.1.1 Da a Sou ces and Documen a ion on Va iables ............ 17 5.1.2 Mul iple Impu a ion and Da a Coding ................ 18 5.1.3 Resul s and Calcula ion o Non-Response Weigh s .......... 19 5.2 Pos -S a i ica ion ................................ 21 5.3 Cha ac e is ics o C oss-Sec ional Weigh ................... 23 6 Appendix 25 Re e ences 26 Lis o Tables 1 La e Regis a ion and New A i als in he Sampling F ame - Adul s and Mino s ...................................... 6 2 Ancho Pe sons Ac oss T anches/Subsamples ................. 7 3 Sample Fac o s Ac oss Coun ies o O igin and Legal S a us ........ 9 4 Mino s and Adul s in Sampling F ame by Coun y o O igin ........ 10 5 Fieldwo k Resul s Wi h and Wi hou QNA .................. 12 6 Use o Visual T ansla ions in Ne Sample ................... 13 7 Use o Audio Files in Ne Sample ....................... 14 8 Cha ac e is ics o Household Non-Response Weigh - M5 .......... 21 9 Popula ion Cha ac e is ics Used in he Raking P ocess ........... 22 10 Summa y S a is ics o Household Weigh s .................. 23 11 Va iables Used Es ima ing he P opensi y Sco e ............... 25 Lis o Figu es 1 Response Ra es a he S a e and Municipal Le el .............. 13 2 Reduced Non-Response Model ......................... 19 3 Dis ibu ion o Indi idual Weigh s Ac oss All Th ee Weigh ing S eps . . . 23 The Ins i u e o Employmen Resea ch (Ins i u ü A bei sma k - und Be u s o schung, IAB) is an in- dependen ins i u e o he Fede al Employmen Agency in Nu embe g, Ge many. The IAB conduc s labo ma ke and occupa ional esea ch in Ge many on opics such as labo ma ke policy and social inequali y, and also does esea ch in he ields o s a is ical me hods and su ey me hodology (IAB 2017). The Mig a ion, In eg a ion and Asylum Resea ch Cen e o he Fede al O ice o Mig a ion and Re ugees (BAMF-FZ) conduc s esea ch on mig a ion o and om Ge many as well as esea ch on in eg a ion p ocesses in Ge many. The esul s a e used o mig a ion managemen and poli ical consul ancy. The Fede al O ice o Mig a ion and Re ugees (BAMF) is a ede al au ho i y wi hin he emi o he Minis y o he In e io (BMI) and he Cen e o Excellence o Asylum, Mig a ion and In eg a ion in Ge many (BAMF 2017). The Socio-Economic Panel (SOEP) is a longi udinal su ey o p i a e households in Ge many based a he Ge man Ins i u e o Economic Resea ch (DIW Be lin). I has been conduc ed annually since 1984. The SOEP has been ecei ing ongoing unding since 2002 om he ede al go e nmen and he s a e o Be lin h ough he Join Science Con e ence. The su ey p o ides in o ma ion on a ious opics such as household composi ion, employmen , heal h, and social a i udes. In 2017, a o al o a ound 38,000 indi idual esponden s in 23,000 households we e in e iewed (Göbel e al. 2018). 4 SOEP Su ey Pape s 605 34 1 In oduc ion The i s wa e o he IAB-BAMF-SOEP Su ey o Re ugees in Ge many was conduc ed in 2016. The da a we e made a ailable o he scien i ic communi y in 2017 (Sample M3 & M4 o he Socio-Economic Panel (SOEP) (K oh e al. 2017)). I is a andom sample o Ge man households ha include adul asylum seeke s, e ugees, and people wi h subsidia y p o ec ion o p o ec ion on humani a ian g ounds who a i ed in Ge many be ween 2013 and Janua y 2016 and we e en e ed in o he cen al egis e o o eigne s (AZR) by Ap il 2016 (mino s we e addi ionally sampled in in June 2016). The su ey has been used ex ensi ely in policy and academic esea ch on he ecen in lux o e ugees o Ge many (B ücke e al. 2017,B ücke e al. 2016,Baie /Siege 2018,Bü mann e al. 2018,B ücke e al. 2019). To compensa e o changes in he unde lying popula ion o e ugees in Ge many in a longi udinal su ey, he IAB-BAMF-SOEP Su ey needed o in eg a e indi iduals who we e no pa o he sampling ame in he ini ial wa e o samples M3 & M4. In wa e 2, 2017, his applied o wo g oups in pa icula : i s , e ugees who a i ed in Ge many be ween Janua y and Decembe 2016 (new a i als) and, second, e ugees who a i ed be ween Janua y 2013 and Janua y 2016 bu who egis e ed in Ge many la e (la e egis- a ion). Sample M5, as he 2017 supplemen a y sample o he IAB-BAMF-SOEP Su ey, co e s hese wo g oups. As Table 1shows, hese wo g oups o new a i als and la e egis a ions co e a ound 435,000 pe sons (mino s included), 290,000 o whom a e adul s1. Wi h his enla gemen , we a e able o ex end he a ge popula ion o he IAB-BAMF-SOEP su ey o pe sons who a i ed in Ge many h ough he yea 2016 and also include la e asylum applican s. 1Please no e ha in he ollowing ables, we do no ely on o als as es ima ed om he sampling ame bu on an AZR e sion ha was compiled in Decembe 2017, as he in eg a ed weigh o all sub-samples M3-M5 should be calcula ed based on he same AZR e sion. The i s wa e o IAB-BAMF-SOEP used an AZR e sion om Janua y 2017. Hence, he second wa e uses o als ha ely on he AZR one yea la e . The e o e, he e is a sligh de ia ion in M5 be ween he o als o he sampling ame and he o als used in he aking p ocess. The e o e, as able 1indica es, we also acknowledge pe sons who do no ha e e ugee s a us. These a e pa icula ly pe sons who had applied o a p olonga ion o hei legal s a us and he e o e possessed empo a ily no legal s a us we conside ed in he sampling. This also holds o indi iduals who had changed ack unde mig a ion law bu had ini ially immig a ed as e ugees. These special cases esul om he ac ha some indi iduals in he a ge popula ion ha e li ed in Ge many o some ime bu changed s a us a some poin . 5 SOEP Su ey Pape s 605 34 a e o a ound 52.1%. Looking a he wo di e en ocus g oups, we see ha he esponse a e in he la e egis a ion sample (56.1%) was sligh ly highe han o he new a i als (47.1%). The 2,915 households we e con ac ed by 30 di e en in e iewe s, each o whom inished 51 household in e iews success ully on a e age (min = 12 & max = 149). 1,256 o he sampled ancho pe sons ei he declined o we e no able o pa icipa e. I is s iking ha in o al, only a ound 11% did no pa icipa e due o a “ha d e usal” o due o ime cons ain s. The o e all esponse a e o he IAB-BAMF-SOEP Su ey o Re ugees was 54.7% i he quali y-neu al d op-ou s a e igno ed. This is an ou s anding esponse a e compa ed o o he sub-samples in he SOEP. Figu e 1shows he esponse a es a he s a e (Lände ) and coun y le els. Gene ally speaking, we ound lowe esponse a es in no heas e n Ge many. The esponse a es on he s a e le el a ied om a ound 30% (Mecklenbu g-Wes e n Pome ania) o 65% pe cen (Ba a ia)2. Table 5: Fieldwo k Resul s Wi h and Wi hou QNA Household Response Ra e M5 O e al (Wi h and wi hou QNA) La e Regis a ion (Wi h and wi hou QNA) New A i als (Wi h and wi hou QNA) Quali y-Neu al D op-Ou Deceased 0.0 (1) - 0.1 (1) - - - Nonexis en /in alid add ess 4.8 (139) - 4.5 (74) - 5.1 (65) - Sub o al 4.8 (140) - 4.6 (75) - 5.1 (65) - Response Full/Pa ial Response 52.1 (1519) 54.7 (1519) 56.1 (914) 58.8 (914) 605 (47.1) 49.6 (605) Sub o al 52.1 (1519) 54.7 (1519) 56.1 (914) 58.8 (914) 605 (47.1) 49.6 (605) Non-Response No Loca able/Accessible 27.7 (808) 29.1 (808) 25.5 (415) 26.7 (415) 30.6 (393) 32.2 (393) Illness o Nu sing Ca e 1.0 (28) 1.0 (28) 0.8 (13) 0.8 (13) 1.2 (15) 1.2 (15) Language P oblems 3.5 (102) 3.7 (102) 2.5 (41) 2.6 (41) 4.8 (61) 5.0 (61) No ime/Re usal 10.7 (313) 11.3 (313) 10.4 (169) 10.9 (169) 11.2 (144) 11.8 (144) O he 0.2 (5) 0.2 (5) 0.2 (3) 0.2 (3) 0.2 (2) 0.2 (2) Sub o al 44.5 (1279) 45.3 (1256) 39.3 (641) 41.2 (641) 47.9 (615) 50.4 (615) To al 100 (2915) 100 (2775) 100 (1630) 100 (1555) 100 (1285) 100 (1220) 4.2.1 Ve sions o he Ques ionnai e in Di e en Languages All ield ma e ials we e p o ided in se en di e en languages, as a signi ican numbe o esponden s we e no p o icien in Ge man a he ime o he su ey (see Table 6). The same languages we e used as in he M3 & M4 sub-samples. A he beginning o each 2In small s a es like B emen and Saa land, no household in e iews we e comple ed. The absence o household in e iews in small s a es is un o una e, bu is due o he ela i ely low a ia ion o PSUs in hose a eas. Fu he , he g oss sample was compa a i ely small in compa ison o M3 & M4. I many esponden s a e also unwilling o pa icipa e o i he add ess quali y in a PSU is low, no in e iews can be conduc ed. 12 SOEP Su ey Pape s 605 34 Response in % 90-100 (12) 80-90 (9) 70-80 (11) 60-70 (19) 50-60 (21) 40-50 (28) 30-40 (16) 20-30 (9) 0-20 (5) no included (272) Figu e 1: Response Ra es a he S a e and Municipal Le el in e iew, one o he languages was chosen. In e iews we e conduc ed using he CAPI (Compu e -Assis ed Pe sonal In e iewing) mode, and he Ge man and chosen language e sions we e displayed side-by-side on he sc een. This allowed language ba ie s o be o e come qui e easily. Mo eo e , Kan a Public p o ided an in e p e e ho line ha could be con ac ed i any di icul ies a ose du ing he in e iew. Table 6: Use o Visual T ansla ions in Ne Sample Visual T ansla ion Pe cen (absolu e) Ge man / English 10.1 (228) Ge man / A abic 66.8 (1,505) Ge man / Fa si 17.4 (392) Ge man / Pash o 0.8 (17) Ge man / U du 2.0 (46) Ge man / Ku manji 2.8 (64) To al 100 (2,252) In addi ion, we p o ided audio iles o deal wi h po en ial illi e acy (see Table 7). As 13 SOEP Su ey Pape s 605 34 he numbe s indica e, mos esponden s did no use he audio iles a all (75%). The in e iewe epo ed ha only 8% used he audio iles wi h e e y ques ion. Table 7: Use o Audio Files in Ne Sample Audio-Files Pe cen (absolu e) Wi h e e y ques ion 8.0 (180) Wi h a ound 2/3 5.0 (113) Wi h a ound hal 4.2 (95) Wi h ewe han hal 8.2 (185) No a all 74.6 (1,679) To al 100 (2,252) Besides language ba ie s, in e iewe s encoun e ed o he di icul ies. A ound 47 % o he ne household sample s ill li ed in public g oup housing (47.4 % public g oup housing and 52.6 % in p i a e housing). Some in e iewe s he e o e needed o i s ob ain pe mission o en e hese acili ies. To do so, i was c ucial ha hey make con ac wi h he local wel a e and o he o ganiza ions ha un he shel e s. 14 SOEP Su ey Pape s 605 34 5 C oss-Sec ional Weigh ing As desc ibed abo e, we assigned a ying sampling p obabili ies o a ge popula ion mem- be s. Design weigh s accoun o hese unequal sampling p obabili ies o households wi hin sample poin s. The wo-s age sampling design ini ially applies equal selec ion p obabil- i ies (PPS, p obabili y p opo ional o size). Only in he subsequen s ep o andomly selec ing seconda y sampling uni s (SSUs) pe sample poin a e unequal sampling ac- o s in oduced ac oss coun ies o o igin, legal s a us, gende , and a ge popula ion (see Table 3in sec ion 3.3). Please no e ha he Cen al Regis e o Fo eigne s lis s indi iduals wi hou any in o - ma ion abou hei household con ex (ma i al s a us o ela i es). Howe e , as wi h all o he o he samples ha make up he Socio-Economic Panel s udy, he IAB-BAMF-SOEP Su ey o Re ugees is designed as a household panel su ey in which all adul household membe s a e in e iewed. Hence, a household wi h wo e ugees, o ins ance, has a highe sampling p obabili y han single households. In o de o de e mine a household’s sampling p obabili y, we he e o e need o assign sample p obabili ies o all membe s o exis ing households who a e pa o he sampling ame, e en hough we did no ini ially sample hese indi iduals as ancho esponden s. To his end, in o ma ion on coun y o o igin, gende , da e o asylum applica ion, and a i al da e is needed o all household membe s. While comple e in o ma ion is a ailable on ancho esponden s, i is no always a ail- able on o he household membe s, o ins ance, because some household membe s did no pa icipa e in an in e iew. In many cases, howe e , p oxy in o ma ion can be ob ained om o he household membe s. Fu he mo e, he in o ma ion on whe he he esponden is iden i ied as being pa o he la e egis a ions o he new a i als can be in e ed om he ancho esponden . I da a was no a ailable om he household con ex , single impu a ion p ocedu es we e used o eplace missing alues wi h plausible and consis en in o ma ion. As desc ibed abo e, we assigned a ying sampling p obabili ies ac oss coun ies o o igin, legal s a us, gende , and a ge sub-popula ion. Combining all he componen s allowed us o calcula e each indi idual’s p obabili y o be sampled in he s udy. A e 15 SOEP Su ey Pape s 605 34 ha ing iden i ied he sampling p obabili y o each household membe , we could calcu- la e household sampling p obabili ies o each o he 1,519 households. I a leas one household membe is sampled, he whole household is pa o he s udy. By knowing indi idual sampling p obabili ies (pi) o each household membe and assuming sampling independen o he household con ex , he sampling p obabili y o a household (phh) de- no es a unc ion o he complemen a y p oduc o he complemen a y indi idual sampling p obabili ies: phh = 1 −Y i (1 −pi) Household weigh s (whh) esul om he in e se household sample p obabili y: whh =1 phh Please no e ha he numbe o households in he a ge popula ion is no known and canno be ex ac ed om he sampling ame, as no household iden i ie is a ailable in he Cen al Regis e o Fo eigne s. Thus, we es ima ed he o al numbe o households based on ou en i e weigh ing and pos -s a i ica ion p ocedu e on pe sonal le el a 210,254 households. 5.1 Non-Response Weigh ing No all sampled households can be con ac ed du ing ieldwo k, and hose ha a e con- ac ed may choose o no pa icipa e in he su ey. Non- esponse may in oduce bias in o es ima es i esponden s di e sys ema ically om non- esponden s. Non- esponse weigh s a e cons uc ed o add ess his issue wi h he aim o minimizing po en ial bias due o non- esponse. O e ecen yea s, non- esponse has been inc easing cons an ly in su eys like he SOEP, and e y ew s udies ha e achie ed esponse a es abo e 40 pe cen (Schnell 2012). The esponse a e in he i s wa e o he IAB-BAMF-SOEP Su ey o Re ugees was oughly 50 pe cen o M3 & M4, whe eas he esponse a e o M5 was sligh ly highe 16 SOEP Su ey Pape s 605 34 (55%). We specula e ha e ugees ha e a s ong mo i a ion o pa icipa e in such su eys because i allows hem o p o ide in o ma ion abou hei cu en si ua ion. Se e al echniques ha e been p oposed o adjus o non- esponse. We used a s a - egy simila o hose used p e iously in he SOEP: p opensi y sco e es ima ion. This equi ed in o ma ion on bo h g oups, esponden s and non- esponden s. Applying logis- ic eg ession analysis, we es ima ed he p opensi y o each household o pa icipa e/no pa icipa e. Non- esponse weigh s we e hen calcula ed by ans e ing he p opensi ies o p obabili ies; he in e se o his p obabili y o each household is he inal household non- esponse weigh (Kim/Kim 2007). 5.1.1 Da a Sou ces and Documen a ion on Va iables The main da a sou ces o es ima ing p opensi y sco es s em om ou sampling ame, he Cen al Regis e o Fo eigne s, as well as an in e iewe ques ionnai e. In addi ion, we made use o ex e nal da abases a he coun y (INKAR, INKAR 2018) and municipali y le el (Regionalda enbank, Die Regionalda enbank Deu schland 2018). Da a Sou ces Used in he Modelling o Non-Response: Municipali y: Fede al S a is ical O ice: The "Regionalda enbank Deu schland" (Regional Da abase Ge many) con ains in o ma ion on di e en le els o geog aphic uni s ha was collec ed as pa o a join p ojec o he Fede al S a is ical O ice and i s sub-na ional coun e pa s a he s a e (Lände ) le el. We made use o a iables compiled a coun y and neighbo hood le els o he analysis o non- esponse. Coun y: Fede al S a is ical O ice (INKAR): The da abase "Indika o en und Ka en zu Raum und S ad en wicklung in Deu schland und in Eu opa" (INKAR) p o ides in o ma ion on egional economic ac i i y (e.g., p ope y p ices, household income, wel a e bene i s) as well as popula ion cha ac e is ics (e.g., educa ional da a) o he di e en egional en i ies. F om his sou ce, he in o ma ion is a ailable a he coun y le el and was compiled in 2014 and 2015. Cen al Regis e o Fo eigne s: Addi ionally, Kan a Public anonymized he g oss sample, and we used in o ma ion om he AZR ega ding he ancho esponden s. We we e able o use he na ionali y, age, and asylum s a us a he ime o sampling o es ima e a p opensi y sco e. a) Asylum s a us a he ime o sampling b) Coun y o o igin c) Gende 17 SOEP Su ey Pape s 605 34 d) Time o a i al in Ge many e) Age In e iewe Ques ionnai e: Fu he mo e, in e iewe s we e eques ed o ill ou a ques ionnai e on each household hey had a emp ed o con ac . This allowed us o gain a pic u e o he household’s physical su oundings and he in e iewe ’s eelings abou hese su oundings. A documen a ion o all he a iables used is p o ided in he appendix (see Table 11). As was he case wi h he es ima ion in he p e ious sub-samples, we ocused on indi- idual cha ac e is ics, because e ugees a e andomly dis ibu ed ac oss Ge many using he “Königs eine Schlüssel”. We he e o e assume ha con ex in o ma ion has a less sys ema ic impac on esponse beha io . 5.1.2 Mul iple Impu a ion and Da a Coding In o de o gene a e a non- esponse weigh , we need comple e in o ma ion on all espon- den s and non- esponden s, as indi iduals wi h missing alues would o he wise d op ou o any mul i a ia e es ima ion. To a oid such exclusion, we impu ed missing da a using he “Mul iple Impu a ion by Chained Equa ions” (Roys on 2009) me hodology implemen ed in S a a 14. We wo ked wi h en di e en p edic ions o deal adequa ely wi h he un- ce ain y in he impu a ion p ocess. Fo he inal es ima ion o he non- esponse weigh , we selec ed one o he en p edic ions andomly. Thus only one non- esponse weigh was es ima ed. Some a iables we e ecoded and condensed. Me ic a iables we e ca ego ized, e- sul ing in h ee dis inc ca ego ies, wi h he middle ca ego y as a e e ence. O dinal indi- ca o s we e condensed o a maximum o i e ca ego ies, and each ca ego y was included as a dummy a iable wi h one ca ego y as e e ence. The same p ocedu e was used wi h nominal a iables. Using ca ego ized a iables and hei espec i e bina y indica o s in eg ession has se e al ad an ages. Non-linea e ec s we e con olled o because indi- idual pa ame e s we e es ima ed o each g oup. Also, he ca ego iza ion p e en ed he es ima ion o ex eme p obabili ies e y close o ze o o one because o single ou lie s on a a iable. This was necessa y in o de no o in la e he es ima ed weigh s inapp op ia ely 18 SOEP Su ey Pape s 605 34 ( o an example, see K oh e al. 2015). 5.1.3 Resul s and Calcula ion o Non-Response Weigh s In o de o es ima e household esponse p opensi ies, logis ic eg ession analysis was used. In his p ocess, we accoun ed o mul iple impu a ions. We used obus s anda d e o s, clus e ed a PSU le el, o accoun o a iance due o p ima y sampling uni s om which ancho esponden s we e sampled. The ull sample o 2,775 eligible households was used in e e y es ima ion. Fi s , a ull model was es ima ed using all o he a iables a hand. Then a educed model was es ima ed. We es ima ed a Pseudo-R2o .044 o he inal non- esponse model. I is aluable o ake a close look a he ac o s in luencing he chance o esponding o he in e iew inqui y. Ba a ia Feeling Secu e Medium U ban A ea Only One Immig a ion o ice in Poin En y 2013 Low Wel a e Expendi u es o Re ugees E i ea, Somalia Tole a ion O he Na ionali y Balkan In e cep P edic o -2 -1 0 1 2 Coe icien and 95%-c.i. Reduced Model Figu e 2: Reduced Non-Response Model 1) Coun y o O igin We ound an e ec o he esponden ’s coun y o o igin. Especially indi iduals 19 SOEP Su ey Pape s 605 34 om E i ea, Somalia, and he Balkans had a highe isk o no esponding o he in e iewe inqui y. 2) Asylum S a us Rega ding he asylum applica ion p ocess, we ound an e ec only o hose espon- den s who ha e a pos ponemen o emo al. They ha e a sligh ly highe chance o no esponding o ou inqui y. 3) In e iewe Assessmen When he in e iewe el sa e in he household’s neighbo hood su oundings, we measu ed a highe chance o he esponden aking pa in he in e iew. All he o he a iables desc ibing he in e iewe ’s eelings and assessmen s we e no signi - ican . The e o e, we can conclude ha in e iewe e ec s a e ela i ely low. 4) Sampling Uni s Rega ding he sampling uni s, we measu ed a highe p obabili y o pa icipa ion o e ugees who a e pa o a p ima y sampling uni in which he e is only one immig a ion o ice. 5) Cha ac e is ics o he Household’s Neighbo hood Su oundings When he household is loca ed in a medium-sized u ban a ea (be ween 150 and 100 inhabi an s pe km2), he e is a highe chance o aking pa in he in e iew. 6) Time o A i al People who en e ed Ge many in 2013 had he highes chance o aking pa in an in e iew compa ed o o he en y ime poin s. 7) Cu en Place o Residence On a s a e (Lände ) le el, we ound a highe chance o people ha li e in Ba a ia. 8) Con ex In o ma ion Fu he mo e, in a eas whe e public wel a e expendi u es on e ugees a e low, he esponse a e is lowe . In his ega d, wel a e expendi u es ep esen p oxy in o - ma ion o he o al numbe o e ugees and asylum seeke s li ing in an a ea. 20 SOEP Su ey Pape s 605 34 The educed non- esponse model was used o p edic household pa icipa ion p ob- abili ies. The in e ses o hese p obabili ies unc ion as non- esponse weigh s. Table 8 shows he summa y s a is ics o he weigh s. Table 8: Cha ac e is ics o Household Non-Response Weigh - M5 Quan iles Min 10% 25% 50% 75% 90% Max Mean SD N Non- esponse Weigh 1.3 1.4 1.6 1.7 2.0 2.3 5.3 1.8 0.5 1,519 We es ima ed a non- esponse weigh o all pa icipa ing households. Due o ou esponse a e o a ound 54 pe cen , he mean o he weigh is close o wo. 5.2 Pos -S a i ica ion In addi ion o he a o emen ioned combina ion o design and non- esponse weigh s, he weigh s we e co ec ed o mee known cell dis ibu ions, o ma ginal o als. These we e de i ed om he Cen al Regis e o Fo eigne s solely o his pu pose. To coun he o als used in he p ocedu e desc ibed below, we used he Decembe 2017 e sion o he Cen al Regis e o Fo eigne s. Following he pos -s a i ica ion p ocedu es used p e iously in he SOEP, we used he echnique o “i e a i e p opo ional i ing” (Deming/S ephan 1944), also e e ed o as “ aking”. This is a special ype o pos -s a i ica ion and is usually used “when pos - s a a a e o med using mo e han one a iable, bu only he ma ginal popula ion o als a e known” (Loh 2010). Table 9shows a lis o cha ac e is ics ha we e used in he aking p ocess o he indi idual weigh : coun y o o igin, sex, age as a g ouped a iable, he da e o a i al in Ge many, and a a iable o he egion in which he household li es. These a iables apply o he aking p ocess a he indi idual le el. O en in household su eys, no all household membe s ill ou a ques ionnai e. These indi iduals ha e also been pa o he weigh ing s a egy so a , meaning ha hey we e assigned a design weigh and a household non- esponse weigh as well. Due o he ac ha hese indi iduals a e no included in he ne sample, we ha e o co ec o his 21 SOEP Su ey Pape s 605 34