Germany Going Postal? Comparing Postal and Election Day Voters in the 2017 German Federal Election

Wagne , Aiko; Lich eblau, Josephine A icle — Published Ve sion Ge many Going Pos al? Compa ing Pos al and Elec ion Day Vo e s in he 2017 Ge man Fede al Elec ion Ge man Poli ics P o ided in Coope a ion wi h: WZB Be lin Social Science Cen e Sugges ed Ci a ion: Wagne , Aiko; Lich eblau, Josephine (2022) : Ge many Going Pos al? Compa ing Pos al and Elec ion Day Vo e s in he 2017 Ge man Fede al Elec ion, Ge man Poli ics, ISSN 1743-8993, Taylo & F ancis, London, Vol. 31, Iss. 4, pp. 602-625, h ps://doi.o g/10.1080/09644008.2020.1759553 This Ve sion is a ailable a : h ps://hdl.handle.ne /10419/218878 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/4.0/ Ge many Going Pos al? Compa ing Pos al and Elec ion Day Vo e s in he 2017 Ge man Fede al Elec ion* AIKO WAGNER and JOSEPHINE LICHTEBLAU The p opo ion o o e s cas ing hei o es by mail days o e en weeks be o e elec- ion day has isen om ini ially fi e pe cen o nea ly hi y pe cen in he 2017 Ge man ede al elec ion. A he same ime, he di e ences in elec ion esul s be ween pos al and elec ion day o e s a e inc easing, eaching a eco d a he Bun- des ag elec ion in 2017. Agains his backg ound, i seems pa icula ly su p ising ha he e a e only a small numbe o empi ical s udies on pos al o ing in Ge many. Fo an e alua ion (o he de elopmen ) o pos al o ing, i is c ucial o in es iga e he ac o s esponsible o he di e en elec ion esul s be ween he wo o e g oups. This equi es mo e p ecise knowledge o pos al o e s and how hey di e om o e s who cas hei ballo s on elec ion day. Using da a om he Ge man Longi udinal Elec ion S udy (GLES), we in es iga e in how a composi ion e ec s –di e ences be ween he o e ypes in he composi ion o ac o s po en ially ele an o pa y choice –and associa ion e ec s –di e ences ega ding hei o e unc ions o c i e ia applied o o e choice –can help o explain he di e en elec- o al ou comes. We conclude ha he o e ypes do no di e subs an ially om each o he wi h ega ds o hei decision c i e ia bu ha hei composi ion di e s wi h ega ds o ce ain cha ac e is ics ele an o pa y choice. INTRODUCTION In o de o s eng hen he p inciple o uni e sal su age, pos al o ing was in oduced in Ge many a he 1957 Bundes ag elec ions. Vo e s unable o cas hei o e a he polling s a ion on elec ion day should be enabled o o e by mail in ad ance. One o he main goals associa ed wi h he in oduc ion o pos al o ing (and la e o abandoning he necessi y o s a e easons when applying o pos al ballo documen s in 2008), was hus o inc ease o e u nou . Since hen, pos al o ing in Ge many has de eloped in o a ue success s o y. Since eunifica ion in 1990, he p opo ion o pos al o e s has isen s eadily, eaching a eco d high in he las Bundes ag elec ion in 2017, in which almos one hi d o all o es we e cas by mail. A he same ime, howe e , *An ea lie d a o his pape was p esen ed a he 2018 con en ion o he Ge man Poli ical Science Associ- a ion in F ank u . Ge man Poli ics, Vol.31, No.4, 2022, pp.602–625 h ps://doi.o g/10.1080/09644008.2020.1759553 © 2020 The Au ho (s). Published by In o ma UK Limi ed, ading as Taylo & F ancis G oup This is an Open Access a icle dis ibu ed unde he e ms o he C ea i e Commons A ibu ion License (h p://c ea i ecommons.o g/licenses/by/4.0/), which pe mi s un es ic ed use, dis ibu ion, and ep oduc ion in any medium, p o ided he o iginal wo k is p ope ly ci ed. di e ences in he elec ion esul s o pos al o e s and hose o ballo box o e s a e inc easing. In his espec , oo, he 2017 Bundes ag elec ions p esen a eco d, he di e - ences in he pa ies’sha es o he second o es be ween he wo elec o al g oups ha ing been g ea e han in any p e ious Bundes ag elec ion. Agains his backg ound, i is a he su p ising ha he e a e ha dly any empi ical s udies examining pos al o ing in Ge many. In in e na ional esea ch, pos al o ing is commonly analysed as one o m o ea ly o ing o con enience o ing. In his esea ch s and, mos s udies deal wi h no ma i e expec a ions asking whe he ea ly o ing in ac leads o highe pa ici- pa ion a es o educes social selec i i y in elec o al pa icipa ion and, hus, helps o educe unde ep esen a ion o ce ain g oups’p e e ences. Bu how can we e alua e pos al o ing in Ge many? The answe o his ques ion depends la gely on he ac o s esponsible o he di e en elec ion esul s be ween pos al and elec ion day o e s. The objec i e o his a icle, he e o e, is o de elop and es an explana o y model o he di e ences in he pa ies’ o e sha es be ween he wo g oups o o e s. Explaining hese beha iou al di e ences equi es mo e p ecise knowledge o pos al o e s and how hey di e om elec ion day o e s. In p inciple, he e a e wo ypes o e ec s ha can cause he di e ences in o ing beha iou o wo aspec s in which o e ypes a e di e en . Fi s , composi ion e ec s, i.e. di e ences in he composi ion o po en ially decision- ele an ea u es, can cause di e en o ing decisions. Second, he li e a u e on o e he e ogenei y sugges s ha di e ences can also be he esul o di e en decision c i e ia o o e unc ions. In his line o hinking he di e ences in pa y s eng hs can be a ibu ed o so-called associa ion e ec s. To e alua e pos al o ing in Ge many, we hus ask wha ole composi ion and associa ion e ec s espec i ely play o explaining di e ences in pa y choice be ween pos al and elec ion day o e s. Fi s , we b iefly discuss he cu en s a e o esea ch on ea ly o ing in Ge many and, based on ha , conside which c i e ia can be used o e alua e pos al o ing in Ge many in a no ma i e sense. In his con ex , we p esen ou explana o y model o he di e en elec ion esul s in mo e de ail. We use da a om he pos -elec ion su ey o he Ge man Longi udinal Elec ion S udy 2017 o es ou explana o y model. In a fi s s ep, we in es iga e possible composi ion e ec s. To ha end, we compa e he wo ypes o o e s in e ms o ac o s ha influence bo h o ing ype and pa y choice in o de o shed ligh on he ques ion o whe he pos al o e s a e simply a di e en subse o he elec o a e han elec ion day o e s. Second, we examine which pa ies benefi om which ac o s associa ed wi h an inc eased p ob- abili y o pos al o ing. In a hi d s ep, we analyse and compa e he decision c i e ia o o e unc ions o he wo g oups o o e s o see whe he associa ion e ec s can help explain he di e en elec ion esul s in 2017. The esul s o ou analyses sugges ha pos al and elec ion day o e s do no di e subs an ially om each o he wi h ega ds o he c i e ia hey apply o o e choice. The di e ences in o e sha es be ween he wo g oups seem composi ional in na u e: As some sociodemog aphic cha ac e is ics a e ela ed o bo h an inc eased p obabili y o pos al o ing and o an inc eased p obabili y o o ing o pa icula pa ies, hese pa ies pe o m be e among pos al han elec ion day o e s. GERMANY GOING POSTAL? 603 THE SIGNIFICANCE OF POSTAL VOTING IN GERMANY, STATE OF RESEARCH AND THEORETICAL CONSIDERATIONS The significance o pos al o ing in Ge many and o Ge man o e s can be illus a ed by looking a how i has been used since i s in oduc ion in 1957. Figu e 1 shows his de elopmen . S a ing a fi e pe cen in 1957, he p opo ion o o e s who cas hei o es by mail be o e elec ion day 1 (black line) ose con inuously un il 1980, ell sligh ly un il 1990, and has isen again in e e y elec ion e e since. In 2008, he necessi y o jus i y he use o pos al o ing was abandoned. A e wa ds, he p opo ion o pos al o e s ose o one-fi h in he 2009 Bundes ag elec ions, hen again o one-qua e in 2013, and in he 2017 Bundes ag elec ions nea ly 30 pe cen o o e s used he pos al o e op ion. As a esul , pos al o ing should by now be conside ed an essen ial pa o Ge man elec o al law. Ano he in e es ing finding becomes appa en when com- pa ing he de elopmen o he p opo ion o pos al o e s among all o e s wi h he de elopmen o he p opo ion among all eligible o e s ep esen ed by he g ey line. The wo lines emain close un il eunifica ion bu di e ge a e wa ds. Be ween 2002 and 2009, o example, he p opo ion o pos al o e s among eligible o e s emained s able a a ound 15 pe cen , i.e. he e was nei he a decline in pa icipa ion no an inc ease in pa icipa ion among pos al o e s. Howe e , he e was an inc ease in he p o- po ion o all o e s, as o e u nou had allen. Thus, he compa a i ely weake inc ease in he g ey line indica es ha o e u nou among pos al o e s has inc eased since 1990 ela i e o o e u nou a elec ion day. FIGURE 1 DEVELOPMENT OF POSTAL VOTING IN GERMANY Sou ce: Bundeswahllei e (2017b). 604 GERMAN POLITICS The significance o pos al o ing in Ge many in gene al and o he inc ease in usage in pa icula becomes e en clea e when compa ing elec ion esul s be ween elec ion day o e s and pos al o e s. Figu e 2 shows he di e ence index o he dis ibu ions o lis o es be ween pos al and elec ion day o e s since eunifica ion. Fo each Bun- des ag elec ion, he absolu e di e ences in elec ion esul s be ween pos al and elec ion day o e s we e added up o all pa ies and subsequen ly di ided by wo (since he gains o one pa y a e he losses o he o he ). While he wo elec o al g oups became mo e dissimila wi h ega d o hei o e choices om elec ion o elec ion un il 1998, hey con e ged in he 2002 elec ion. Since hen, howe e , he e has again been an inc ease in he di e si y index, wi h –as in he case o usage o he pos al o ing op ion – he highes alue in he 2017 Bundes ag elec ions. This means ha , a leas since he 2002 Bundes ag elec ions, pos al and elec ion day o e s inc easingly di e in hei pa y p e e ences, wi h he di e ences in he 2017 Bundes ag elec ions being e en la ge han in 1998 and hus ep esen ing a new maximum. Consequen ly, in Figu e 3, we ake a close look a he 2017 dis ibu ion o o es as well as a which pa ies we e a ec ed by he di e ences in o ing decisions be ween pos al o e s and ballo box o e s. Fo mos pa ies, he p opo ions o second o es a e a he simila wi h di e ences be ween pos al and elec ion day o e s usually below wo pe cen age poin s. Howe e , in he case o he Ch is ian democ a s (CDU/ CSU) and he igh -wing populis A D, mo e ele an di e ences can be disce ned. While he CDU/CSU won 36.4 pe cen o pos al o e s, i only gained 31.5 pe cen among elec ion day o e s (5 pe cen age poin s less). The opposi e is ue o he A D: FIGURE 2 DIFFERENCES IN THE VOTING BEHAVIOUR OF POSTAL AND BALLOT BOX VOTERS SINCE 1990 Sou ce: own calcula ions based on da a om he Bundeswahllei e (2015,2017a). GERMANY GOING POSTAL? 605 i only ecei ed sligh ly less han en pe cen o he second o es o pos al o e s bu was able o secu e almos 14 pe cen o he o es on elec ion day –a plus o 4.3 pe cen age poin s, co esponding o abou 45 pe cen . The CDU/CSU hus benefi ed om he pos al o ing op ion. Albei a a lowe le el, he same is ue o he FDP and he G eens. The SPD and he Le we e sligh ly mo e success ul among elec ion day o e s. 2 Ano he poin o conside when assessing ( he significance o ) pos al o ing is he iming o he elec ion decision. Ob iously, people who cas hei o e weeks be o e he elec ion canno ake in o accoun he de elopmen s o he las days o he elec ion campaign. E en s sho ly be o e he elec ion, such as changes in he poli ical si ua ion, e.g. new in o ma ion om polls and ela ed s a egic o ing calcula ions o mobilisa ion s a egies o he pa ies, can hus no longe influence pos al o e s’elec ion decision (c . Sheppa d and Beau ega d 2018). Based on hose conside a ions, Figu e 4 shows he ex en o which he wo o e g oups di e wi h ega d o he ime o he elec ion decision in 2017. Fo almos wo- hi ds o bo h g oups, he elec ion decision was made a long ime o mon hs ago –well be o e a pos al o e was possible. Almos a qua e o elec ion day o e s made hei decision in he las ew days be o e o on elec ion day i sel , whe eas his was he case o less han 14 pe cen o pos al o e s. Con e sely, significan ly mo e pos al o e s han elec ion day o e s made hei decision in he weeks be o e he elec ion (24 compa ed o 16 pe cen ). On a e age, elec ion day o e s decide la e han pos al o e s which pa y o o e o . The o me a e, hus, mo e suscep ible FIGURE 3 SECOND VOTE SHARES BY VOTER TYPE IN THE 2017 BUNDESTAG ELECTIONS Sou ce: Bundeswahllei e (2017a). 606 GERMAN POLITICS o e en s in he las days o he campaign whe eas hose e en s canno ha e an impac on a ballo al eady cas , o cou se. In sho , a la ge (and g owing) p opo ion o o e s in Ge many make use o he pos al o ing op ion. They do no only di e om ballo box o e s in e ms o hei o ing beha iou , bu also end o a i e a hei o e choice ea lie . While pos al o e s a e an empi ically ele an phenomenon, he e has been su p isingly li le esea ch on his segmen o he elec o a e in Ge many so a . In e na ionally, he e is an es ablished esea ch s and on op ions ha acili a e he ac o o ing o educe i s cos s by enabling ci izens o cas hei o e in ad ance and ou side he polling s a ion –gene ally e e ed o as ea ly o con enience o ing. No ma i e a gumen s o such o ing op ions usually e e o he assump ion ha hey inc ease o e u nou and hus s eng hen he equali y p inciple o democ acy, since ci izens who, o example due o employmen o lack o mobili y canno cas hei o e a he polling s a ion on elec ion day, a e enabled o do so by o he means. Fu he mo e, ad oca es poin ou ha con enience o ing op ions can coun e ac he social selec i i y o elec o al pa ici- pa ion, since lowe cos s would encou age hose segmen s o he popula ion ha would o he wise end o s ay away om he elec ion (G onke e al. 2008). Empi ical s udies es ing hese assump ions sugges ha he in oduc ion o absen ee and ad ance o ing does indeed inc ease o e u nou a he na ional and sub-na ional le el (e.g. Ga ne 2018; Goodman and S okes 2018; Luechinge , Rosinge , and S u ze 2007; Magleby 1987; Sou hwell 2000). So a , howe e , he e is no e idence ha con enience FIGURE 4 TIME OF THE ELECTION DECISION BY VOTER TYPE Sou ce: GLES pos -elec ion c oss-sec ion su ey 2017 (Roß eu sche e al., ‘Pos -Elec ion C oss Sec ion,’2018). GERMANY GOING POSTAL? 607 o ing educes he social selec i i y o o ing. Some s udies find no o only mino di e ences in he composi ion o p esence and con enience o e s wi h ega d o pa - icipa ion- ele an socio-demog aphic cha ac e is ics (McAllis e and Mulle 2018; Neeley and Richa dson 2001). O he s udies show ha absen ee o ad ance o ing op ions a e inc easingly used by he socio-economically be e -o , mo e highly edu- ca ed and poli ically in e es ed people (Dyck and Gimpel 2005; Elle mann 2004; Gieble 2014; Lich eblau and Wagne 2019; McAllis e and Mulle 2018). Such obse - a ions sugges ha con enience o ing migh e en inc ease a he han dec ease social selec i i y (Be insky 2005), implying ha hey migh in ac ail o achie e he desi ed no ma i e goals. Bu how easonable a e he gene al no ma i e expec a ions ega ding pos al o ing in he con ex o Ge many? Due o he compa a i ely s ic equi emen s o applying o and filling in pos al ballo documen s, i does no seem plausible o expec a com- pensa o y e ec agains he social selec i i y o o e u nou . No su p isingly, Elle - mann (2004) concludes in he s udy o he 2002 Bundes ag elec ions ha pos al o ing is mos widesp ead among people wi h high o mal educa ion, high p o essional s a us (ci il se an s and he sel -employed) and om u ban esiden ial a eas. Gieble (2014) comes o simila conclusions in his s udy on pos al o e s in he 2013 ede al elec ions. Fo a no ma i e e alua ion o pos al o ing in Ge many, i is necessa y o ask how he di e en elec ion esul s o pos al and ballo box o e s come abou . 3 Wha is he di e ence be ween pos al o e s and ballo box o e s in Ge many? Figu e 5 illus a es he explana o y model o di e ences in elec ion esul s be ween pos al o e s and elec- ion day o e s which o ms he basis o his pape . As we know om Figu es 2 and 3, he elec ion esul s a e unequal. On he one hand, his can be caused by composi ion e ec s o o e cha ac e is ics; his is why knowledge o possible di e ences be ween o e g oups wi h ega d o he ac o s influencing he elec ion decision is el- e an : I di e ences can be ound in he cha ac e is ics ha de e mine pa y choice, hen di e en elec ion esul s a e ine i able. I , o example, people wi h a mo e le is FIGURE 5 EXPLANATORY MODEL FOR DIFFERENT ELECTION RESULTS OF POSTAL AND ELECTION DAY VOTERS 608 GERMAN POLITICS poli ical leaning inc easingly used pos al o ing, i would no be su p ising i le -wing pa ies sco ed be e among pos al o e s. On he o he hand, esea ch on o e he e ogenei y shows ha ci izens use di e en decision-making s a egies: While o some pe sons economic policy issues can be deci- si e o hei o e, o he s a e mo e s ongly guided by hei assessmen o he leading poli icians. Thus, o e s wi h iden ical p e e ences migh in he end suppo di e en pa ies (Blumens iel and Plischke 2015; Ri e s 1988). In his line o esea ch, Ba le (2005) showed o he B i ish case ha mo e poli ically sophis ica ed o e s ely mo e s ongly on hei e alua ion o policy issues han on he e alua ion o he poli ical pe sonnel. The c ucial poin in ligh o ou ques ion is hus ha o e g oups ha e he - e ogeneous mo i es o o ing, in o he wo ds, ha hei o e unc ions a y. Acco d- ingly, i seems plausible o ask whe he he di e en pa y s eng hs be ween pos al and elec ion day o e s (also) esul om hem applying di e en decision c i e ia o o e unc ions. The e o e, in he ollowing we will also ask o wha ex en he di e ences in pa y choice can be a ibu ed o associa ion e ec s. These associa ion e ec s a e shown in Figu e 5 by he e ical a ows poin ing upwa ds. E en i he dis ibu ion o o e cha ac e is ics ha a e po en ially ele an o pa y choice was iden ical be ween elec ion day and pos al o e s, he e ogeneous decision c i e ia could help explain he di e en elec ion esul s in he wo g oups. Re e ing o he example abo e, le us assume ha he le - igh a i udes be ween he wo g oups do no di e subs an ially and ha poli ical a i udes a e gene ally simila . I pos al o e s ely mo e s ongly on long- e m, s able ideological le - igh posi ions o hei o e choice, while o elec ion day o e s, he e alua ion o op candida es is mo e ele an , his migh lead o di e ences in pa y s eng hs. In he end, i is also possible ha bo h he o e unc ions and he dis ibu ion o cha ac e is ics a y be ween he wo g oups o o e s. Conside ing he di e ences be ween elec ion day and pos al o e s ega ding he iming o hei o e choice (Figu e 4), i seems plausible ha hose di e ences a e a po en ial sou ce o bo h composi ional as well as associa ional di e ences. I e en s sho ly be o e elec ion day, e.g. poli ical scandals, d ama ically change he e alua ion o he incumben , we should expec composi ional di e ences ega ding he e alua ion o he incumben –be e e alua ions o pos al o e s as many o hem al eady decided weeks be o e he elec ion and wo se e alua ions o elec ion day o e s many o whom decided only sho ly be o e he elec ion day –leading o di e ences in o e sha es be ween he wo g oups. I majo ex e nal e en s, such as a nuclea disas e , happen in he las days be o e he elec ion and change he ele ance o ce ain mo i es o elec- o al beha iou , we should disco e di e ences in he associa ion e ec s –in his case g ea e ele ance o a i udes on nuclea powe o elec ion day o e s han o pos al o e s who al eady made hei choice. Consequen ly, he iming o he o e choice migh ha e a significan impac on he dis ibu ion o ele an cha ac e is ics and on he o e unc ion o pos al and elec ion day o e s. 4 DATA AND METHOD To in es iga e he causes o di e en elec ion esul s be ween pos al and elec ion day o e s, we use da a om he pos -elec ion c oss-sec ion su ey 2017 o he Ge man GERMANY GOING POSTAL? 609 models also e eal simila a e age ma ginal e ec s o ou fi e p edic o s o bo h o e g oups (see Figu e A1 and Table A5 in he Appendix C). The only di e ence conce ns he ele ance o e ospec i e go e nmen pe o mance e alua ions: While i eaches he usual fi e-pe cen significance h eshold o elec ion day o e s, i does no o pos al o e s (wi h simila poin es ima es o he e ec ’s s eng h). The di e ence is, howe e , (a) qui e small and (b) possibly d i en by he lowe numbe o pos al o e s (and he esul ing la ge s anda d e o s), which is why i should no be o e - in e p e ed. The e o e, ou conclusion wi h ega d o he di e ences in he decision c i- e ia is nega i e –pos al and elec ion day o e s base hei pa y choice on a simila se o c i e ia. CONCLUSION AND OUTLOOK The significance o pos al o ing in Ge many has clea ly inc eased in ecen decades. By now, almos a hi d o he o e s cas hei ballo be o e elec ion day. Mo eo e , he di e ences in elec o al beha iou be ween pos al o e s and elec ion day o e s a e con- side able wi h ega ds o some pa ies as he sha e o second o es di e s by up o fi e pe cen age poin s be ween he o e g oups. Agains his backg ound, i seems pa icu- la ly su p ising ha he e a e only ew empi ical-analy ical s udies on pos al o ing in Ge many, making he pos al o e a la gely unknown being. The in e na ional li e a u e TABLE 4 RESULTS OF CONDITIONAL LOGIT MODELS FOR VOTE CHOICE (LOGIT COEFFICIENTS) Vo e choices PID 1.75*** (0.17) Le /Righ dis ance −0.04*** (0.01) E alua ion op candida es 0.34*** (0.03) P oblem sol ing compe ence 0.76*** (0.11) E alua ion go e nmen pe o mance 0.07*** (0.02) Pos al o e x PID −0.14 (0.31) Le /Righ dis ance −0.02 (0.02) E alua ion op candida es 0.11 (0.07) P oblem sol ing compe ence 0.19 (0.21) E alua ion go e nmen pe o mance −0.03 (0.04) Obse a ions 8,401 Pseudo R-squa ed 0.47 No e: S anda d e o s in pa en heses. ***p< 0.001, **p< 0.01, *p< 0.05. 616 GERMAN POLITICS on ea ly o ing ocuses p ima ily on possible e ec s on o e u nou –and finds mixed esul s. As a as we a e awa e, he e is ha dly any esea ch dealing wi h he ques ion p esen ed in his pape as o how he di e ences in o e sha es be ween pos al and elec- ion day o e s can be explained. Fo his pu pose, we examined he wo ypes o o e s wi h espec o bo h, po en ially di e en cha ac e is ics (composi ion e ec s) and po en ially di e en mo i es o pa y choice (associa ion e ec s). Based on su ey da a om he GLES, we conclude ha pos al and elec ion day o e s do no di e sub- s an ially om each o he wi h ega ds o he c i e ia hey apply o o e choice. The o e unc ions o bo h g oups ha dly a y, hence, associa ion e ec s canno con ibu e o an explana ion o he di e en elec ion ou comes. Howe e , he composi ion o he o e g oups wi h ega ds o ce ain cha ac e is ics di e s. Among pos al o e s, Wes Ge mans, economically be e -o , sel -employed people as well as pensione s and pupils/s uden s a e o e ep esen ed. Since hese cha ac e is ics a e also ela ed o pa y choice, we can conclude ha o some ex en he di e ences in o ing beha iou s em om hose composi ional di e ences. The poli ical implica ions o he inc eased use o he pos al o ing op ion a e clea and se ious. I pa ies and candida es a e no longe able o win o e a la ge pa o he elec o a e in he las days o weeks o he elec ion campaign, hey need o add ess hose o e s ea lie . Fo his eason, among o he s, a ge ing becomes mo e demanding o pa ies in Ge many. In addi ion, pos al o e s a e no able o conside he e en s o he final phase o he elec ion campaign o hei o e choice. While his migh no ha e been ele an o explaining he di e ences in second o e sha es be ween elec ion day and pos al o e s in he Ge man pa liamen a y elec ions in 2017, he ea lie iming o he o e choice o he la e se s limi s o sho - e m e ospec i e o ing in case o e.g. poli ical scandals du ing he las days o he campaign. 8 Mo eo e , he simila i y o o e unc ions o pos al and elec ion day o e s implies ha he ype o ballo used did no change he o e all elec ion esul : As bo h o e ypes elied on a simila se o e alua ion c i e ia, o ing beha iou i sel seems no o ha e been a ec ed by pos al o ing. Ra he , we obse e a sel -selec ion o ce ain g oups o ci izens in o absen ee o ing, bu he ype o ballo i sel does no impac pa y choice. On he basis o ou p elimina y esul s –especially conside ing he mode a e expla- na o y powe o ou model p edic ing he o e ype –we see need o u he esea ch in h ee di ec ions: Fi s ly, exis ing s udies analysing he easons why people decide o o e by mail ha e p ima ily ocused on he ce ain y o o e choice –people wi h fixed, s ong pa y p e e ences can cas hei o es wi hou any u he need o in o - ma ion p o ided by he elec ion campaign. Howe e , many elec ion day o e s a e ce ain abou hei decision mon hs be o e he elec ion akes place, oo. This may help explain he (in some cases) small di e ences be ween g oups (bo h ega ding hei cha ac- e is ics and hei o e unc ions). He e, u he di e en ia ion is needed in u u e esea ch. Secondly, aspec s o mobili y, such as he dis ance o he polling s a ion in combina ion wi h a pe son’s physical condi ion could no be adequa ely analysed so a . Simila ly, aspec s o con enience, such as he use o pos al o ing due o possible spon aneous holi- days o he a oidance o po en ial collisions o planned ac i i ies, could no be in es i- ga ed on he basis o exis ing da a. Thi dly, he causal mechanisms a e no equally ob ious o canno be su ficien ly heo ised in all cases. Ques ions conce ning he GERMANY GOING POSTAL? 617 easons as o why, o example, Wes Ge mans use pos al o ing much mo e equen ly han Eas Ge mans, ha e no been a he cen e o p e ious s udies. In iew o he no ma- i e, p ac ical-poli ical and poli ical-scien ific ele ance o he subjec , i is wo h opening up his field o esea ch sys ema ically along hese sugges ed lines. ACKNOWLEDGEMENTS We would like o hank all pa icipan s o hei eedback. Fu he mo e, we a e g a e ul o A ne Ca s ens and Lena Röllicke o hei help p epa ing he manusc ip and o he anonymous e iewe s o hei use ul sugges ions o imp o e he pape . DISCLOSURE STATEMENT No po en ial conflic o in e es was epo ed by he au ho (s). FUNDING This wo k was suppo ed by Deu sche Fo schungsgemeinscha : [g an numbe 90948644]. ABOUT THE AUTHORS Aiko Wagne is a esea ch ellow a he WZB Be lin Social Science Cen e in he depa men ‘Democ acy & Democ a iza ion’. He is a membe o he Ge man Longi udinal Elec ion S udy (GLES). Josephine Lich eblau is a esea ch ellow a he WZB Be lin Social Science Cen e in he depa men ‘Democ acy & Democ a iza ion’. She is associa ed wi h he Ge man Longi udinal Elec ion S udy (GLES). NOTES 1. Following he Fede al Re u ning O fice (Bundeswahllei e ), we subsume all ypes o ea ly o ing as pos al o ing. 2. The di e ences in second o e sha es be ween pos al and elec ion day o e s in he sample a e based on a e e y simila o hose based on he o ficial elec ion esul s epo ed by he Bundeswahllei e (c . Appendix A). 3. While Elle mann (2004) only men ions he di e ences in elec ion esul s be ween pos al and elec ion day o e s, Gieble (2014) concludes ha he socio-demog aphic di e ences he ound be ween he g oups alone can ha dly explain he di e en o e sha es. 4. Using da a om he 2017 GLES’Rolling C oss Sec ion (Roß eu sche e al., ‘Rolling C oss Sec ion Cam- paign S udy,’2018), we es ed whe he he di e ences in he o e sha es o CDU/CSU and A D be ween pos al and elec ion day o e s a e simply a unc ion o he iming o he decision making. The e o e, we eg essed indi idual o e in en ion on he numbe o days un il he elec ion day o he las mon h o he campaign. The esul s e eal ha he numbe o days un il he elec ion did no ha e an e ec on he o e in en ions o hese wo pa ies. We ake his as an indica ion ha he be e esul o he A D among elec- ion day o e s compa ed o pos al o e s is no a me e eflec ion o he pa y’s posi i e end in he las weeks and days o he campaign and ice e sa o he CDU/CSU (c . Appendix B). 5. We use egional and socio-s uc u al pos -s a ifica ion weigh s in all ou analyses. 6. The sis e pa ies CDU and CSU a e ea ed as one pa y. 7. Al hough s ongly co ela ed wi h occupa ion (esp. pensione s and pupils/s uden s), we included age as a con ol a iable in a sepa a e model, oo (c . Table A3 in he Appendix C). Howe e , including he age o a 618 GERMAN POLITICS esponden has nei he a significan impac on he likelihood o o ing by mail no does i change he mag- ni ude o significance o one o he o he explana o y a iables. 8. Howe e , he same also applies o elec ion day o e s who make hei decision on which pa y o o e o long be o e elec ion day. ORCID Josephine Lich eblau h p://o cid.o g/0000-0002-9699-6791 REFERENCES Ba le, John. 2005.“Homogeneous Models and He e ogeneous Vo e s.”Poli ical S udies 53 (4): 653–675. Be insky, Adam J. 2005.“The Pe e se Consequences o Elec o al Re o m in he Uni ed S a es.”Ame ican Poli ics Resea ch 33 (4): 471–491. doi:10.1177/ 1532673X04269419. Blumens iel, Jan E ic, and Thomas Plischke. 2015.“Changing Mo i a ions, Time o he Vo ing Decision, and Sho -Te m Vola ili y –The Dynamics o Vo e He e ogen- ei y.”Elec o al S udies 37 (1): 28–40. doi:10.1016/j.elec s ud.2014.11.003. Budge, Ian, and Da id Fa lie. 1983.Explaining and P edic ing Elec ions: Issue E ec s and Pa y S a egies in Twen y-Th ee Democ acies. London: Allen & Unwin. Bundeswahllei e [Fede al Re u ning O fice ]. 2015.Wahl zum Deu schen Bundes ag am 22. Sep embe 2013, He 5, Teil 2. Tex liche Auswe ung (Rep äsen a i e Wahls a is ik und Wahlbezi kss a is ik). Wiesbaden: S a is isches Bundesam . h ps://www.bundeswahllei e .de/dam/jc /a5dccee9-06a -4932-a1a9- 499505c0b818/b w13_he 5-2.pd . Bundeswahllei e [Fede al Re u ning O fice ]. 2017a.E gebnisse in den U nen- und B ie wahlbezi ken nach Lände n.h ps://www.bundeswahllei e .de/en/dam/jc / dd7 37e-90d0-4306-b 62-07a5400c1 ad/b w17_e gebnisse_bezi ksa .pd . Bundeswahllei e [Fede al Re u ning O fice ]. 2017b.Pos al ballo . h ps://www. bundeswahllei e .de/en/se ice/glossa /b/b ie wahl.h ml. Campbell, Angus, Phillip E. Con e se, Wa en Mille , and Donald E. S okes. 1960.The Ame ican Vo e . New Yo k: Wiley. Dal on, Russell J. 1996.Ci izen Poli ics: Public Opinion and Poli ical Pa ies in Ad anced Indus ial Democ acies. Cha ham, NJ: Cha ham House. Downs, An hony. 1957.An Economic Theo y o Democ acy. New Yo k: Ha pe and Row. Dyck, Joshua J., and James G. Gimpel. 2005.“Dis ance, Tu nou , and he Con enience o Vo ing.”Social Science Qua e ly 86 (3): 531–548. Elle mann, Sil ia. 2004.“Die bedeu ung de B ie wähle bei de Bundes agswahl 2002.”In Die Bundes agswahl 2002. Analysen de Wahle gebnisse und des Wahlk- amp es, edi ed by F ank B e schneide , Jan an De h, and Edel aud Rolle , 249– 275. Wiesbaden: VS Ve lag. Fio ina, Mo is P. 1981.Re ospec i e Vo ing in Ame ican Na ional Elec ions. New Ha en: Yale Uni e si y P ess. GERMANY GOING POSTAL? 619 Ga ne , Holly Ann. 2018.“Ad ance Vo ing Wi h Same-Day Regis a ion: The Impac on Tu nou in New Zealand.”Poli ical Science 70 (2): 102–116. doi:10.1080/ 00323187.2018.1561154. Gelman, And ew. 2008.“Scaling Reg ession Inpu s by Di iding by Two S anda d De ia ions.”S a is ics in Medicine 27 (15): 2865–2873. doi:10.1002/sim.3107. Gieble , Heiko. 2014.“Die B ie wähle .”In Zwischen F agmen ie ung und Konzen a- ion: Die Bundes agswahl 2013, edi ed by Rüdige Schmi -Beck, Hans Ra inge , Sig id Roß eu sche , Be nha d Weßels, Ch is o Wol , Ina Biebe , Manuela S. Blu- menbe g, e al., 169–177. Baden-Baden: Nomos. Goodman, Nicole, and Leah C. S okes. 2018.“Reducing he Cos o Vo ing: An E alu- a ion o In e ne Vo ing’s E ec on Tu nou .”B i ish Jou nal o Poli ical Science, 1–13. doi:10.1017/S0007123417000849. G onke, Paul, E a Galanes-Rosenbaum, Pe e A. Mille , and Daniel To ey. 2008. “Con enience Vo ing.”Annual Re iew o Poli ical Science 11 (1): 437–455. doi:10.1146/annu e .polisci.11.053006.190912. K ause, We ne , Ma cus Spi le , and Aiko Wagne . 2017.A ak ion und Repulsion. Anhänge Innen ech s- und linkspopulis ische Pa eien im eu opäischen Ve gle- ich: Le ia han Sonde band 32 | 2017. Lich eblau, Josephine, and Aiko Wagne . 2019.“Die B ie wähle .”In Zwischen Pola - isie ung und Beha ung: Die Bundes agswahl 2017, edi ed by Sig id Roß eu sche , Rüdige Schmi -Beck, Ha ald Schoen, Be nha d Weßels, and Ch is o Wol , 169–179. Baden-Baden: Nomos. Luechinge , Simon, My a Rosinge , and Alois S u ze . 2007.“The Impac o Pos al Vo ing on Pa icipa ion: E idence o Swi ze land.”Swiss Poli ical Science Re iew 13 (2): 167–202. doi:10.1002/j.1662-6370.2007. b00075.x. Magleby, Da id B. 1987.“Pa icipa ion in Mail Ballo Elec ions.”The Wes e n Poli i- cal Qua e ly 40 (1): 79–91. doi:10.2307/448554. McAllis e , Ian, and Damon Mulle . 2018.“Ea ly Vo ing, Elec ion Campaigning and Pa y Ad an age in Aus alia.”Elec o al S udies 52: 103–110. doi:10.1016/j. elec s ud.2018.02.003. Neeley, G an W., and Lillia d E. Richa dson. 2001.“Who is Ea ly Vo ing? An Indi id- ual Le el Examina ion.”The Social Science Jou nal 38 (3): 381–392. doi:10.1016/ S0362-3319(01)00135-5. Ri e s, Douglas. 1988.“He e ogenei y in Models o Elec o al Choice.”Ame ican Jou nal o Poli ical Science 32: 737–757. Roß eu sche , Sig id, and Philipp Sche e . 2014.“Ideologie.”In Zwischen F agmen ie - ung und Konzen a ion: Die Bundes agswahl 2013, edi ed by Rüdige Schmi - Beck, Hans Ra inge , Sig id Roß eu sche , Be nha d Weßels, Ch is o Wol , Ina Biebe , and Manuela S. Blumenbe g, e al., 213–223. Baden-Baden: Nomos. Roß eu sche , Sig id, Rüdige Schmi -Beck, Ha ald Schoen, Be nha d Weßels, Ch is- o Wol , and Alexande S aud . 2018.Rolling C oss Sec ion Campaign S udy (GLES 2017): GESIS Da a A chi e, Cologne. ZA6801 Da a file e sion 3.0.1. doi:10.4232/1.13051. Roß eu sche , Sig id, Rüdige Schmi -Beck, Ha ald Schoen, Be nha d Weßels, Ch is- o Wol , Aiko Wagne , Reinhold Melche , and Heiko Gieble . 2018.Pos -Elec ion 620 GERMAN POLITICS C oss Sec ion (GLES 2017): GESIS Da a A chi e, Cologne. ZA6801 Da a file e sion 4.0.0. doi:10.4232/1.13138. Schä e , Anne, and R. Schmi -Beck. 2014.“Pa eibindungen.”In Zwischen F agmen- ie ung und Konzen a ion: Die Bundes agswahl 2013, edi ed by Rüdige Schmi - Beck, Hans Ra inge , Sig id Roß eu sche , Be nha d Weßels, Ch is o Wol , Ina Biebe , Manuela S. Blumenbe g, e al., 203–211. Baden-Baden: Nomos. Sheppa d, Jill, and Ka ine Beau ega d. 2018.“Ea ly Vo ing in Aus alia: The Cos s and Benefi s o Con enience.”Poli ical Science 70 (2): 117–134. doi:10.1080/ 00323187.2018.1561155. Sou hwell, P iscilla L. 2000.“The E ec o All-Mail Elec ions on Vo e Tu nou .” Ame ican Poli ics Qua e ly 28 (1): 72–79. doi:10.1177/1532673X00028001004. Wagne , Aiko, and Be nha d Weßels. 2012.“Pa ies and Thei Leade s. Does I Ma e How They Ma ch? The Ge man Gene al Elec ions 2009 in Compa ison.”Elec o al S udies 31 (1): 72–82. doi:10.1016/j.elec s ud.2011.06.007. Weßels, Be nha d. 2014.“Wahl e hal en soziale G uppen.”In Zwischen F agmen ie - ung und Konzen a ion: Die Bundes agswahl 2013, edi ed by Rüdige Schmi - Beck, Hans Ra inge , Sig id Roß eu sche , Be nha d Weßels, Ch is o Wol , Ina Biebe , Manuela S. Blumenbe g, e al., 187–202. Baden-Baden: Nomos. APPENDIX A TABLE A1 DIFFERENCES IN SECOND VOTE SHARES BETWEEN POSTAL AND ELECTION DAY VOTERS IN THE ESTIMATION SAMPLE (BASED ON SELF-REPORTED VOTE CHOICE) No e: Di e ences in pe cen age poin s. Nega i e alues depic a highe o e sha e among pos al han elec ion day o e s. GERMANY GOING POSTAL? 621 APPENDIX B In o de o es whe he he conside able di e ences in he o e sha es o CDU/CSU and A D be ween pos al and elec ion day o e s a e simply a unc ion o he iming o he decision making, we eg essed indi idual o e in en ion on he numbe o days un il he elec ion day o he las mon h o he campaign using da a om he GLES Rolling C oss Sec ion Campaign S udy. Table A2 displays he esul s. The ime a iable nei he has an s a is ically significan e ec on he likelihood o o e o he CDU/CSU no he like- lihood o o e o he A D. Hence, i seems unlikely ha he A D’s be e esul and he CDU/CSU’s wo se esul among elec ion day o e s was me ely due o pa ies’posi- i e o nega i e ends in he las weeks and days o he campaign. APPENDIX C TABLE B1 LOGIT MODELS FOR PREDICTING VOTE CHOICE FOR THE CDU/CSU AND AFD BY THE NUMBER OF DAYS UNTIL THE ELECTION (LOGIT COEFFICIENTS) CDU/CSU A D Field day −0.01 (0.01) −0.01 (0.02) Cons an −0.18 (0.46) −0.33 (0.70) Obse a ions 2,570 2,570 Pseudo R-squa ed 0.00 0.00 No e: S anda d e o s in pa en heses. ***p< 0.001, **p< 0.01, *p< 0.05. Socio- and egional-s uc u al weigh s applied. Sou ce: GLES olling c oss-sec ion su ey 2017 (Roß eu sche e al., ‘Rolling C oss Sec ion Campaign S udy,’2018). TABLE C1 RESULTS OF LOGISTIC REGRESSION PREDICTING THE VOTING TYPE (POSTAL VOTING = 1) Main model Main model including age Age –0.23 (0.27) Male −0.27 (0.15) −0.27 (0.15) Wes 0.40* (0.17) 0.40* (0.17) Occupa ional s a us ( e e ence: wo ke s) Ci il se an s/whi e colla 0.49 (0.32) 0.51 (0.32) Sel -employed 1.11** (0.40) 1.09** (0.40) Pensione s 0.78* (0.31) 0.66* (0.32) Pupils/s uden s 0.96* (0.43) 1.08* (0.45) Educa ion ( e e ence: high) Low 0.10 (0.21) 0.05 (0.22) Con inued 622 GERMAN POLITICS TABLE C1 CONTINUED. Main model Main model including age Medium 0.07 (0.17) 0.05 (0.17) Income 0.33* (0.13) 0.32* (0.13) U ban esiden ial a ea 0.19 (0.15) 0.19 (0.15) Chu ch a endance ( e e ence: ne e ) Once/a ew imes a yea 0.06 (0.15) 0.05 (0.15) Once a mon h/mo e o en −0.12 (0.23) −0.13 (0.23) Union membe 0.36 (0.21) 0.36 (0.21) Pola i y o le - igh posi ion −0.08 (0.16) −0.08 (0.16) PID 0.01 (0.18) 0.00 (0.18) Poli ical in e es (Ve y) s ong −0.13 (0.22) −0.12 (0.22) Weak/none 0.02 (0.16) 0.01 (0.16) Sa is ac ion wi h democ acy ( e e ence: nei he /no ) (Ve y) sa isfied −0.10 (0.31) −0.08 (0.31) No e y/no a all sa isfied 0.31 (0.18) 0.31 (0.18) Cons an −2.36*** (0.42) −2.60*** (0.50) Obse a ions 1,319 1,318 Pseudo R-squa ed 0.04 0.04 No e: Logi coe ficien s. S anda d e o s in pa en heses. ***p< 0.001, **p< 0.01, *p< 0.05. GERMANY GOING POSTAL? 623 TABLE C3 RESULTS OF SEPARATE CONDITIONAL LOGIT MODELS PREDICTING PARTY CHOICE Elec ion day o e s Pos al o e s PID 0.14*** (0.17) 1.61*** (0.27) Le - igh dis ance −0.04*** (0.01) −0.07** (0.02) E alua ion o op candida es 0.34*** (0.03) 0.45*** (0.06) P oblem sol ing compe ence 0.76*** (0.11) 0.95*** (0.18) E alua ion go e nmen pe o mance 0.07*** (0.02) 0.04 (0.04) Obse a ions 5,928 2,419 Pseudo R-squa ed 0.46 0.51 No e: Logi coe ficien s. S anda d e o s in pa en heses. ***p< 0.001, **p< 0.01, *p< 0.05. TABLE C2 RESULTS OF LOGISTIC REGRESSIONS PREDICTING PARTY CHOICE WITH REGION, OCCUPATIONAL STATUS AND INCOME CDU/CSU SPD Le G eens FDP A D Wes 0.37* (0.14) 0.42* (0.17) −0.97*** (0.19) 0.62** (0.22) 0.07 (0.20) −0.39 (0.21) Occupa ional s a us ( e e ence: wo ke s) Ci il se an s/whi e colla 0.19 (0.25) 0.17 (0.30) −0.05 (0.34) 1.14* (0.48) 0.44 (0.38) −0.69* (0.30) Sel -employed 0.14 (0.35) −1.29* (0.55) −0.02 (0.50) 1.54** (0.54) 0.94* (0.47) −0.67 (0.48) Pensione s 0.52* (0.26) 0.51 (0.30) −0.60 (0.36) 0.07 (0.50) 0.76* (0.38) −1.21*** (0.32) Pupils/s uden s −0.78 (0.40) −0.69 (0.44) 0.33 (0.42) 1.15* (0.54) 0.28 (0.53) −2.22*** (0.66) Income 0.05 (0.13) −0.28 (0.15) 0.28 (0.20) 0.33* (0.17) 0.70*** (0.20) −0.19 (0.21) Cons an −1.98* (0.89) −0.05 (1.04) −3.54** (1.36) −5.76*** (1.24) −7.81*** (1.42) −0.05 (1.45) Obse a ions 1,573 1,573 1,573 1,573 1,573 1,573 Pseudo R-squa ed 0.02 0.03 0.03 0.06 0.03 0.04 No e: Logi coe ficien s. Robus s anda d e o s in pa en heses. ***p< 0.001, **p< 0.01, *p< 0.05. 624 GERMAN POLITICS FIGURE C1 RESULTS OF SEPARATE CONDITIONAL LOGIT MODELS FOR VOTE CHOICE (AVERAGE MARGINAL EFFECTS, AME) GERMANY GOING POSTAL? 625