Privatizing Disability Insurance

Seibold, A hu ; Sei z, Sebas ian; Siegloch, Sebas ian
A icle — Published Ve sion
P i a izing Disabili y Insu ance
Econome ica
P o ided in Coope a ion wi h:
John Wiley & Sons
Sugges ed Ci a ion: Seibold, A hu ; Sei z, Sebas ian; Siegloch, Sebas ian (2025) : P i a izing Disabili y
Insu ance, Econome ica, ISSN 1468-0262, Wiley, Hoboken, NJ, Vol. 93, Iss. 5, pp. 1697-1737,
h ps://doi.o g/10.3982/ECTA22113
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/329806
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/
Econome ica, Vol. 93, No. 5 (Sep embe , 2025), 1697–1737
PRIVATIZING DISABILITY INSURANCE
ARTHUR SEIBOLD
Depa men o Economics, Ludwig-Maximilians Uni e si y Munich and CEPR
SEBASTIAN SEITZ
Depa men o Economics, Uni e si y o Manches e and ZEW
SEBASTIAN SIEGLOCH
Depa men o Economics, Uni e si y o Cologne, CEPR, and ZEW
Public disabili y insu ance (DI) p og ams in many coun ies ace g owing iscal p es-
su es, p omp ing e o s o educe spending. In his pape , we in es iga e he wel a e
e ec s o expanding he ole o p i a e insu ance ma ke s in he ace o public DI cu s.
We exploi a e o m ha abolished one pa o Ge man public DI and use unique da a
om a la ge insu e . We documen modes c owding-ou e ec s o he e o m, such
ha p i a e DI ake-up emains incomple e. We ind no ad e se selec ion in he p i-
a e DI ma ke . Ins ead, p i a e DI ends o a ac indi iduals wi h high income, high
educa ion, and low disabili y isk. Using a e ealed p e e ence app oach, we es ima e
indi idual insu ance alua ions. Ou wel a e analysis inds ha pa ial DI p o ision ia
he olun a y p i a e ma ke can imp o e wel a e. Howe e , dis ibu ional conce ns
may jus i y a ull public DI manda e.
KEYWORDS: Disabili y insu ance, social insu ance, p i a iza ion, isk-based selec-
ion.
1. INTRODUCTION
ACROSS THE DEVELOPED WORLD, he numbe o indi iduals ecei ing public disabili y
insu ance (DI) bene i s has isen apidly o e he pas decades. This has made DI one
o he la ges social insu ance p og ams in OECD coun ies (OECD (2024)). Due o he
inc easing iscal bu den, go e nmen s ace p essu e o enac e o ms educing he gen-
e osi y o public DI p og ams. While such e o ms help imp o e iscal sus ainabili y, hey
na u ally come a he cos o p o iding less insu ance o indi iduals a isk o disabili y.
As a consequence, some economis s and policymake s ha e p oposed a la ge ole o
p i a e DI (e.g., GAO (2018)). Sizable p i a e DI ma ke s al eady exis in many coun-
ies, including he U.S. and Ge many. Howe e , p i a e DI p o ision aces se e al po-
en ial p oblems, including ma ke ailu es due o ad e se selec ion and equi y conce ns
A hu Seibold: [email p o ec ed]
Sebas ian Sei z: [email p o ec ed]
Sebas ian Siegloch: [email p o ec ed]
We hank Ma ika Cab al, And eas Halle , Da id Koll, Camille Landais, Ch is ina Meye , Johannes Spin-
newijn, De ek Wu, Jose Zweimülle as well as semina pa icipan s a Ox o d, LSE, CREST, LMU Munich,
Manches e , S . And ews, Tinbe gen Ins i u e Ams e dam, S ockholm, Copenhagen, FU Be lin, F ank u
School o Finance/Goe he, Hohenheim, KIT, he NBER Summe Ins i u e, he CEPR Public Economics Sym-
posium, NHH, ZEW, CESi o, he Bocconi-CEPR Social Secu i y Wo kshop, Bonn-Mannheim CRC, NTA,
IIPF, EEA, and Bonn-Cologne ECON ibu e o help ul commen s and sugges ions. We especially hank Eck-
ha d Janeba o his suppo in ob aining he da a used in his pape . A hu Seibold g a e ully acknowledges i-
nancial suppo om he Daimle and Benz Founda ion and he Ge man Resea ch Founda ion (DFG) h ough
CRC TR 224 (P ojec C01). Sebas ian Siegloch g a e ully acknowledges inancial suppo om he DFG unde
he Excellence S a egy—EXC 2126/1-390838866.
© 2025 The Au ho s. Econome ica published by John Wiley & Sons L d on behal o The Econome ic Socie y.
A hu Seibold is he co esponding au ho on his pape . This is an open access a icle unde he e ms o
he C ea i e Commons A ibu ion License, which pe mi s 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.
1698 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
(Liebman (2015)). Despi e he signi icance o his policy deba e, he e is ema kably li le
empi ical e idence on he unc ioning o p i a e DI ma ke s.1
In his pape , we b eak new empi ical g ound on hese issues. We in es iga e key ea-
u es o p i a e DI ma ke s and s udy he wel a e consequences o expanding he ole o
p i a e DI. We exploi a unique e o m ha abolished one pa o public DI o Ge man
wo ke s. Ou analysis is based on no el mic oda a om a la ge p i a e insu e and ad-
minis a i e da a on he uni e se o public DI claims. This se ing and da a allow us o
ob ain h ee pieces o empi ical e idence ha go e n he wel a e impac o p i a e DI
p o ision. Fi s , we es ima e c owding-ou e ec s be ween public and p i a e DI. Second,
we documen subs an ial he e ogenei y in p i a e DI ake-up, which aises dis ibu ional
conce ns. Thi d, we ca e ully es o isk-based selec ion, which p o ides c ucial in o -
ma ion abou he e iciency o he p i a e DI ma ke . We adop a e ealed p e e ence
app oach o es ima e indi idual insu ance alua ions based on hese empi ical ac s. Ou
wel a e analysis shows ha pa ly p i a izing DI can imp o e wel a e. Howe e , a ull
public DI manda e could be jus i ied by equi y conce ns.
Assessing he po en ial o p i a e DI o complemen public DI is challenging o wo
easons. Fi s , sui able a ia ion in public DI co e age is needed. Second, comp ehensi e
and eliable da a on p i a e insu ance is necessa y o quan i y he in e ac ion be ween
public and p i a e DI. This pape is he i s o o e come bo h o hese challenges.
We exploi exogenous a ia ion induced by a e o m ha emo ed one pa o Ge man
public DI and eplaced i wi h a olun a y p i a e DI ma ke . Speci ically, he e o m o
2001 p i a ized own-occupa ion DI o younge wo ke s. Recei ing own-occupa ion DI
bene i s equi es wo ke s o be unable o wo k in hei p e ious occupa ion. In con as ,
gene al DI bene i s a e based on s ic e eligibili y c i e ia: being unable o wo k in any
occupa ion. Be o e he e o m, bo h own-occupa ion and gene al DI we e pa o he
social insu ance sys em. The e o m comple ely emo ed public own-occupa ion DI o
bi h coho s 1961 and younge . Impo an ly, he Ge man p i a e DI ma ke o e s con-
ac s including own-occupa ion DI co e age, o which wo ke s can op i hey wish o
compensa e o he loss o public DI co e age.
To add ess he second challenge, we ob ained a no el da a se on all p i a e DI con ac s
issued by a op-10 insu e in he Ge man p i a e DI ma ke . This da a con ains high-
quali y in o ma ion on p i a e DI co e age, p ices, and indi idual cha ac e is ics, which
a e c i ical inpu s in o ou empi ical analysis. We demons a e ha he insu e mic oda a
is ep esen a i e o he o e all p i a e DI ma ke in key dimensions using agg ega e da a
on he en i e p i a e DI ma ke om a leading a ing agency and ep esen a i e house-
hold su ey da a. Complemen ing he da a on p i a e DI, we use adminis a i e da a on
he uni e se o public DI claims.
We p o ide h ee pieces o empi ical e idence on he p i a e DI ma ke . Fi s , we s udy
c owding-ou e ec s o he e o m, ha is, he impac o public DI cu s on p i a e DI ake-
up. On agg ega e, we ind subs an ial g ow h in he p i a e DI ma ke a e he e o m.
To iden i y a causal e ec , we use a di e ence-in-di e ence s a egy exploi ing he coho
cu o o he e o m. We ind ha ea ed indi iduals bo n abo e he cu o signi ican ly
inc ease p i a e DI pu chases compa ed o con ol coho s below he cu o . Ye , e en
15 yea s a e he e o m, o e all ake-up emains modes , as only 26% o wo ke s ha e
p i a e DI. Thus, c owding-ou is a om comple e.
1Fo ins ance, in hei epo o Cong ess, he U.S. Go e nmen Accoun abili y O ice (GAO (2018)) con-
cludes ha he implica ions o a ious p oposals o expand p i a e DI canno be ully assessed due o “an a ay
o complex ac o s ha could in luence p i a e DI expansions and SSDI cos sa ings— ac o s o which da a,
me hods, and assump ions [...] a e ei he un eliable, unsuppo ed, o una ailable.”
PRIVATIZING DISABILITY INSURANCE 1699
Second, we documen subs an ial he e ogenei y in p i a e DI ake-up. Modes o e all
p i a e DI ake-up is mainly d i en by low ake-up among indi iduals wi h low income and
low educa ion. Fo ins ance, ake-up is 65% in he op income quin ile bu only 7% o 12%
in he bo om h ee quin iles. Mo eo e , he e is impo an he e ogenei y in ake-up by
p iced isk g oups, which insu e s assign o wo ke s based on occupa ions. Indi iduals in
high- isk g oups acing highe p emiums a e much less likely o ake up insu ance. These
pa e ns indica e po en ial equi y issues in he p i a e DI ma ke .
Thi d, we ca e ully in es iga e isk-based selec ion in o p i a e DI using wo comple-
men a y empi ical s a egies. The i s s a egy is a “posi i e co ela ion es ” (Chiappo i
and Salanié (2000)), eg essing p i a e DI ake-up on disabili y isk a he le el o ine-
g ained occupa ions. As a second s a egy, we conduc a “cos cu e es ” (Eina , Finkel-
s ein, and Cullen (2010)), which exploi s a ia ion in p i a e DI p emiums o e ime o
es o di e ences in isk ac oss indi iduals. We do no ind any e idence o ad e se selec-
ion. In bo h empi ical designs, isk-based selec ion is o a small and insigni ican deg ee.
A i s glance, he lack o ad e se selec ion may be su p ising, as insu ance should, in
p inciple, be mo e aluable o highe - isk indi iduals. We discuss se e al po en ial expla-
na ions o his esul . Fo ins ance, he e could be he e ogeneous p e e ences o insu -
ance ha a e nega i ely co ela ed wi h isk. In pa icula , he ac ha indi iduals wi h
high educa ion and high income a e mo e likely o pu chase p i a e DI seems o coun e
ad e se selec ion. We also explo e he possibili y o he e ogeneous ea nings losses upon
disabili y ha could make p i a e DI mo e aluable o wo ke s in high-skill occupa ions.
We hen u n o he wel a e consequences o p i a izing DI p o ision. The analysis
builds on Eina , Finkels ein, and Cullen (2010), who show ha insu ance demand and
cos cu es a e su icien s a is ics o assess wel a e in insu ance ma ke s. Ou se ing wi h
insu ance choice p o ides a unique oppo uni y o implemen a e ealed p e e ence ap-
p oach and di ec ly es ima e indi iduals’ willingness o pay o he DI co e age o e ed by
he p i a e ma ke . To ace ou he slope o he p i a e DI demand cu e, we implemen
an e en s udy design a ound occupa ion eclassi ica ions, which en ail la ge changes in
p i a e DI p emiums. This es ima ion yields a demand elas ici y o −1.06.
Ou coun e ac ual analysis compa es he pos - e o m s a us quo, whe e DI is pa ly
p o ided ia he p i a e ma ke , o a ull public DI manda e, including his ex a co e age.
We conside wo wel a e measu es: (i) he ne alue o DI, which exp esses he alue o
DI ela i e o i s di ec cos , and (ii) he ma ginal alue o public unds (Hend en and
Sp ung-Keyse (2020)). Fo he la e , we calib a e addi ional indi ec iscal e ec s based
on ou da a and es ima es om he li e a u e. Ou main ne alue es ima e implies ha
e ealed insu ance alua ions among indi iduals ecei ing ex a co e age amoun o only
70.5% o he di ec cos o insu ance. Simila ly, he ma ginal alue o public unds o a
manda e is 0.617.
This ini ial wel a e analysis igno es dis ibu ional conce ns. In an ex ension, we inco -
po a e wel a e weigh s based on a U ili a ian social wel a e unc ion. We ind ha a ull
public DI manda e has a social alue exceeding i s cos s, e en unde small deg ees o eq-
ui y conce n. Impo an ly, edis ibu i e e ec s hinge on he design o social insu ance.
Ap i a e DI manda e does no inc ease social wel a e since insu ance bene i s o high-
isk g oups a e coun e ac ed by isk- a ed p emiums. In con as , a public DI manda e
inanced by income-based social insu ance con ibu ions e ec i ely edis ibu es o low-
income, high- isk indi iduals.
O e all, his pape shows ha se e al ma ke ailu es discussed in p io li e a u e can-
no jus i y manda ing ex a DI co e age in his con ex . Mos impo an ly, we do no ind
any e idence o ad e se selec ion in he p i a e DI ma ke , which is o en conside ed
1700 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
he canonical a ionale o a public DI manda e. We also accoun o ma ke powe and
adminis a i e cos s, which con ibu e o low ake-up in o he p i a e insu ance ma ke s
(B aun, Kopecky, and Ko eshko a (2019)), and we calib a e se e al po en ial ex e nali-
ies. We ind ha none o hese ac o s wa an a manda e in ou se ing. Hence, he p i-
a e ma ke co e s he majo i y o indi iduals wi h a su icien ly high willingness o pay,
and a manda e would p edominan ly en oll hose wi h alua ions below ma ginal cos . I
is impo an o no e ha ou wel a e analysis ollows exis ing li e a u e and main ains he
assump ions o he e ealed p e e ence app oach, which abs ac s om beha io al biases.
S udying such biases is beyond he scope o his pape .
We a gue ha ou indings a e ele an beyond he Ge man con ex o wo main ea-
sons. Fi s , ou sys ema ic e iew o DI sys ems ac oss coun ies e eals ha bo h he
Ge man public DI sys em and he Ge man p i a e DI ma ke sha e key cha ac e is ics
wi h o he coun ies, making ou se ing a b oadly ep esen a i e case s udy o p i a izing
DI. We also p o ide e idence ha ou esul s a e no d i en by speci ic ins i u ional ea-
u es, such as means- es ed social assis ance. Second, own-occupa ion disabili y isk— he
sub isk ou se ing cen e s on—is s ongly posi i ely co ela ed wi h o e all disabili y isk
in he da a. Hence, he ypes o indi iduals who would bene i om p i a e DI end o be
hose who would bene i om DI mo e gene ally. One may he e o e expec simila pa -
e ns o eme ge in o he se ings, including hose whe e he de ails o DI co e age di e .
Indeed, he limi ed exis ing e idence on p i a e DI ake-up om o he coun ies poin s
in his di ec ion.2
This pape con ibu es o a la ge and g owing li e a u e on DI (see Low and Pis a e i
(2020), o a e iew). Much o his li e a u e ocuses on public DI and i s e ec on labo
supply and claiming decisions (Bound (1989), G ube (2000), Au o and Duggan (2003,
2006,2007), Chen and an de Klaauw (2008), Au o , Duggan, and Lyle (2011), S aubli
(2011), on Wach e , Song, and Manches e (2011), Ma ie and Vall Cas ello (2012), Maes-
as, Mullen, and S and (2013), F ench and Song (2014), Kos ol and Mogs ad (2014),
Bo ghans, Gielen, and Lu me (2014), Koning and Lindeboom (2015), Liebman (2015),
Au o , Duggan, G eenbe g, and Lyle (2016), Bu khause , Daly, and Zieba h (2016),
Deshpande (2016a,b), Mullen and S aubli (2016), Gelbe , Moo e, and S and (2017), Au-
o , Kos ol, Mogs ad, and Se zle (2019), Ruh and S aubli (2019)). In con as , he e is
much less p io wo k on p i a e DI ma ke s. Excep ions include Au o , Duggan, and G u-
be (2014), S epne (2021), and Sei z (2021), who analyze mo al haza d e ec s o p i a e
DI. A ecen wo king pape by Fische , Geye , and Zieba h (2024) also in es iga es he
Ge man e o m o 2001. Thei app oach is complemen a y o ou s, combining su ey da a
wi h a gene al equilib ium model o he p i a e DI ma ke . Simila o ou esul s, hey
ind modes p i a e DI ake-up a e he e o m, pa icula ly among low-income g oups.
While ou wel a e analysis ocuses on he ade-o be ween p i a e and public DI p o-
ision, hey emphasize he ole o supply-side ac o s o ake-up and wel a e wi hin he
p i a e DI ma ke .
We make h ee main con ibu ions o his li e a u e. Fi s , exploi ing ou unique da a
and se ing, we p o ide no el empi ical e idence on c owding-ou and selec ion in p i a e
DI ma ke s. Ou indings cons i u e he i s di ec empi ical e idence on hese issues,
which a e key in assessing he wel a e impac o policies expanding he ole o p i a e
ma ke s and choice in DI. Second, we u he exploi ou se ing wi h insu ance choice
o es ima e willingness o pay o DI in a e ealed p e e ence app oach. So a , li le is
2Fo ins ance, Au o , Duggan, and G ube (2014)andGAO (2018) epo ha p i a e DI ake-up inc eases
wi h income in U.S. da a, simila o ou indings.

PRIVATIZING DISABILITY INSURANCE 1701
known abou how indi iduals alue DI. One excep ion is Cab al and Cullen (2019), who
es ima e willingness o pay o supplemen al DI co e age wi hin a U.S. i m and de i e
a lowe bound on public DI alua ions.3Ou hi d con ibu ion is o assess he wel a e
consequences o pa ial p i a e DI p o ision s. a ull public manda e. This complemen s
and ex ends exis ing wo k analyzing wel a e and he insu ance-incen i e ade-o wi hin
public DI (Diamond and Sheshinski (1995), Low and Pis a e i (2015), Meye and Mok
(2019), Halle , S aubli, and Zweimülle (2024)).
Mo e b oadly, ou wo k builds on a ich li e a u e es ing o isk-based selec ion
in social insu ance se ings. In Supplemen al Appendix D (Seibold, Sei z, and Siegloch
(2025)), we p o ide a mo e de ailed e iew o his li e a u e and conduc a me a-analysis
o s udies on isk-based selec ion published in leading economics jou nals since 2000.
A la ge majo i y o exis ing wo k ocuses on heal h insu ance. Se e al pape s ha e exam-
ined long- e m ca e insu ance (LTCI), pension annui ies, and unemploymen insu ance.
E idence on isk-based selec ion in DI is pa icula ly sca ce. The only excep ion is Hen-
d en (2013), who p o ides indi ec e idence on po en ial isk-based selec ion in DI by
documen ing ha indi iduals ha e p i a e in o ma ion abou hei disabili y isk.
While p io wo k on heal h insu ance, pension annui ies, and unemploymen insu ance
ends o ind e idence o ad e se selec ion, ou analysis o DI shows some in e es ing
pa allels o LTCI. DI and LTCI sha e se e al ea u es, including isks p edominan ly oc-
cu ing la e in li e, s ong isk- a ing in he p i a e ma ke , and he coexis ence o public
and p i a e schemes. Like us, Finkels ein and McGa y (2006)andBoye , De Donde ,
Flue , Le oux, and Michaud (2020) ind no ad e se selec ion in p i a e LTCI, wi h poin
es ima es indica ing ad an ageous selec ion. B aun, Kopecky, and Ko eshko a (2019)in-
es iga e he in e play o asymme ic in o ma ion, ma ke powe on he supply side, and
o he ans e p og ams in explaining low obse ed LTCI ake-up.
Finally, his pape con ibu es o a nascen li e a u e in es iga ing he wel a e e -
ec s o uni e sal manda es e sus olun a y ma ke s in social insu ance se ings. Exis -
ing s udies on hese issues include wo k on heal h insu ance (e.g., Eina , Finkels ein,
and Cullen (2010), Finkels ein, Hend en, and Shepa d (2019)), unemploymen insu ance
(Landais, Nekoei, Nilsson, Seim, and Spinnewijn (2021), Hend en, Landais, and Spin-
newijn (2021)), and wo ke s’ compensa ion (Cab al, Cui, and Dwo sky (2022)). Ou main
con ibu ion o his wide li e a u e is ha we p o ide he i s wel a e analysis o a p i-
a e ma ke wi h choice e sus a ull public manda e in he con ex o DI, one o he mos
impo an social insu ance p og ams.
The emainde o his pape is o ganized as ollows. Sec ion 2ou lines con ex and da a,
Sec ion 3p esen s e idence on c owding-ou , Sec ion 4discusses he e ogenei y in p i a e
DI ake-up, Sec ion 5 es s o isk-based selec ion, Sec ion 6p esen s he demand and
cos es ima ion, Sec ion 7shows he wel a e analysis, and inally Sec ion 8concludes.
2. INSTITUTIONAL BACKGROUND AND DATA
2.1. Disabili y Insu ance in Ge many
Public Disabili y Insu ance. In Ge many, public disabili y insu ance (DI) is adminis-
e ed by he S a e Pension Fund. En ollmen is manda o y o all employed indi iduals,
while mos sel -employed wo ke s and ci il se an s a e exemp . DI con ibu ions a e
3In addi ion, a ew s udies use indi ec consump ion-based me hods o quan i y he insu ance alue o public
DI (e.g., Meye and Mok (2019), Deshpande and Lockwood (2022)).
1702 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
le ied as pay oll axes oge he wi h old-age pension con ibu ions. En olled wo ke s be-
come eligible o DI bene i s in he e en o a long- e m disabili y limi ing hei wo k
capaci y. Mo eo e , eligibili y equi es ha ing con ibu ed o a leas 5 yea s in o al and
a leas 3 ou o he 5 yea s be o e he onse o disabili y. Upon applica ion, a medical
and wo k capaci y assessmen is ca ied ou by he Pension Fund. Bene i s a e a unc-
ion o wo ke s’ con ibu ions, assuming hey would ha e kep con ibu ing acco ding o
hei a e age p edisabili y ea nings un il age 63. Public DI bene i s a e hus oughly p o-
po ional o indi iduals’ p edisabili y ea nings, wi h an a e age g oss eplacemen a e o
39%. Bene i s can be paid un il a wo ke eaches he no mal e i emen age, when hey
a e con e ed in o an old-age pension. Acco ding o ou adminis a i e da a, 25.1% o
Ge man wo ke s become disabled and claim public DI h oughou hei li e ime.
C ucially o ou pu poses, he public DI sys em consis s o wo sepa a e ie s, gene al
DI and own-occupa ion DI. The i s ie pays bene i s o wo ke s su e ing om a gen-
e al disabili y (E we bsun ähigkei ), such ha hey a e unable o wo k in any occupa ion
o mo e han 3 hou s pe day. Common condi ions leading o gene al disabili y include
degene a i e disc disease o se e e dep ession/bu n-ou . The second ie , on he o he
hand, equi es an own-occupa ion disabili y (Be u sun ähigkei ), de ined as being unable
o wo k in one’s p e ious occupa ion. Fo ins ance, a bus d i e su e ing om ision im-
pai men is unable o wo k in hei occupa ion bu may be able o wo k in o he jobs.
The wo DI ie s equi e sepa a e applica ions. Wo ke s on own-occupa ion DI ecei e
wo- hi ds o gene al DI bene i s bu ace a less s ingen ea nings es .4Own-occupa ion
DI cases make up 13.2% o all public DI claims.
The Re o m o 2001. Be o e 2001, all wo ke s we e co e ed bo h by gene al and own-
occupa ion DI as pa o he public DI manda e. Howe e , ising expendi u e on DI ben-
e i s s oked conce ns abou he iscal sus ainabili y o he p og am in he 1990s. This mo i-
a ed a majo e o m in 2001 aimed a educing public DI spending. Mos impo an ly, he
e o m ea u ed a sha p, coho -based change in he scope o public DI: own-occupa ion
DI co e age was comple ely emo ed o bi h coho s 1961 and younge om 2001 on-
wa d. Besides his main elemen , he e o m ea u ed u he changes equally a ec ing
all coho s, including g adually phased-in changes o bene i calcula ion.5
The iming o he e o m was no ewo hy. Ini ially, he e o m was announced in De-
cembe 1997 o ake e ec in Janua y 1999. The ini ial e o m p oposal in ended o abol-
ish own-occupa ion DI o all wo ke s, no only o younge coho s. A e a change o
ede al go e nmen , he e o m was e ac ed in la e 1998. Howe e , in Decembe 2000,
he e o m was eannounced in i s inal o m ea u ing he coho cu o , and he changes
ook e ec in Janua y 2001.
P i a e Disabili y Insu ance. Acco ding o ou a ing agency da a, a leas 73 insu ance
companies o e p i a e DI con ac s. In 2015, he op 3 p o ide s had a combined ma ke
4Gene al DI bene i s a e educed o mon hly ea nings abo e EUR 400, whe eas wo ke s on own-
occupa ion DI a e allowed o ea n a leas EUR 700, depending on hei p io ea nings. No e ha hese
ea nings es h esholds a e adjus ed e e y ew yea s. The a o emen ioned igu es apply be ween 2008 and
2017.
5Mo e p ecisely, he e o m al e ed wo elemen s o bene i calcula ion. Fi s , an adjus men ac o was g ad-
ually in oduced, ea u ing nega i e bene i adjus men s simila o penal ies o ea ly old-age pension claims.
Second, he hypo he ical con ibu ion pe iod used o bene i calcula ion was g adually ex ended, somewha
coun e ac ing he new penal ies. In addi ion, he e o m in oduced he possibili y o claiming pa ial DI bene-
i s o indi iduals wi h a gene al disabili y who can wo k be ween 3 and 6 hou s pe day. Finally, wo k capaci y
eassessmen s we e in oduced, bu in p ac ice, mos bene icia ies s ill ecei e bene i s pe manen ly.
PRIVATIZING DISABILITY INSURANCE 1703
sha e o 34.2%, and he op 10 p o ide s had a ma ke sha e o 62.7%. C ucially, p i a e
DI always includes co e age o own-occupa ion disabili y isk, closely mi o ing he p e-
e o m public DI sys em. Thus, wo ke s a ec ed by he e o m can choose o pu chase
p i a e DI in o de o compensa e o he emo al o public own-occupa ion DI. P i a e
DI payou s a e independen o he public DI sys em, such ha hey can also se e as a
op-up in case a wo ke is awa ded public DI bene i s. Be o e 2001, p i a e DI pu ely
se ed as such op-up insu ance.
An impo an di e ence o he public DI sys em is ha p i a e DI p emiums a e isk-
based. The p ima y de e minan o p i a e DI p emiums a e indi iduals’ occupa ions,
whe eby insu e s map occupa ions in o a disc e e numbe o isk g oups (see below). Fu -
he mo e, insu ance p emiums can be adjus ed o p eexis ing medical condi ions and
isky leisu e ac i i ies, bu his occu s in equen ly.6Finally, mon hly p emiums a e ac u-
a ially adjus ed o he indi idual’s con ac s a and end da e. This p icing p ac ice has
emained la gely unchanged h oughou ou sample pe iod and applies o all buye s. P i-
a e DI con ac s ypically speci y a ixed amoun o insu ed bene i s, which can be se
indi idually. In p ac ice, mon hly p i a e DI payou s a e o a simila magni ude o pub-
lic DI bene i s (see Sec ion 2.3). Ge man p i a e DI is la gely a nong oup ma ke : he
majo i y o 85% o con ac s a e pu chased indi idually, and he emainde a e ob ained
ia employe s (FAZ (2012)). Excep a ew niche p o ide s, insu e s gene ally o e con-
ac s o all occupa ions. Acco ding o o icial s a is ics, only 3% o con ac applica ions
a e denied by insu e s (GDV (2023)).7In e ms o p ima y ea u es, including he de -
ini ion o disabili y, bene i le els, and con ac du a ion, p i a e DI con ac s a e qui e
homogenous ac oss p o ide s.8P i a e DI can be bough ei he as a s and-alone p oduc
o bundled wi h o he ypes o insu ance, mos commonly li e insu ance.
Table Ip o ides summa y in o ma ion abou p i a e DI p emiums and isk g oups. The
a e age mon hly p emium o insu e EUR 1000 o mon hly bene i s is a ound EUR 73 o
a con ac s a age o 25 and EUR 98 o a s a age o 45. Following s anda d p ac ice
in he indus y, he insu e om which ou mic oda a o igina es uses i e isk g oups o
p ice DI con ac s. Appendix Table A1 shows equen occupa ions in each isk g oup.
Examples o occupa ions classi ied as low- isk include medical doc o s, compu e scien-
is s, and accoun an s. Medium- isk occupa ions include high-school eache s, sec e a ies,
and elec ical enginee s; high- isk occupa ions include bake s, i e igh e s, and wa ehouse
wo ke s. P i a e DI p emiums di e s ongly ac oss isk g oups: o ins ance, an indi id-
ual in he lowes - isk g oup 1 is cha ged EUR 32 a age 25, while p emiums ise up o
EUR 155 o he highes - isk g oup 5. This a ia ion in p emiums is oughly in line wi h
di e ences in disabili y isk ac oss occupa ions. Acco ding o ou adminis a i e da a, he
li e ime disabili y isk o indi iduals in isk g oup 1 is less han 5%, while i is 24% in isk
6P emiums a e adjus ed beyond isk-g oup speci ic p ices in only 4% o p i a e DI con ac s (GDV (2023)).
7This includes a ew ex emely isky occupa ions such as ci cus a is s and explosi es wo ke s, as well as
ejec ions due o p eexis ing condi ions o isky leisu e ac i i ies. Besides co e age denials, ano he 12% o
p i a e DI con ac s ea u e exclusion clauses o p eexis ing condi ions. The ela i ely low denial and exclusion
a es a e an impo an di e ence o U.S. nong oup insu ance ma ke s, whe e hese a e mo e equen and can
a ec b oade se s o occupa ions and income g oups (Hend en (2013), B aun, Kopecky, and Ko eshko a
(2019)).
8The well-known consume ad ice o ganiza ion S i ung Wa en es (2024) inds in hei la es epo ha
con ac s o e ed by 67 p i a e DI p o ide s a e essen ially homogenous in hese p ima y cha ac e is ics. E en
on seconda y cha ac e is ics such as wai ing pe iods, e oac i e bene i adjus men s, and ules on occupa ion
swi ches, con ac s a e qui e simila : he a e age p o ide sa is ies 86% o seconda y c i e ia se by he epo .
Ul ima ely, 85% o p o ide s ecei e he highes (“ e y good”) o second-highes (“good”) a ing.
1704 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
TABLE I
RISK GROUPS,DISABILITY RISK,AND PRIVATE DI PREMIUMS.
(1) (2) (3) (4) (5) (6)
Risk g oup Sha e o
labo o ce Li e ime
disabili y isk
Sha e o
own-occupa ion
DI claims
Mon hly insu ance p emium
o con ac s a a ...
age 25 age 35 age 45
All 10000% 2506% 1320% 7284 8354 9815
Risk G oup 1 972% 481% 1085% 3161 3595 4322
Risk G oup 2 1699% 1535% 806% 4172 4908 5749
Risk G oup 3 3512% 2377% 1256% 6814 7990 9373
Risk G oup 4 3756% 3101% 1574% 10060 11331 13303
Risk G oup 5 062% 3992% 3194% 15524 17578 21068
No e: The able shows in o ma ion on disabili y isk and p i a e DI p emiums by isk g oup. Risk g oups a e assigned by he
insu e o indi iduals based on hei occupa ions. See Appendix Table A1 o equen occupa ions in each isk g oup. Column (1)
shows he sha e o each isk g oup ou o he labo o ce. Column (2) shows he ac ion o indi iduals e e claiming public DI bene i s
in each g oup. Column (3) shows he sha e o own-occupa ion DI claims ou o all DI claims. Columns (4) o (6) show he mon hly
p emium (in EUR) cha ged o an indi idual insu ing EUR 1000 o mon hly bene i s by con ac s a age, o a ixed con ac end age
o 65. The in o ma ion in columns (1) o (3) is om ou adminis a i e da a, and p emiums a e based on he insu e mic oda a.
g oup 3, and 40% in isk g oup 5.9Own-occupa ion disabili y isk is s ongly posi i ely
co ela ed wi h o e all disabili y isk and inc eases e en as e ac oss isk g oups. Fo in-
s ance, 8% o 11% o all disabili y cases in isk g oups 1 and 2 a e due o own-occupa ion
disabili y, bu he sha e inc eases o 32% o isk g oup 5. Finally, we no e ha isk g oups
di e in size, whe e isk g oup 1 and especially isk g oup 5 make up a smalle sha e o
he labo o ce han he middle g oups.
O he Sa e y Ne P og ams. In he Ge man con ex , a numbe o o he sa e y ne p o-
g ams a e po en ially a ailable o indi iduals who a e unable o wo k. Mos impo an ly,
means- es ed social assis ance (A bei slosengeld II) pays a la bene i o indi iduals a isk
o po e y, which could p o ide an al e na i e o DI bene i s o some wo ke s. O he
ans e p og ams, including unemploymen , sickness, and acciden insu ance, ha e a
maximum bene i du a ion o up o 2 yea s, and hus canno subs i u e DI o pe ma-
nen ly disabled wo ke s. Since a la ge wel a e e o m in 2005, social assis ance bene i s
ha e been ela i ely low; o ins ance, he a e age bene icia y ecei ed EUR 469 pe
mon h in 2015. Social assis ance is also subjec o a s ic means es a he household
le el: households owning any asse s wo h mo e han EUR 10k a e ineligible, and ans-
e s a e wi hd awn a a a e o 80% o household income exceeding EUR 100 pe mon h.
Mo eo e , bene i eceip equi es ac i ely sea ching o a job. Thus, social assis ance is
unlikely o be a meaning ul subs i u e o DI o mos wo ke s, excep pe haps o hose
in e y low-income, low-asse households. We e u n o his discussion in Sec ions 3.2 and
4.2, whe e we show ha he a ailabili y o social assis ance canno explain much o ou
esul s.
9No e ha we calcula e disabili y isk among coho s 1960 and olde , who a e obse ed unde ull public
DI co e age, including own-occupa ion DI. Thus, ou isk measu e is no con ounded by endogenous claiming
esponses o he 2001 e o m (see also Sec ion 5).
PRIVATIZING DISABILITY INSURANCE 1711
he e o m cu o . Only ou baseline con ol g oup exhibi s a e y small inc ease ela-
i e o olde coho s, bu he e a e no di e en ial ends in insu ance pu chases be ween
coho s u he below he cu o .
Quan i ying C owding-ou . Ou di e ence-in-di e ence analysis e eals a signi ican
impac o he 2001 e o m on p i a e DI pu chases. To unde s and how his maps in o
c owding-ou e ec s on he le el o p i a e DI ake-up, we can pe o m a simple back-o -
he-en elope calcula ion. We compu e he p edic ed numbe o con ac s held by coho s
1961–1962 in 2015 based on p e- e o m mean pu chases and add he cumula i e causal
impac o e he pos - e o m pe iod implied by ou es ima es. This esul s in a 26% in-
c ease in he s ock o p i a e DI con ac s held by he baseline ea men g oup who we e
ea ed a ages 39 o 40. Pe o ming a simila calcula ion among he ull se o ea ed
coho s om Figu e 3sugges s a subs an ially la ge ise in a e age p i a e DI ake-up
by 194%. Applying his es ima e o he p e- e o m ake-up a e om Figu e 1yields a
c owding-ou e ec o 18 pe cen age poin s, simila o he obse ed inc ease in o e all
ake-up. This implies ha much o he g ow h o he p i a e ma ke can be a ibu ed o a
causal e ec o he e o m, while con i ming ha he magni ude o c owding-ou be ween
public and p i a e DI is modes .
Al e na i e S a egy: Reg ession Discon inui y Design. As an al e na i e empi ical s a -
egy, he coho cu o o he e o m could be exploi ed o implemen a eg ession discon i-
nui y design (RDD). Ou di e ence-in-di e ence s a egy has he ad an age o p o iding
highe s a is ical powe , pa icula ly o subg oup analyses. Ne e heless, Appendix Fig-
u e A2 shows ha an RDD would yield esul s simila o ou main speci ica ion. The
igu e depic s a sha p jump in he numbe o p i a e DI con ac s held by indi iduals p e-
cisely a he Janua y 1961 cu o . The RDD es ima e is highly signi ican and co esponds
o an inc ease o 25% ela i e o a e age ake-up among con ol coho s. This is almos
pe ec ly in line wi h he 26% inc ease in he s ock o p i a e DI con ac s implied by ou
baseline di e ence-in-di e ence esul .
The Wel a e Re o m o 2005. As we explained in Sec ion 2.1, ano he impo an sa e y
ne e o m occu ed in 2005. The e o m made social assis ance subs an ially less gen-
e ous, eplacing he p io income-dependen wel a e sys em wi h a lowe , la bene i .
Mo eo e , s ic means es ing was in oduced, and bene i eceip was made condi ional
on ac i e job sea ch (see, e.g., B adley and Kuegle (2019)). Fo ou pu poses, he 2005
e o m p esen s a aluable oppo uni y o es he po en ial ole o social assis ance in de-
e mining p i a e DI ake-up. I social assis ance se ed as a subs i u e o p i a e DI, one
migh expec p i a e DI ake-up o inc ease a e 2005, when bene i le els we e educed
and many wo ke s became ineligible o social assis ance. Howe e , Figu e 1sugges s
ha much o he g ow h in p i a e DI ake-up occu s igh a ound 2001, wi h no u he
change a e 2005. Simila ly, Figu e 2p o ides no indica ion ha ea ed coho s inc ease
p i a e DI pu chases a e 2005. To es his mo e o mally, Appendix Table A5 pe o ms
a di e ence-in-di e ence es ima ion a ound he 2005 e o m, esul ing in a small and
insigni ican e ec on p i a e DI pu chases. This sugges s ha he gene osi y o social
assis ance is no a majo d i e o p i a e DI ake-up.
4. HETEROGENEITY IN PRIVATE DI TAKE-UP
In his sec ion, we s udy which ypes o indi iduals ake up p i a e DI. The main chal-
lenge in doing so is ha comp ehensi e mic oda a on he o e all p i a e DI ma ke is no

1712 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
a ailable. This challenge is aced by much o he li e a u e in es iga ing p i a e insu ance
ma ke s, which o en uses da a om a speci ic insu e o employe (e.g., Finkels ein and
Po e ba (2004,2014), Eina , Finkels ein, and Cullen (2010), Au o , Duggan, and G u-
be (2014), Cab al and Cullen (2019)). We ollow a simila app oach and eso o he
insu e mic oda a. Supplemen al Appendix B.1 desc ibes in de ail how we use his da a
in combina ion wi h adminis a i e da a and o icial social insu ance s a is ics o calcula e
p i a e DI ake-up o subg oups. Impo an ly, ou app oach equi es he assump ion ha
he ma ke sha e o he insu e is cons an ac oss subg oups (wi hin con ac ype and
yea ). This assump ion is no innocuous and i s alidi y hinges on how ep esen a i e he
insu e is o he o e all ma ke . In Sec ion 2.3, we a gued ha he insu e e lec s he
ma ke well in e ms o con ac design, p icing, occupa ional co e age, and geog aphic
co e age. Mo eo e , we p esen comp ehensi e alida ion checks o he esul ing ake-up
a es using ep esen a i e household su ey da a and o he independen da a sou ces in
Sec ion 4.3.
4.1. P i a e DI Take-up Ac oss G oups
In Figu e 4, we begin by p o iding desc ip i e e idence on he e ogenei y in p i a e DI
ake-up in 2015. Panel (a) displays a s ong posi i e co ela ion be ween p i a e DI ake-
up and income. In he op income quin ile, almos wo- hi ds (65%) o indi iduals ha e
p i a e DI. P i a e DI ake-up is 30% in he ou h quin ile, 11% o 12% in he second
and hi d quin iles, and only 7% in he bo om quin ile. Panel (b) shows a simila g adien
o p i a e DI ake-up by educa ion. 80% o indi iduals in he highes educa ion quin ile
ha e p i a e DI, while ake-up is 26% in he ou h quin ile and only 5% o 8% in he
bo om h ee quin iles.
Panel (c) depic s p i a e DI ake-up by ype o occupa ion. We use he ask-based clas-
si ica ion by Dengle , Ma hes, and Paulus (2014) who adap he me hod o Au o , Le y,
and Mu nane (2003) and dis inguish be ween cogni i e (analy ical o in e ac i e) e -
sus manual asks and ou ine e sus non ou ine asks. Wi h 77% and 45%, espec i ely,
wo ke s in occupa ions pe o ming analy ical o in e ac i e non ou ine asks display he
highes p i a e DI ake-up a es. In cogni i e ou ine occupa ions, 20% o indi iduals
ha e p i a e DI, while ake-up in manual occupa ions is only 8%.
Panel (d) inally in es iga es p i a e DI ake-up by isk g oup. Recall ha he insu e
assigns indi iduals o one o i e isk g oups based on occupa ions, and hese isk g oups
a e he p ima y de e minan o p i a e DI p emiums. We ind ha lowe - isk wo ke s
acing lowe insu ance p emiums a e much mo e likely o pu chase p i a e DI. In he
lowes - isk g oups 1 and 2, 58% and 56% o indi iduals ha e p i a e DI, espec i ely.
Among isk g oup 3, p i a e DI ake-up is 20%, and only 7% o 8% o indi iduals in isk
g oups 4 and 5 ha e p i a e DI.
He e ogeneous Impac o he 2001 Re o m. To unde s and how he 2001 e o m a ec s
p i a e DI ake-up ac oss g oups, we p o ide wo addi ional pieces o e idence. Fi s ,
Appendix Figu e A3 shows p i a e DI ake-up a es be o e he e o m (in 1997). Di e -
ences in ake-up be o e he e o m a e quali a i ely simila o Figu e 4, bu much less
p onounced. Fo ins ance, 17% o 18% o indi iduals in he op wo income quin iles
ha e p i a e DI, compa ed o 6% o 9% in he bo om h ee quin iles. Second, Appendix
Figu e A4 shows esul s om a causal analysis es ima ing he e ogeneous e ec s o he
2001 e o m. Fo his pu pose, we epea he di e ence-in-di e ence es ima ion om
equa ion (1) sepa a ely o each subg oup. To inc ease s a is ical powe , we ex end he
PRIVATIZING DISABILITY INSURANCE 1713
FIGURE 4.—P i a e DI Take-Up ac oss G oups. No es: The igu e shows p i a e DI ake-up a es in 2015 by
income quin ile (panel a), educa ion quin ile (panel b), ype o occupa ion (panel c), and isk g oup (panel
d). In panel (b), educa ion is de ined as yea s o schooling. In panel (c), we use he ask-based classi ica ion
by Dengle , Ma hes, and Paulus (2014) o g oup occupa ions. Take-up a es a e calcula ed among all coho s;
see Supplemen al Appendix B.1 o de ails.
coho window used in he es ima ion o 1957–1964. The ela i e e ec s by subg oup a e
e y simila o he desc ip i e pa e ns om Figu e 4. Fo ins ance, we es ima e ha he
impac on p i a e DI pu chases o he op income quin ile is abou double ha o he
ou h quin ile, while e ec s o he bo om h ee quin iles a e e y small. The impac
o he e o m on ake-up also s ongly inc eases wi h educa ion and non ou ine occupa-
ions, bu dec eases wi h isk g oups. These esul s sugges ha he causal impac o he
2001 e o m by subg oup is closely in line wi h he obse ed pos - e o m he e ogenei y in
p i a e DI ake-up.19
Mul i a ia e He e ogenei y. Ou main he e ogenei y analysis shows ha p i a e DI
ake-up exhibi s s ong uncondi ional co ela ions wi h income, educa ion, ype o oc-
19In addi ion, Appendix Table A2 shows summa y s a is ics o p i a e DI con ac s by ime o pu chase.
Cha ac e is ics o p e- e sus pos - e o m buye s align wi h he he e ogenei y analysis p esen ed he e. Indi-
iduals who ook up p i a e DI be ween he i s e o m announcemen and i s inal implemen a ion (1998 o
2000) display simila cha ac e is ics o p e e o m buye s mo e b oadly.
1714 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
FIGURE 5.—De e minan s o P i a e DI Take-Up. No es: The igu e displays esul s om eg essions o p i-
a e DI ake-up a es on indi idual cha ac e is ics a he h ee-digi occupa ion le el. The coe icien s show
he es ima ed impac o he espec i e a iable wi h 95% con idence in e als. “Basic con ols” a e he se o
cha ac e is ics om he igu e, ha is, log income, educa ion (yea s), indica o s o cogni i e and non- ou ine
occupa ions, and isk g oups. “Ex ended con ols” include hese a iables plus gende , ma i al s a us, an indi-
ca o o economic aining and esidence in Eas Ge many. See Appendix Table A7 o ull eg ession esul s.
cupa ion, and p iced isk g oups. To u he unpack his he e ogenei y, Figu e 5displays
esul s om eg essions o p i a e DI ake-up on hese cha ac e is ics, con olling o
a ying se s o indi idual and occupa ional cha ac e is ics. One key esul om hese
mul i a ia e speci ica ions is ha income is no a signi ican de e minan o p i a e DI
ake-up once educa ion and o he cha ac e is ics a e con olled o . Thus, while he un-
condi ional ake-up g adien by income is impo an o dis ibu ional conside a ions (see
Sec ion 7.2), income pe se does no seem o d i e he e ogeneous p i a e DI demand. On
he con a y, Figu e 5shows ha p i a e DI ake-up emains signi ican ly co ela ed wi h
educa ion, isk g oups, and wo king in a non ou ine occupa ion in all speci ica ions.
4.2. P i a e DI Take-up, Disabili y Risk, and Risk P o ec ion Bene i
Bo h desc ip i e pa e ns and di e ence-in-di e ence esul s sugges ha p i a e DI
ake-up dec eases wi h isk g oups. This aises a po en ial puzzle, as, in p inciple, one
would expec highe - isk indi iduals o place a highe alua ion on insu ance. No e ha
isk- a ed p emiums alone canno explain his inding. P i a e DI p emiums inc ease
ac oss isk g oups in a manne no a om ac ua ially ai . Thus, while one may no nec-
essa ily expec p i a e DI ake-up o inc ease wi h isk g oups, he p icing scheme canno
explain he dec easing pa e n.20 To explo e wha could explain low p i a e DI ake-up
among high- isk wo ke s, we calib a e a comp ehensi e measu e o he po en ial insu -
ance alue o p i a e DI. Ou isk p o ec ion bene i measu e akes in o accoun a a ie y
20This poin is closely ela ed o ou es o isk-based selec ion in Sec ion 5.Oncewecon ol o insu ance
p ices, we ind ha he co ela ion be ween p i a e DI ake-up and disabili y isk becomes la bu does no
u n posi i e, as one would expec unde s anda d models o ad e se selec ion.
PRIVATIZING DISABILITY INSURANCE 1715
o ac o s beyond li e ime disabili y isk, including he e ogenei y in he iming and sub-
isk composi ion o disabili y e en s, o he sa e y ne p og ams, and consump ion d ops
upon disabili y.
Building on exis ing app oaches in he insu ance li e a u e, pa icula ly Mi chell,
Po e ba, Wa shawsky, and B own (1999)andFinkels ein and McKnigh (2008), we se
ou an expec ed li e ime u ili y amewo k in Supplemen al Appendix B.3. In his ame-
wo k, indi iduals ace some isk o disabili y in each pe iod, and hey may quali y o
public DI bene i s o social assis ance in case o disabili y. We use he model o calib a e
he ce ain y equi alen u ili y gain ha isk-a e se indi iduals would de i e om p i a e
DI. To pe o m he calib a ions, we combine he in o ma ion on dynamic disabili y isk
pa hs, income, and p i a e DI pa ame e s con ained in ou a ious da a sou ces, and we
ely on a ange o assump ions abou consump ion d ops upon disabili y and he p oba-
bili y o ecei ing public DI bene i s.21 In all speci ica ions, indi iduals can ecei e basic
social assis ance i hei income is su icien ly low. We ini ially assume ha consump ion
losses and isk p e e ences a e homogeneous ac oss g oups, bu elax hese assump ions
la e . Figu e 6displays he calib a ed isk p o ec ion bene i by isk g oup unde selec ed
speci ica ions, along wi h p i a e DI ake-up and li e ime disabili y isk. Full calib a ion
esul s a e shown in Appendix Table A8.
FIGURE 6.—P i a e DI Take-Up, Disabili y Risk, and Risk P o ec ion Bene i . No es: The igu e shows p i a e
DI ake-up a es, disabili y isk, and he isk p o ec ion bene i o p i a e DI by isk g oup. The e ical ba s
depic p i a e DI ake-up a es o each g oup, calcula ed among all coho s in 2015. The solid line shows
li e ime disabili y isk by g oup. The dashed lines show he isk p o ec ion bene i o p i a e DI as a ac ion o
income unde h ee selec ed calib a ion scena ios, namely (i) unde ou baseline assump ions, (ii) in a scena io
whe e income losses upon own-occupa ion disabili y inc ease somewha p og essi ely wi h income, and (iii) in
a scena io wi h ex emely p og essi e income losses. See Appendix Table A8 o de ailed calib a ion esul s.
21Since no es ima es o consump ion losses upon disabili y a e a ailable in he Ge man con ex , we ely on
esul s om he U.S. by Meye and Mok (2019) and esul s om Denma k by Humlum, Munch, and Pla o
(2025) in he calib a ion.
1716 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
Unde he baseline calib a ion scena io, which assumes a coe icien o ela i e isk
a e sion o h ee, we ind ha he expec ed u ili y gain om p i a e DI inc eases mono-
onically om 3.2% o income o isk g oup 1 o 19% o isk g oup 5. These ela i e
magni udes a e oughly aligned wi h he simple isk measu e, which implies ha aking
in o accoun ull dynamic disabili y isk pa hs and he composi ion o claims does no
subs an ially change he ela i e u ili y gain om p i a e DI.22 In Appendix Table A8, he
g adien o he isk p o ec ion bene i ac oss g oups emains simila unde al e na i e
assump ions abou isk a e sion, public DI ejec ions, and he size o consump ion losses.
Nex , he a ailabili y o o he sa e y ne p og ams has been shown o play a ole o
insu ance choices, o example, in long- e m ca e insu ance (B aun, Kopecky, and Ko-
eshko a (2019)). In ou se ing, wo key p og ams could ma e o p i a e DI ake-up.
Fi s , basic social assis ance pays a la bene i , which could p o ide some implici in-
su ance pa icula ly o high- isk g oups, gi en hei lowe a e age income (see Fische ,
Geye , and Zieba h (2024)). In Figu e 6, he calib a ed isk p o ec ion bene i indeed
inc eases less s eeply han he simple isk measu e o he highes isk g oups 4 and 5, bu
his e ec is quan i a i ely small. A limi ed ole o social assis ance is in line wi h he low
implici eplacemen a e o he p og am and wi h he e idence om Sec ion 3.2,whe e
we show ha a la ge wel a e e o m ha dly a ec s p i a e DI ake-up. Second, he a ail-
abili y o public DI bene i s could in luence p i a e DI ake-up. Howe e , pos - e o m
public DI only co e s gene al disabili y isk and pays an app oxima ely cons an eplace-
men a e o all g oups. Thus, i is unlikely o induce di e en ial p i a e DI demand. This
is con i med by Appendix Table A8, whe e a ying he p obabili y o quali ying o public
DI ( ejec ion a es) ha dly impac s he ela i e bene i o p i a e DI ac oss isk g oups.
Ou baseline calib a ion assumes ha consump ion losses upon disabili y a e a cons an
pe cen age o income o all isk g oups. Howe e , he consequences o own-occupa ion
disabili y may a y. In pa icula , low- isk occupa ions a e mo e likely o pe o m cog-
ni i e non ou ine asks, which end o equi e specialized skills (see Dengle , Ma hes,
and Paulus (2014)). I is hus possible ha wo ke s in low- isk g oups expe ience la ge
ea nings losses i an own-occupa ion disabili y necessi a es swi ching o a di e en oc-
cupa ion.23 Un o una ely, we do no know o any exis ing es ima es o he e ogeneous
consump ion losses upon disabili y ac oss occupa ions o isk g oups. Hence, we conside
wo calib a ion scena ios o examine his channel. As a benchma k, we make he ex eme
assump ion ha in he e en o own-occupa ion disabili y, all indi iduals a e only able o
wo k in a basic low-skill occupa ion whe e hey ea n he a e age bo om-quin ile income.
We also conside an in e media e scena io in be ween his ex eme case and ou base-
line speci ica ion. Figu e 6shows ha such “p og essi e” ea nings losses can subs an ially
modi y he ela i e bene i o p i a e DI. Unde he in e media e scena io, he isk p o-
ec ion bene i inc eases less s eeply wi h isk g oups, and unde he ex eme scena io, i
becomes i ually la .
Finally, obse ed p i a e DI ake-up pa e ns could be explained by he e ogeneous isk
p e e ences. In pa icula , high- isk wo ke s may ha e lowe isk a e sion. One po en ial
eason behind a nega i e co ela ion be ween disabili y isk and isk a e sion could be
ha wo ke s selec in o mo e o less isky occupa ions based on hei isk ole ance. To
22Two key calib a ion inpu s illus a e why his occu s. Appendix Figu e A5 shows ha he iming o disabili y
e en s is simila ac oss isk g oups. Table Ishows ha , in addi ion o o e all disabili y isk, he sha e o own-
occupa ion claims also inc eases wi h isk g oups.
23This would be consis en wi h e idence om he li e a u e on ea nings losses upon job displacemen . Fo
ins ance, Huck eld (2022) documen s pa icula ly la ge losses o wo ke s wi h speci ic human capi al who
ind eemploymen in a di e en occupa ion wi h lowe skill equi emen s.

PRIVATIZING DISABILITY INSURANCE 1717
assess whe he plausible a ia ion in isk p e e ences could explain he e ogeneous p i a e
DI ake-up, we calib a e he deg ee o isk a e sion ha would make indi iduals indi e -
en be ween li e ime expec ed u ili y wi h and wi hou p i a e DI a ma ke p emiums.
This yields an implied coe icien o ela i e isk a e sion o he ma ginal buye in each
isk g oup. De ails o he calib a ion a e shown in Supplemen al Appendix B.4, and e-
sul s a e displayed in Appendix Table A8. Ac oss calib a ion scena ios wi h p opo ional
consump ion d ops, he implied coe icien o isk a e sion is be ween 1.09 and 4.96 o
isk g oup 1 and be ween 0.66 and 2.87 o isk g oups 4 and 5. E en when allowing o
p og essi e ea nings losses, implied isk a e sion o highe - isk g oups emains below ha
o isk g oup 1.24 This sugges s ha lowe isk a e sion among high- isk g oups could in-
deed explain hei lack o p i a e DI demand. Ou isk a e sion es ima es o hese g oups
a e low bu no implausibly a om esul s in he li e a u e.25
Implica ions o Wel a e Analysis. Ou baseline calib a ions sugges ha , gi en he size
and s uc u e o disabili y isk and he low le el o social assis ance in he Ge man con-
ex , wo ke s in high- isk g oups should bene i subs an ially om p i a e DI. Howe e ,
low ake-up among hese wo ke s could be explained by he e ogeneous ea nings losses
upon own-occupa ion disabili y o by he e ogeneous isk p e e ences. A key ad an age
o he e ealed p e e ence app oach we ollow la e on is ha he unde lying easons be-
hind indi idual choices will no ma e o he wel a e implica ions, as long as obse ed
p i a e DI ake-up e lec s wo ke s’ ue insu ance alua ions. Besides he ac o s dis-
cussed abo e, beha io al biases could be ano he po en ial eason o low insu ance ake-
up among ce ain g oups. S udying such biases is ou side he scope o his pape . We
b ie ly e u n o his discussion in ou conclusion.
4.3. Valida ion Checks
Ou empi ical esul s on he e ogenei y ely on he insu e mic oda a, as indi idual-
le el da a on he en i e ma ke is no a ailable. As we discussed be o e, he alidi y o
hese indings depends on how ep esen a i e he insu e is o he o e all ma ke . In his
sec ion, we p esen se e al alida ion checks using addi ional, independen da a sou ces.
To begin wi h, o e all p i a e DI ake-up in ou da a is e y simila o es ima es om
o he sou ces. A su ey conduc ed by TNS In a es (2015), a p i a e su ey company,
inds ha 26% o wo king adul s had p i a e DI in 2015, co esponding p ecisely o ou
main ake-up a e es ima e o he same yea om Sec ion 3.1. In he ep esen a i e
EVS su ey da a, o e all p i a e DI ake-up by Ge man households was 31% in 2013.
Household-le el ake-up is na u ally highe han ou indi idual-le el es ima e since he
a e age household has a ound wo membe s (see Appendix Table A4).
O he da a sou ces also sugges quali a i ely simila p i a e DI ake-up pa e ns by
subg oups. Appendix Figu e A6 shows ha ake-up a es clea ly inc ease wi h income in
he household su ey, albei wi h a somewha la e g adien . We ma ch ake-up a es
by gende well, conside ing ha he su ey igu es a e a he household le el. To alida e
24In e es ingly, calib a ed isk a e sion does no dec ease mono onically wi h isk g oups. In pa icula , isk
a e sion o he ma ginal buye in g oup 5 ends o be highe han in g oups 2 o 4. This likely occu s due
o a combina ion o social assis ance p o iding mo e sizable implici insu ance and ela i ely high p i a e DI
p emiums cha ged o his g oup.
25S udies on insu ance choices ypically yield es ima es o ela i e isk a e sion anging be ween 1 and 8
(e.g., F ench (2005), Lockwood (2018), Jacobs (2023), Landais and Spinnewijn (2021)). Some wo k implies
la ge alues (e.g., Cohen and Eina (2007), Sydno (2010)).
1718 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
p i a e DI ake-up a es by isk g oup, we use he a ing agency da a, which includes a
b eakdown by “ha monized” isk g oups o he en i e ma ke . This in o ma ion is based
on insu e s epo ing he numbe o con ac s in ou isk g oups de ined by he a ing
agency. Ha monized g oups co espond oughly o he isk g oups used by he insu e
p o iding ou mic oda a, bu he insu e addi ionally di e en ia es among he highes
isks. Ou main ake-up es ima es o he la ges isk g oups 2 and 3 a e i ually he
same as hose om he a ing agency da a. Fo he lowes and highes - isk g oups, he
a ing agency da a displays e en s onge he e ogenei y in ake-up han ou main esul s.
Finally, as an addi ional piece o e idence, we p esen a compa ison o p i a e DI p ic-
ing by di e en insu e s. Fo his exe cise, we collec da a on p ices cha ged o he 10
mos equen occupa ions in each isk g oup o hose o he op 10 insu e s o e ing
online p ice calcula o s. Appendix Figu e A6 plo s he a e age mon hly p emium by isk
g oup o he insu e p o iding ou mic oda a and ou la ge compe i o s. Rela i e p ices
cha ged o di e en occupa ions a e ai ly simila ac oss insu e s. This sugges s ha di -
e en isk g oups should ha e li le eason o selec speci ically in o he insu e p o iding
ou mic oda a, as i s p icing e lec s he o e all ma ke .
5. RISK-BASED SELECTION: POSITIVE CORRELATION TEST
A c ucial ques ion o he e icien unc ioning o p i a e DI ma ke s is whe he and
how indi iduals selec in o pu chasing insu ance based on hei disabili y isk. S anda d
ad e se selec ion models p edic ha high- isk indi iduals a e mo e likely o pu chase
insu ance, which can lead o unde p o ision o insu ance o e en comple e ma ke un-
a eling (Ake lo (1970), Ro hschild and S igli z (1976)). To o mally es o isk-based
selec ion, we implemen a posi i e co ela ion es (Chiappo i and Salanié (2000)) in his
sec ion. La e , we p o ide addi ional e idence om a cos cu e es (see Sec ion 6.2).
Ou aim is o es o a posi i e co ela ion be ween p i a e DI ake-up and disabili y
isk, which would indica e ad e se selec ion. Speci ically, we un he ollowing eg ession:
Qj=β0+β1πj+
5
∑︂
k=2
γk1( isk g oupj=k)+ϵj(2)
whe e Qjdeno es p i a e DI ake-up o indi iduals in h ee-digi occupa ion jin 2015,
πjis li e ime disabili y isk measu ed in adminis a i e da a, and 1( isk g oupj=k)isan
indica o o occupa ion jbeing assigned o isk g oup kby he insu e .26 No e ha we
de ine πjas o al isk (including bo h own-occupa ion DI and gene al DI op-up payou s),
as his is he ele an isk measu e om he pe spec i e o insu e cos .
Ou se ing and da a enable us o add ess wo key challenges o en aced by simila
posi i e co ela ion es s in he li e a u e. Fi s , in assessing whe he he e is ad e se se-
lec ion, i is impo an o es ima e he co ela ion be ween p i a e DI ake-up and isk
wi hin g oups o indi iduals acing he same insu ance p ices. Indeed, we ound a s ong
nega i e co ela ion be ween p i a e DI ake-up and isk g oups in he p e ious sec ion.
Risk g oups e lec an obse ed componen o isk based on which insu ance con ac s
a e p iced. In equa ion (2), we con ol lexibly o p ices by including a se o isk g oup
dummies, such ha we can in e p e β1as cap u ing selec ion on unp iced isk. Second,
26Mo e p ecisely, we measu e isk g oupjas he modal isk g oup in an occupa ion. Risk g oups a e no nec-
essa ily he same o all indi iduals wi hin a h ee-digi occupa ion in he da a because he insu e some imes
changes isk g oup assignmen o e ime (see Sec ion 6.2).
PRIVATIZING DISABILITY INSURANCE 1719
FIGURE 7.—Risk-Based Selec ion: Posi i e Co ela ion Tes . No es: The igu e shows a binned sca e plo
depic ing he co ela ion be ween p i a e DI ake-up and esidual (unp iced) disabili y isk, con olling o
p i a e DI p emiums. We un he eg ession a he le el o h ee-digi occupa ions. As explained in Sec ion 5,
ake-up a es a e calcula ed among ea ed coho s in 2015, while disabili y isk is measu ed only among con ol
coho s. The igu e also includes he es ima ed slope coe icien bwi h i s s anda d e o in pa en heses. See
Appendix Table A9 o de ails o he co esponding eg ession esul s.
a well-known di icul y wi h he posi i e co ela ion es is ha ex pos measu es o isk
based on obse ed insu ance claims may con ound selec ion on ex an e isk and mo al
haza d esponses (see, e.g., Landais e al. 2021). A co ela ion o DI ake-up and ex pos
claiming p obabili ies may be d i en by ce ain isk ypes selec ing in o insu ance (se-
lec ion) o hose wi h mo e insu ance co e age becoming mo e likely o claim (mo al
haza d). To add ess his challenge and isola e isk-based selec ion, we calcula e ake-up
among he ea ed coho s 1961 and younge bu measu e disabili y isk πjas he ac ion
claiming DI only among he con ol coho s 1960 and olde . This isk measu e should no
be con ounded by mo al haza d esponses o di e en ial ake-up among ea ed coho s
since all indi iduals in he con ol coho s a e obse ed unde ull public DI co e age.
Figu e 7depic s he es ima ion esul s in a binned sca e plo . The e is conside able
esidual unp iced a ia ion in disabili y isk along he ho izon al axis. C ucially, he es-
ima ed ela ionship be ween occupa ion-le el p i a e DI ake-up and unp iced isk is
qui e la , and he es ima ed slope coe icien co esponding o β1in equa ion (2)issmall
and s a is ically insigni ican . In o he wo ds, we do no ind any ad e se selec ion om
he poin o iew o he insu e : wi hin p iced isk g oups, indi iduals wi h highe isk a e
no mo e likely o selec in o pu chasing insu ance. The poin es ima e on isk is nega i e,
which would imply sligh ly ad an ageous selec ion in o p i a e DI, i any hing.
A i s glance, he lack o ad e se selec ion may seem su p ising, as insu ance should
in p inciple be mo e aluable o highe - isk indi iduals. Howe e , some o he ac o s we
discussed in Sec ion 4.2 could explain his empi ical inding. E idence om o he insu -
ance ma ke s poin s o he e ogenei y in non isk ela ed componen s o indi idual insu -
ance alua ions. I hese co ela e nega i ely wi h unde lying isk, hey can undo po en ial
ad e se selec ion (e.g., Finkels ein and McGa y 2006,Cu le , Finkels ein, and McGa y
2008). Ou inding ha high-income and highly educa ed indi iduals a e mo e likely o
pu chase p i a e DI could, o ins ance, e lec he e ogeneous isk a e sion o ea nings
1720 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
losses ha a e nega i ely co ela ed wi h isk. Appendix Table A9 p esen s u he sugges-
i e e idence along hese lines by explo ing how isk-based selec ion changes condi ional
on di e en se s o obse ables. In pa icula , we ind ha he coe icien u ns posi i e
(bu emains insigni ican ) once we con ol o educa ion. This sugges s ha no condi-
ioning p ices on educa ion induces some ad an ageous selec ion, coun e ing po en ial
ad e se selec ion wi hin isk g oups. Hence, e en hough i is somewha coa se, he p ic-
ing scheme de ised by insu e s seems o be an impo an ac o in p e en ing ad e se
selec ion.27
6. VALUE AND COST OF DI
6.1. Basic Concep ual F amewo k
Nex , ou aim is o quan i y he alue and cos o he DI co e age o e ed by he p i-
a e ma ke . These wo a e he main componen s en e ing ou wel a e analysis la e on.
Ou concep ual app oach builds on Eina , Finkels ein, and Cullen (2010), who show ha
in o de o e alua e wel a e in insu ance ma ke s, he key su icien s a is ics a e gi en by
insu ance demand and cos cu es. Simila amewo ks ha e been used in a numbe o so-
cial insu ance se ings (Hend en, Landais, and Spinnewijn (2021), Eina and Finkels ein
(2023)).
We conside a popula ion o he e ogeneous indi iduals indexed by θi,whe eF(θi)de-
no es he dis ibu ion o ypes. He e ogenei y may include a ia ion bo h in indi idual
disabili y isk and in o he ac o s in luencing DI alua ions, such as isk a e sion. The
i s key componen o ou wel a e analysis is demand o DI. Deno e by (θi) he u ili y
o consume i om DI, and by pk he insu ance p emium cha ged o indi iduals in isk
g oup k. In a p i a e ma ke wi h insu ance choice, he indi idual pu chases DI i hei
willingness o pay exceeds he p emium, (θi)≥pk. Agg ega e demand o p i a e DI in
g oup kcan be w i en as
Dk(pk)=∫︂1(︁ (θ)≥pk)︁dFk(θ)=P k(︁ (θi)≥pk)︁
The second componen we equi e is he cos o p o iding DI. We deno e by c(θi) he
expec ed cos o insu ing indi idual i’s isk. The a e age cos a p ice pkis
ACk(pk)=1
Dk(pk)∫︂c(θ)1(︁ (θ)≥pk)︁dFk(θ)=𝔼k(︁c(θi)| (θi)≥pk)︁
In addi ion, we can w i e ma ginal cos as MCk(pk)=𝔼k(c(θi)| (θi)=pk).
Be o e we p oceed o he empi ical implemen a ion, h ee aspec s a e wo h no ing.
Fi s , we assume ha indi iduals make a disc e e choice whe he o buy insu ance, and we
abs ac om he choice o insu ed bene i amoun s in p i a e DI con ac s. This assump-
ion is mo i a ed by ou esul s om Sec ion 3.2, which sugges ha his ex ensi e ma gin
is he empi ically ele an dimension o insu ance choice. Second, ou main analysis ol-
lows he li e a u e ega ding he cos o p o iding DI and abs ac s om any o he cos
incu ed by insu e s. We discuss he po en ial ole o adminis a i e cos in Sec ion 7.3.
Thi d, since insu ance p ices depend on isk g oups o which insu e s assign indi iduals
27In line wi h his a gumen , Bundo , Le in, and Mahoney (2012) show ha isk- a ed p emiums can im-
p o e e iciency by limi ing isk-based selec ion in o heal h insu ance.
PRIVATIZING DISABILITY INSURANCE 1727
he insu ed ela i e o he cos o he insu e . Ne alue in he p i a e ma ke is
NVp i =∑︂
k
nk[︃∫︂ (θ)1(︁ (θ)≥pk)︁dFk(θ)]︃
∑︂
k
nk[︃∫︂c(θ)1(︁ (θ)≥pk)︁dFk(θ)]︃(6)
whe e nkdeno es he size o isk g oup k. In he p i a e ma ke , he ne alue is gi en by
he alue o hose who choose o pu chase DI, ha is, o whom (θ)≥pk, di ided by he
cos o p o iding DI o hem. Since we es ima e p i a e DI alua ions in he p esence o
baseline public DI co e age, NVp i shouldbe in e p e ed as he ne alue o he addi ional
co e age o e ed by he p i a e ma ke .
Ou main coun e ac ual o in e es is he in oduc ion o an insu ance manda e p o id-
ing he co e age o e ed by p i a e DI o all wo ke s.33 S a ing om he p i a e ma ke
equilib ium, he ne alue o in oducing he manda e is
NVmand =∑︂
k
nk[︃∫︂ (θ)1(︁ (θ)<p
k)︁dFk(θ)]︃
∑︂
k
nk[︃∫︂c(θ)1(︁ (θ)<p
k)︁dFk(θ)]︃(7)
Indi iduals wi h willingness o pay abo e he ma ke p ice al eady pu chased p i a e DI
when hey had he choice. Thus, a manda e expands co e age o hose indi iduals whose
willingness o pay is below he p ice.
Based on his wel a e measu e, a e o m is wel a e-imp o ing i i s ne alue is g ea e
han one; ha is, i gene a es alue exceeding i s cos . Fo ou coun e ac ual, NVmand >
1 would imply ha manda ing p i a e DI co e age is wel a e-imp o ing. In con as ,
NVmand <1 would mean ha lea ing his co e age o he olun a y p i a e ma ke is
p e e able.
Ou baseline wel a e analysis ollows Eina , Finkels ein, and Cullen (2010)and o-
cuses on he di ec alue and cos o p o iding ex a DI. Howe e , a ious ypes o in-
di ec e ec s could be associa ed wi h DI p o ision. To accoun o hese as much as
possible, we also calcula e a ma ginal alue o public unds (MVPF) measu e (Hen-
d en and Sp ung-Keyse (2020), Finkels ein and Hend en (2020)). To do his, we aug-
men equa ion(7) and include ele an iscal ex e nali ies in he denomina o .34 As a
i s ex e nali y, manda ing p i a e DI co e age is likely o impose indi ec mo al haz-
a d cos s on o public DI, since op-up insu ance in case he wo ke also quali ies o
33While he e migh be al e na i e policies o in e es , o example, manda ing pa s o p i a e DI co e age,
we ocus on his coun e ac ual o wo easons. Fi s , he ai es compa ison is a guably be ween wo policies
p o iding he same insu ance co e age, so i is na u al o conside a manda e o ac ual p i a e DI co e age.
Second, ou empi ical es ima es a e di ec ly ela ed o his coun e ac ual since we quan i y selec ion, insu ance
demand, and cos o he co e age p o ided by he p i a e DI ma ke . Thus, we conside a coun e ac ual based
on ac ual p i a e DI co e age he mos empi ically c edible.
34We con inue using he en i e es ima ed demand and cos cu es when calcula ing he MVPF, as in equa-
ions (6)and(7). This is impo an because ou coun e ac ual is a la ge e o m mo ing insu ance ake-up o
100%, such ha he app oxima ion o willingness o pay based on he en elope heo em o en used in MVPF
calcula ions would no be app op ia e.

1728 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
public (gene al) DI bene i s is included. To quan i y his channel, we use he es ima e
o Sei z (2021), who inds ha p i a e DI inc eases public DI claims by 8.6%. Sec-
ond, a iscal ex e nali y may a ise because addi ional DI claims en ail educ ions in
labo supply, which in u n lowe ax e enue. Un o una ely, empi ical es ima es o
he labo supply e ec s o DI a e no a ailable in he Ge man con ex , so we use an
a e age o es ima es om he li e a u e. Thi d, co e ing all wo ke s wi h p i a e DI
may educe hei p opensi y o claim social assis ance i hey become unable o wo k,
implying a posi i e iscal ex e nali y. Supplemen al Appendix B.5 p o ides ull de ails
on how we calib a e each channel based on ou da a and es ima es om he li e a-
u e.
Wel a e in he p i a e DI ma ke can be g aphically illus a ed using ou es ima ed
demand and cos cu es. Panel (a) o Figu e 10 depic s ne alue in he p i a e DI ma ke
o he case o isk g oup 3. The o al a ea unde he demand cu e up o equilib ium
ake-up co esponds o he nume a o in equa ion (6), and he a ea unde he ma ginal
cos cu e co esponds o he denomina o . In addi ion, he igu e shows a decomposi ion
o willingness o pay in o consume su plus (a ea A), p oduce su plus (B), and cos (C).
Appendix Figu e A8 shows analogous g aphs o all isk g oups. O e all, he p i a e DI
ma ke gene a es a la ge su plus, as indi iduals wi h he highes willingness o pay choose
o pu chase p i a e DI. Consume su plus is pa icula ly sizable in isk g oups 1 and 2,
whe e indi iduals exhibi he highes insu ance alua ions. P oduce s ecei e he la ges
su plus om isk g oups 1, 4, and 5, whe e ma kups o e ma ginal cos a e highes (see
Sec ion 7.3).
FIGURE 10.—Wel a e Calcula ions. No es: The igu e illus a es ou wel a e calcula ions o he case o isk
g oup 3. Panel (a) depic s wel a e in he p i a e DI ma ke , whe e he ne alue is gi en by he o al a ea unde
he demand cu e (A+B+C) di ided by he a ea unde he cos cu e (C). Panel (b) illus a es he ne alue
o a e o m manda ing p i a e DI co e age. The manda e inc eases DI ake-up om he ma ke equilib ium
o 1. The ne alue o he e o m is gi en by he addi ional a ea unde he demand cu e (D+G) di ided
by he addi ional cos (F+G). In bo h panels, he ne alue can be u he decomposed, as explained in he
espec i e legend. Appendix Figu es A8 and A9 show co esponding g aphs o all isk g oups.
PRIVATIZING DISABILITY INSURANCE 1729
TABLE IV
WELFARE EFFECTS OF INSURANCE MANDATES.
(1) (2) (3) (4)
P i a e DI Manda e
(Risk-Based P emiums)
Public DI Manda e
(Income-Based
Con ibu ions)
Ne Value MVPF Ne Value MVPF
Baseline alue 0.705 0.617 0.705 0.617
Social alue by isk a e sion σ
σ=1 0.634 0.553 1.215 1.057
σ=2 0.575 0.499 1.547 1.343
σ=3 0.527 0.455 1.749 1.518
σ=5 0.459 0.393 1.936 1.678
σ=8 0.395 0.333 2.010 1.742
No e: The able shows he wel a e e ec s o manda ing he DI co e age o e ed by he p i a e insu ance ma ke . Column (1)
p esen s ne alue es ima es o a p i a e DI manda e and column (3) p esen s es ima es o a ull public DI manda e inanced by
income-based social insu ance con ibu ions. Columns (2) and (4) p esen ma ginal alue o public unds (MVPF) es ima es o he
same manda es, aking in o accoun iscal ex e nali ies. The i s ow shows he baseline alue calcula ed as in equa ion (7). The
emaining ows show he social alue calcula ed as in equa ions (8)and(9), using wel a e weigh s om a U ili a ian social wel a e
unc ion unde di e en alues o he coe icien o ela i e isk a e sion σ.
Panel (b) o Figu e 10 illus a es he wel a e e ec s o a manda e, again o he case o
isk g oup 3. S a ing om he p i a e ma ke s a us quo, insu ing all indi iduals en ails
addi ional cos s gi en by he a ea unde he cos cu e be ween equilib ium ake-up and
manda ed ake-up o 100% (a eas F+G). Expanding insu ance yields addi ional alue
(D+G), bu his is exceeded by addi ional p emium paymen s (D+E+F+G), implying
a ne loss in consume su plus. Insu e s, on he o he hand, gain su plus (D+E). Thus,
he o e all ne alue o he manda e is gi en by D+Go e F+G, which is clea ly below
one. Appendix Figu e A9 shows ha he ne alue o a manda e is below one o all isk
g oups excep g oup 1.
The i s ow o Table IV quan i ies hese wel a e esul s. We ind an o e all ne alue
o in oducing a p i a e DI manda e o 0.705. This implies ha pa ial DI p i a iza ion,
simila o he 2001 e o m, is wel a e-imp o ing compa ed o a ull public DI manda e.
Simila ly, we ind an MVPF o a manda e o 0.617. This e lec s ha he join e ec o
iscal ex e nali ies inc eases he ne cos o p o iding ex a DI, ein o cing ou conclusion
ha he p i a e DI ma ke is wel a e-imp o ing.35
7.2. The Social Value o a DI Manda e
As an ex ension o he wel a e analysis, we in oduce dis ibu ional conce ns. Recall
ha he p i a e DI ma ke disp opo iona ely co e s high-income and low- isk indi idu-
als. A manda e would hus ex end co e age o mo e low-income and high- isk indi iduals,
on whom a social planne conce ned wi h equi y would place pa icula weigh . We w i e
35As a poin o compa ison, Hend en and Sp ung-Keyse (2020) epo simila MVPF es ima es o ma ginal
e o ms p o iding addi ional DI bene i s be ween 0.74 and 0.96.
1730 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
he social ne alue o in oducing a manda e as
SNVmand
=∑︂
k
nk[︃λk
Consume su plus
⏟⏞⏞ ⏟
∫︂(︁ (θ)−pk)︁1(︁ (θ)<p
k)︁dFk(θ)+
Insu e e enue
⏟⏞⏞ ⏟
∫︂pk1(︁ (θ)<p
k)︁dFk(θ)]︃
∑︂
k
nk[︃∫︂c(θ)1(︁ (θ)<p
k)︁dFk(θ)]︃(8)
The i s e m in he nume a o cap u es he addi ional ne u ili y indi iduals in isk g oup
kde i e unde a manda e, co esponding o hei alua ion minus he p ice o ex a DI.
The change in consume su plus among isk g oup kis mul iplied by λk, he social wel a e
weigh o indi iduals in his g oup. The second e m in he nume a o e lec s addi ional
e enue o he insu e . The social ne alue hen ela es he sum o hese wo e ms o he
addi ional cos o p o iding insu ance.36
Equa ion (8) conside s a p i a e insu ance manda e whe e indi iduals a e compelled
o pu chase p i a e DI a ma ke p ices. Howe e , in ou se ing, ex a DI co e age was
p o ided ia he social insu ance sys em be o e he 2001 e o m, whe e indi iduals a e
manda ed o pa icipa e and pay social insu ance con ibu ions. To e alua e such a public
DI manda e, we ha e o conside ha con ibu ions can di e om isk- a ed p i a e DI
p emiums pk. Fo mally, he social ne alue o a public DI manda e is gi en by
SNVpub
=(︃∑︂
k
nk{︃λk[︃∫︂(︁ (θ)−pk)︁1(︁ (θ)<p
k)︁dFk(θ)+
P icing e ec
⏟⏞⏞ ⏟
∫︂(︁pk−ppub
k)︁dFk(θ)]︃
+∫︂ppub
k1(︁ (θ)<p
k)︁dFk(θ)}︃)︃/︂(︃∑︂
k
nk[︃∫︂c(θ)1(︁ (θ)<p
k)︁dFk(θ)]︃)︃(9)
whe e ppub
kdeno es con ibu ions paid by indi iduals in isk g oup k.Compa ed oequa-
ion (8), a public DI manda e hus en ails an addi ional p icing e ec en e ing consume
su plus. Speci ically, we suppose ha con ibu ions a e le ied as a p opo ion o an in-
di idual’s g oss income, as is he case in ou se ing and o he social insu ance sys ems.
We calcula e he equi ed con ibu ion a e such ha o al con ibu ions equal he cos o
p o iding he ex a co e age o all indi iduals.
In o de o ob ain wel a e weigh s, we equi e a social wel a e unc ion. As is common
in he li e a u e, we assume U ili a ian social wel a e, such ha wel a e weigh s a e gi en
by ma ginal u ili y om consump ion. Mo eo e , we assume cons an ela i e isk a e -
sion u ili y. We calcula e social wel a e weigh s o each isk g oup based on expec ed
li e ime income. Appendix Table A12 shows in o ma ion on income and esul ing social
wel a e weigh s by g oup. Expec ed income dec eases mono onically wi h isk g oups. On
a e age, indi iduals in isk g oup 1 ea n mo e han double he income o hose in isk
36Insu e e enue and cos ca y a weigh o one, co esponding o he a e age social wel a e weigh in he
popula ion.
PRIVATIZING DISABILITY INSURANCE 1731
g oup 5. We conside a ange o alues o he coe icien o ela i e isk a e sion σbe-
ween 1 and 8, whe e a highe σimplies s onge equi y conce n, and hus la ge wel a e
weigh s on highe - isk g oups.
Resul s om he social ne alue calcula ions a e shown in Table IV. Column (1) sug-
ges s ha a p i a e DI manda e would lowe wel a e unde any deg ee o equi y con-
ce n. S onge equi y conce n dec eases he social ne alue, indica ing ha a p i a e DI
manda e would be a eg essi e policy. As can be seen in Appendix Figu e A9, o cing
all indi iduals o pu chase insu ance a ma ke p ices en ails la ge educ ions in con-
sume su plus among highe - isk g oups, since hey ha e o pay highe p ices ela i e
o a low willingness o pay. Column (3) shows he wel a e e ec s o a public DI man-
da e wi h income-based con ibu ions. In ou baseline ne alue calcula ions shown in
he i s ow o he able, he wel a e impac o a public DI manda e is he same as ha
o a p i a e DI manda e because he di e ence in p icing does no a ec o al su plus.
Howe e , in he p esence o equi y conce ns, a public DI manda e imp o es wel a e el-
a i e o he p i a e ma ke . In ui i ely, he social insu ance sys em wi h income-based
con ibu ions aises e enue om low- isk, high-income g oups and edis ibu es owa ds
high- isk, low-income g oups by p o iding hem wi h addi ional insu ance a p ices be-
low isk- a ed p emiums. This edis ibu ion is highly alued by he social planne . E en
unde low isk a e sion o σ=1, he social ne alue is al eady 1.215. As expec ed, he
social ne alue ises wi h he deg ee o equi y conce n; o ins ance, i becomes 1.749
unde σ=3 and 2.010 unde σ=8.
We also calcula e an analogous “social” MVPF measu e. To his end, we augmen equa-
ions (8)and(9) and include he a ious iscal ex e nali ies desc ibed in Sec ion 7.1.Ta-
ble IV shows ha he social MVPF is gene ally somewha lowe han he social ne alue.
As be o e, his occu s because he join e ec o iscal ex e nali ies inc eases ne cos .
Howe e , ou main wel a e esul s emain simila : he social MVPF o a p i a e DI man-
da e is a below one, whe eas he social MVPF o a public DI manda e is be ween 1.057
and 1.742.
Ex ensions and Robus ness. Appendix Table A13 shows esul s om h ee ex ensions
o ou wel a e analysis. Fi s , we sepa a e he e ec s o he di e en ypes o indi ec cos s
included in ou MVPF calcula ions. As expec ed, mo al haza d spillo e s on o public DI
and he iscal ex e nali y due o educed labo supply lowe he alue o a manda e, bu
he posi i e ex e nali y due o educed social assis ance claims inc eases i s alue. Second,
we check whe he ou esul s a e obus o elaxing he assump ion o a cons an demand
elas ici y ac oss isk g oups. We ind ha esul s a e quali a i ely una ec ed when using
he poin es ima es o each isk g oup ins ead. Thi d, we allow o some isk-based se-
lec ion in he p i a e DI ma ke . To quan i y he po en ial ange o cos cu e slopes, we
use he 95% con idence in e al a ound he claim e ec om panel (b) o Figu e 8.Con-
e ed in o li e ime claiming p obabili ies, he con idence in e al anges om ad e se
selec ion wi h a 6.9% di e ence in cos s be ween insu ed and uninsu ed indi iduals o
ad an ageous selec ion wi h a −11.1% di e ence. Ad e se selec ion somewha inc eases
he ne alue o a manda e, and ad an ageous selec ion somewha dec eases i , bu again,
esul s emain simila .
7.3. Supply-Side Fac o s: Ma ke Powe and Adminis a i e Cos s
Ou wel a e analysis elies on he su icien -s a is ics amewo k by Eina , Finkels ein,
and Cullen (2010), which akes insu ance supply and he cha ac e is ics o insu ance con-
1732 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
ac s as gi en. E idence om o he insu ance ma ke s, in pa icula long- e m ca e insu -
ance, sugges s ha he supply side can ma e o ake-up and wel a e (B aun, Kopecky,
and Ko eshko a (2019)). Key supply-side ac o s emphasized by p io li e a u e include
he deg ee o ma ke powe and he ex en o adminis a i e cos s.
Se e al empi ical obse a ions sugges ha bo h ma ke powe and adminis a i e cos s
a e less c i ical issues in ou se ing. Fi s , he Ge man p i a e DI ma ke is ela i ely
compe i i e: he op 3 p o ide s ha e a combined ma ke sha e o 34.2%. This is subs an-
ially lowe han in o he ma ke s, o example, B aun, Kopecky, and Ko eshko a (2019)
epo a op-3 ma ke sha e o 66% in U.S. long- e m ca e insu ance. Second, adminis a-
i e cos s a e modes in Ge man p i a e DI. Acco ding o Fische , Geye , and Zieba h
(2024), ixed and a iable adminis a i e cos s amoun o 3% and 10% o p emiums, e-
spec i ely, compa ed o 20% and 13% in B aun, Kopecky, and Ko eshko a (2019). Thi d,
co e age denials, which can esul om adminis a i e cos s, a e in equen in ou se ing.
The o e all denial a e is only 3% (see Sec ion 2.1), e sus 56% in U.S. long- e m ca e
insu ance (B aun, Kopecky, and Ko eshko a (2019)) and up o 64% in U.S. p i a e DI
(Hend en (2013)).
These empi ical ac s ansla e in o modes insu ance loads, a commonly used me -
ic combining insu e p o i s and adminis a i e cos s.37 Ac oss all isk g oups, we ind
an a e age load o 0.27, implying ha insu ed indi iduals can expec o ecei e 73% o
p emiums in bene i s. This is conside ably lowe han he a e age load o 0.42 epo ed
in B aun, Kopecky, and Ko eshko a (2019). No e ha hese loads a e al eady implici ly
inco po a ed in ou main wel a e analysis. Ma ke powe and adminis a i e cos s can
lead o ine icien ly low p i a e DI ake-up, which policy in e en ions aimed a inc eas-
ing ake-up such as a manda e can add ess. Howe e , ou wel a e esul s imply empi ical
insu ance loads a e no su icien ly la ge o jus i y a manda e.
To shed mo e ligh on he ole o adminis a i e cos s o p i a e DI ake-up and wel-
a e, we none heless pe o m an addi ional coun e ac ual simula ion building on B aun,
Kopecky, and Ko eshko a (2019)andFische , Geye , and Zieba h (2024). Some p opo-
nen s o egula o y ools such as minimum bene i a ios a gue ha hese could educe
adminis a i e cos s. In he coun e ac ual, we se adminis a i e cos s o ze o and as-
sume ha his cos educ ion is ully passed h ough o consume s. As i is unlikely ha
such an ex eme educ ion can be achie ed in p ac ice, we in e p e he esul s om his
simula ion as an uppe bound on he po en ial impac o adminis a i e cos s. Appendix
Table A14 shows ha he dec ease in p i a e DI p emiums due o he emo al o admin-
is a i e cos s leads o a modes ise in p i a e DI ake-up o 33%. The posi i e impac
on consume su plus and insu e p o i s exceeds he addi ional (di ec ) cos s o p o iding
DI, such ha he ne alue o he coun e ac ual is 1.359, indica ing an imp o emen in
o e all wel a e. While hese esul s a e quali a i ely unsu p ising, hey con i m ha he
impac o adminis a i e cos s is limi ed in ou se ing.
8. CONCLUSION
In his pape , we p o ide no el empi ical e idence on he unc ioning o p i a e DI
ma ke s. We documen signi ican c owding-ou a e a e o m ha cu public DI co e -
age. Ye , p i a e DI ake-up emains a om comple e, especially among low-income,
37We ollow B own and Finkels ein (2008) and calcula e insu ance loads as 1 −(𝔼(cik )/𝔼(pik )), whe e cik
a e expec ed DI bene i s (co esponding o he di ec cos o p o iding insu ance) om equa ion (5)andpik
a e expec ed insu ance p emium paymen s om equa ion (4).

PRIVATIZING DISABILITY INSURANCE 1733
low-educa ed, and high- isk indi iduals. Ou wel a e analysis highligh s he policy impli-
ca ions o hese indings. Ou baseline wel a e esul is ha lea ing p i a e DI co e age
o he olun a y ma ke is wel a e-imp o ing. This is closely ela ed o ou main empi i-
cal indings. Fi s , we do no ind ad e se selec ion, which would lead o ine icien ly low
insu ance ake-up in he p i a e ma ke and which would be he canonical a ionale o
a manda e. Second, he alue o ex a DI co e age e ealed by insu ance choices is low
o many indi iduals, especially in highe - isk g oups, and ou sizable demand elas ici y
es ima es imply ha insu ance alua ions decline as among he uninsu ed. Thi d, ma -
ke powe and adminis a i e cos s a e o insu icien magni ude o jus i y a manda e. In
o he wo ds, he p i a e DI ma ke co e s he majo i y o indi iduals whose willingness
o pay is abo e ma ginal cos .
Howe e , equi y conce ns can p o ide a a ionale o including he DI co e age cu -
en ly o e ed by he p i a e ma ke in he public DI manda e. Fo such a e o m o
imp o e social wel a e, i is c ucial o implemen income-based con ibu ions as in eal-
wo ld social insu ance sys ems. Such a manda e edis ibu es no only om high- o low-
income g oups bu also om low- o high- isk g oups. Impo an ly, his analysis akes he
design o o he ax and ans e schemes as gi en. I he go e nmen aims a edis ibu ing
ac oss income le els, he e a e likely mo e e icien ways o achie e equi y gains. Ne e -
heless, edis ibu ion ac oss disabili y isk ypes is a dis inc i e ea u e o public DI. I
would be wo hwhile o u u e esea ch o explo e hese ade-o s u he .
Th oughou he wel a e analysis, we ollow he li e a u e on manda es e sus ma ke s
in social insu ance se ings and main ain he assump ions o he e ealed p e e ence ap-
p oach (Eina , Finkels ein, and Cullen (2010), Finkels ein, Hend en, and Shepa d (2019),
Landais e al. (2021), Cab al, Cui, and Dwo sky (2022)). C ucially, his includes he as-
sump ion ha indi iduals make op imal p i a e DI pu chase decisions, such ha obse ed
demand e lec s indi iduals’ ue insu ance alua ions. As discussed in Sec ion 4.2,ob-
se ed p i a e DI ake-up could be explained by se e al ac o s consis en wi h e ealed
p e e ences. Howe e , i beha io al biases lead o ine icien unde insu ance o some
wo ke s, his could po en ially p o ide an al e na i e a ionale o a manda e beyond he
classic ma ke ailu es conside ed by he e ealed p e e ence app oach. S udying such bi-
ases is ou side he scope o his pape , bu his may be ano he a ea whe e u u e wo k is
needed o in o m policy.38
REFERENCES
AKERLOF,GEORGE A. (1970): “The Ma ke o “Lemons”: Quali y Unce ain y and he Ma ke Mechanism,”
Qua e ly Jou nal o Economics, 84 (3), 488–500. [1718]
ANONYMOUS FIRM (2019): “Da a on P i a e Disabili y Insu ance Con ac s,” Unpublished da a, accessed 30-
01-2019. [1705]
AUTOR,DAVID,MARK DUGGAN,KYLE GREENBERG,AND DAVID S. LYLE (2016): “The Impac o Disabili y
Bene i s on Labo Supply: E idence F om he VA’s Disabili y Compensa ion P og am,” Ame ican Economic
Jou nal: Applied Economics, 8 (3), 31–68. [1700]
AUTOR,DAVID,MARK DUGGAN,AND JONATHAN GRUBER (2014): “Mo al Haza d and Claims De e ence
in P i a e Disabili y Insu ance,” Ame ican Economic Jou nal: Applied Economics, 6 (4), 110–141. [1700,
1712]
AUTOR,DAVID,ANDREAS RAVNDAL KOSTOL,MAGNE MOGSTAD,AND BRADLEY SETZLER (2019): “Disabil-
i y Bene i s, Consump ion Insu ance, and Household Labo Supply,” Ame ican Economic Re iew, 109 (7),
2613–2654. [1700]
38In an ea lie e sion o his pape (Seibold, Sei z, and Siegloch (2022)), we conside ed beha io al biases
as an al e na i e a ionale o a manda e mo e explici ly. As i is di icul o di ec ly quan i y such biases in ou
da a, we de e his opic o u u e wo k.
1734 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
AUTOR,DAVID H., AND MARK G. DUGGAN (2003): “The Rise in Disabili y Rolls and he Decline in Unem-
ploymen ,” Qua e ly Jou nal o Economics, 118 (1), 157–205. [1700]
(2006): “The G ow h in he Social Secu i y Disabili y Rolls: A Fiscal C isis Un olding,” Jou nal o
Economic Pe spec i es, 20 (3), 71–96. [1700]
(2007): “Dis inguishing Income F om Subs i u ion E ec s in Disabili y Insu ance,” AEA Pape s &
P oceedings, 97 (2), 119–124. [1700]
AUTOR,DAVID H., MARK G. DUGGAN,AND DAVID S. LYLE (2011): “Ba le Sca s? The Puzzling Decline in
Employmen and Rise in Disabili y Receip Among Vie nam E a Ve e ans,” AEA Pape s & P oceedings, 101
(3), 339–344. [1700]
AUTOR,DAVID H., FRANK LEVY,AND RICHARD J. MURNANE (2003): “The Skill Con en o Recen Techno-
logical Chance: An Empi ical Exploi a ion,” Qua ely Jou nal o Economics, 118 (4), 1279–1334. [1712]
BORGHANS,LEX,ANNE C. GIELEN,AND ERZO F. P. LUTTMER (2014): “Social Suppo Subs i u ion and he
Ea nings Rebound: E idence F om a Reg ession Discon inui y in Disabili y Insu ance Re o m,” Ame ican
Economic Jou nal: Economic Policy, 6 (4), 34–70. [1700]
BOUND,JOHN (1989): “The Heal h and Ea nings o Rejec ed Disabili y Insu ance Applican s,” Ame ican Eco-
nomic Re iew, 79 (3), 482–503. [1700]
BOYER,M.MARTIN,PHILIPPE DEDONDER,CLAUDE FLUET,MARIE-LOUISE LEROUX,AND PIERRE-CARL
MICHAUD (2020): “Long-Te m Ca e Insu ance: In o ma ion F ic ions and Selec ion,” Ame ican Economic
Jou nal: Economic Policy, 12 (3), 134–169. [1701]
BRADLEY,JAKE,AND ALICE KUEGLER (2019): “Labo Ma ke Re o ms: An E alua ion o he Ha z Policies
in Ge many,” Eu opean Economic Re iew, 113 (C), 108–135. [1711]
BRAUN,R.ANTON,KAREN A. KOPECKY,AND TATYANA KORESHKOVA (2019): “Old, F ail, and Uninsu ed:
Accoun ing o Fea u es o he U.S. Long-Te m Ca e Insu ance Ma ke ,” Econome ica, 87 (3), 981–1019.
[1700,1701,1703,1716,1732]
BROWN,JEFFREY R., AND AMY FINKELSTEIN (2008): “The In e ac ion o Public and P i a e Insu ance: Med-
icaid and he Long-Te m Ca e Insu ance Ma ke ,” Ame ican Economic Re iew, 98 (3), 1083–1102. [1707,
1732]
BUNDORF,M.KATE,JONATHAN LEVIN,AND NEALE MAHONEY (2012): “P icing and Wel a e in Heal h Plan
Choice,” Ame ican Economic Re iew, 102 (7), 3214–3248. [1720]
BURKHAUSER,RICHARD V., MARY CDALY,AND NICOLAS R. ZIEBARTH (2016): “P o ec ing Wo king-Age
People Wi h Disabili ies: Expe iences o Fou Indus ialized Na ions,” Jou nal o Labou Ma ke Resea ch,
49, 367–386. [1700]
CABRAL,MARIKA,AND MARK R. CULLEN (2019): “Es ima ing he Value o Public Insu ance Using Com-
plemen a y P i a e Insu ance,” Ame ican Economic Jou nal: Economic Policy, 11 (3), 88–129. [1701,1712,
1726]
CABRAL,MARIKA,CAN CUI,AND MICHAEL DWORSKY (2022): “The Demand o Insu ance and Ra ionale
o a Manda e: E idence F om Wo ke s’ Compensa ion Insu ance,” Ame ican Economic Re iew, 112 (5),
1621–1668. [1701,1733]
CHEN,SUSAN,AND WILBERT VAN DER KLAAUW (2008): “The Wo k Disincen i e E ec s o he Disabili y
Insu ance P og am in he 1990s,” Jou nal o Econome ics, 142, 757–784. [1700]
CHIAPPORI,PIERRE-ANDRÉ,AND BERNARD SALANIÉ (2000): “Tes ing o Asymme ic In o ma ion in Insu -
ance Ma ke s,” Jou nal o Poli ical Economy, 108 (1), 56–78. [1699,1718]
COHEN,ALMA,AND LIRAN EINAV (2007): “Es ima ing Risk P e e ences F om Deduc ible Choice,” Ame ican
Economic Re iew, 97 (3), 745–788. [1717]
CUTLER,DAVID,AMY FINKELSTEIN,AND KATHLEEN MCGARRY (2008): “Mul iple Dimensions o P i a e
In o ma ion: E idence F om he Long-Te m Ca e Insu ance Ma ke ,” AEA Pape s & P oceedings,98(2),
157–162. [1719]
CUTLER,DAVID M., AND JONATHAN GRUBER (1996): “Does Public Insu ance C owd ou P i a e Insu ance?”
Qua e ly Jou nal o Economics, 111 (2), 391–430. [1707]
DE CHAISEMARTIN,CLEMENT,AND XAVIER D’HAULTFOEUILLE (2020): “Two-Way Fixed E ec s Es ima o s
Wi h He e ogeneous T ea men E ec s,” Ame ican Economic Re iew, 110 (9), 2964–2996. [1721]
(2024): “Di e ence-in-Di e ences Es ima o s o In e empo al T ea men E ec s,” Re iew o Eco-
nomics and S a is ics ( o hcoming). [1721-1723]
DENGLER,KATHARINA,BRITTA MATTHES,AND WIEBKE PAULUS (2014): “Occupa ional Tasks in he Ge man
Labo Ma ke : An Al e na i e Measu emen on he Basis o an Expe Da abase,” FDZ Me hoden epo
12/2014. [1712,1713,1716]
DESHPANDE,MANASI (2016a): “Does Wel a e Inhibi Success? The Long- Te m E ec s o Remo ing Low-
Income You h F om he Disabili y Rolls,” Ame ican Economic Re iew, 106 (11), 3300–3330. [1700]
PRIVATIZING DISABILITY INSURANCE 1735
(2016b): “The E ec o Disabili y Paymen s on Household Ea nings and Income: E idence F om he
SSI Child en’s P og am,” Re iew o Economics and S a is ics, 98 (4), 638–654. [1700]
DESHPANDE,MANASI,AND LEE LOCKWOOD (2022): “Beyond Heal h: Non-Heal h Risk and he Value o
Disabili y Insu ance,” Econome ica, 90 (4), 1781–1810. [1701]
DEUTSCHE AKTUARVEREINIGUNG (DAV) (2018): “Übe p ü ung de Angemessenhei de DAV 1997 I als
Rese ie ungs a el ü Be u sun ähigkei s e siche ung,” DAV-Mi eilung. [1725]
DIAMOND,PETER,AND EYTAN SHESHINSKI (1995): “Economic Aspec s o Op imal Disabili y Bene i s,” Jou -
nal o Public Economics, 57 (1), 1–23. [1701]
EINAV,LIRAN,AND AMY FINKELSTEIN (2023): “Empi ical Analyses o Selec ion and Wel a e in Insu ance
Ma ke s: A Sel -Indulgen Su ey,” Gene a Risk and Insu ance Re iew, 48, 167–191. [1720]
EINAV,LIRAN,AMY FINKELSTEIN,AND MARK R. CULLEN (2010): “Es ima ing Wel a e in Insu ance Ma ke s
Using Va ia ions in P ices,” The Qua e ly Jou nal o Economics, 125 (3), 877–921. [1699,1701,1712,1720,
1723,1724,1727,1731,1733]
FAZ (2012): “Be u sun ähigkei : Schu z om Che ,” Accessed 13/09/2024, h ps://www. az.ne /ak uell/ka ie e-
hochschule/ ech -und-gehal /be u sun aehigkei -schu z- om-che -11739000.h ml.[1703]
FINKELSTEIN,AMY,AND NATHANIEL HENDREN (2020): “Wel a e Analysis Mee s Causal In e ence,” Jou nal
o Economic Pe spec i es, 34 (4), 146–167. [1727]
FINKELSTEIN,AMY,AND KATHLEEN MCGARRY (2006): “Mul iple Dimensions o P i a e In o ma ion: E i-
dence F om he Long-Te m Ca e Insu ance Ma ke ,” Ame ican Economic Re iew, 96 (4), 938–958. [1701,
1719]
FINKELSTEIN,AMY,AND ROBIN MCKNIGHT (2008): “Wha Did Medica e Do? The Ini ial Impac o Medica e
on Mo ali y and ou o Pocke Medical Spending,” Jou nal o Public Economics, 92 (7), 1644–1668. [1715]
FINKELSTEIN,AMY,AND JAMES POTERBA (2004): “Ad e se Selec ion in Insu ance Ma ke s: Policyholde E -
idence F om he U.K. Annui y Ma ke ,” Jou nal o Poli ical Economy, 112 (1), 183–208. [1712]
(2014): “Tes ing o Asymme ic In o ma ion Using “Unused Obse ables” in Insu ance Ma ke s:
E idence F om he U.K. Annui y Ma ke ,” Jou nal o Risk and Insu ance, 81 (4), 709–734. [1712]
FINKELSTEIN,AMY,NATHANIEL HENDREN,AND MARK SHEPARD (2019): “Subsidizing Heal h Insu ance o
Low-Income Adul s: E idence F om Massachuse s,” Ame ican Economic Re iew, 109 (4), 1530–1567. [1701,
1733]
FISCHER,BJOERN,JOHANNES M. GEYER,AND NICOLAS R. ZIEBARTH (2024): “Fundamen ally Re o ming
he DI Sys em: E idence F om Ge man No ch Coho s,” NBER Wo king Pape No. 30812. [1700,1716,
1732]
FORSCHUNGSDATENZENTRUM DER RENTENVERSICHERUNG (FDZ-RV) (2015): “On-Si e Ve sion Ve -
siche en en enzugang 1992–2014,” Resea ch Da a Cen e o he Ge man S a e Pension Fund, accessed
14-08-2020. [1706]
FRENCH,ERIC (2005): “The E ec s o Heal h, Weal h, and Wages on Labou Supply and Re i emen Be-
ha iou ,” Re iew o Economic S udies, 72 (2), 395–427. [1717]
FRENCH,ERIC,AND JAE SONG (2014): “The E ec o Disabili y Insu ance Receip on Labo Supply,” Ame i-
can Economic Jou nal: Economic Policy, 6 (2), 291–337. [1700]
GAO (2018): “Social Secu i y Disabili y Insu ance – In o ma ion on Po en ial Implica ions o Expanding P i-
a e Disabili y Insu ance,” Repo o Cong essional Reques e s, GAO-18-248, U.S. Go e nmen Accoun -
abili y O ice. [1697,1698,1700]
GDV (2023): “7 Fak en zu Be u sun ähigkei s e siche ung,” Repo , Gesam e band Deu sche Ve siche e .
Accessed 15/07/2024, h ps://www.gd .de/gd / hemen/leben/7- ak en-zu -be u sun aehigkei s e siche ung-
34338.[1703]
GELBER,ALEXANDER,TIMOTHY J. MOORE,AND ALEXANDER STRAND (2017): “The E ec o Disabili y In-
su ance Paymen s on Bene icia ies’ Ea nings,” Ame ican Economic Jou nal: Economic Policy, 9 (3), 229–261.
[1700]
GRUBER,JONATHAN (2000): “Disabili y Insu ance Bene i s and Labo Supply,” Jou nal o Poli ical Economy,
108 (6), 1162–1183. [1700]
HALLER,ANDREAS,STEFAN STAUBLI,AND JOSEF ZWEIMÜLLER (2024): “Designing Disabili y Insu ance Re-
o ms: Tigh ening Eligibili y Rules o Reducing Bene i s?” Econome ica, 92 (1), 79–110. [1701]
HENDREN,NATHANIEL (2013): “P i a e In o ma ion and Insu ance Rejec ions,” Econome ica,81(5),
1713–1762. [1701,1703,1732]
HENDREN,NATHANIEL,AND BEN SPRUNG-KEYSER (2020): “A Uni ied Analysis o Go e nmen Policies,”
Qua e ly Jou nal o Economics, 135 (3), 1209–1318. [1699,1727,1729]
HENDREN,NATHANIEL,CAMILLE LANDAIS,AND JOHANNES SPINNEWIJN (2021): “Choice in Insu ance Ma -
ke s: A Pigou ian App oach o Social Insu ance Design,” Annual Re iew o Economics, 13, 457–486. [1701,
1720]
1736 A. SEIBOLD, S. SEITZ, AND S. SIEGLOCH
HUCKFELDT,CHRISTOPHER (2022): “Unde s anding he Sca ing E ec o Recessions,” Ame ican Economic
Re iew, 112 (4), 1273–1310. [1716]
HUMLUM,ANDERS,JAKOB R. MUNCH,AND PERNILLE PLATO (2025): “Changing T acks: Human Capi al In-
es men A e Loss o Abili y,” Ame ican Economic Re iew 115 (5), 1520–1554. [1715]
JACOBS,LINDSAY (2023): “Occupa ions, Re i emen , and he Value o Disabili y Insu ance,” Jou nal o Public
Economics, 225. [1717]
KONING,PIERRE,AND MAARTEN LINDEBOOM (2015): “The Rise and Fall o Disabili y Insu ance En ollmen
in he Ne he lands,” Jou nal o Economic Pe spec i es, 29 (2), 151–172. [1700]
KOSTOL,ANDREAS RAVNDAL,AND MAGNE MOGSTAD (2014): “How Financial Incen i es Induce Disabili y
Insu ance Recipien s o Re u n o Wo k,” Ame ican Economic Re iew, 104 (2), 624–655. [1700]
LANDAIS,CAMILLE,AND JOHANNES SPINNEWIJN (2021): “The Value o Unemploymen Insu ance,” Re iew o
Economic S udies, 88 (6), 3041–3085. [1717]
LANDAIS,CAMILLE,ARASH NEKOEI,PETER NILSSON,DAVID SEIM,AND JOHANNES SPINNEWIJN (2021):
“Risk-Based Selec ion in Unemploymen Insu ance: E idence and Implica ions,” Ame ican Economic Re-
iew, 111 (4), 1315–1355. [1701,1719,1724,1733]
LIEBMAN,JEFFREY B. (2015): “Unde s anding he Inc ease in Disabili y Insu ance Bene i Receip in he
Uni ed S a es,” Jou nal o Economic Pe spec i es, 29 (2), 123–150. [1698,1700]
LOCKWOOD,LEE (2018): “Inciden al Beques and he Choice o Sel -Insu e La e-Li e Risks,” Ame ican Eco-
nomic Re iew, 108 (9), 2513–2550. [1717]
LOW,HAMISH,AND LUIGI PISTAFERRI (2015): “Disabili y Insu ance and he Dynamics o he Incen i e Insu -
ance T ade-o ,” Ame ican Economic Re iew, 105 (10), 2986–3029. [1701]
(2020): “Disabili y Insu ance: Theo e ical T ade-O s and Empi ical E idence,” Fiscal S udies,41(1),
129–164. [1700]
MAESTAS,NICOLE,KATHLEEN J. MULLEN,AND ALEXANDER STRAND (2013): “Does Disabili y Insu ance Re-
ceip Discou age Wo k? Using Examine Assignmen o Es ima e Causal E ec s o SSDI Receip ,” Ame i-
can Economic Re iew, 103 (5), 1797–1829. [1700]
MARIE,OLIVIER,AND JUDIT VALL CASTELLO (2012): “Measu ing he (Income) E ec o Disabili y Insu ance
Gene osi y on Labo Ma ke Pa icipa ion,” Jou nal o Public Economics, 96 (1–2), 198–210. [1700]
MEYER,BRUCE D., AND WALLACE K. C. MOK (2019): “Disabili y, Ea nings, Income and Consump ion,” Jou -
nal o Public Economics, 171, 51–69. [1701,1715]
MITCHELL,OLIVIA S., JAMES M. POTERBA,MARK J. WARSHAWSKY,AND JEFFREY R. BROWN (1999): “New
E idence on he Money’s Wo h o Indi idual Annui ies,” Ame ican Economic Re iew, 89 (5), 1299–1318.
[1715]
MORGEN &MORGEN (2021): “Da a on P i a e Disabili y Insu ance Ma ke ,” Unpublished da a, accessed
13-07-2021. [1706]
MULLEN,KATHLEEN J., AND STEFAN STAUBLI (2016): “Disabili y Bene i Gene osi y and Labo Fo ce Wi h-
d awal,” Jou nal o Public Economics, 143, 49–63. [1700]
OECD (2024): “OECD Social Expendi u e Da abase,” Da abase. [1697]
ROTHSCHILD,MICHAEL,AND JOSEPH STIGLITZ (1976): “Equilib ium in Compe i i e Insu ance Ma ke s: An
Essay on he Economics o Impe ec In o ma ion,” Qua e ly Jou nal o Economics, 90 (4), 629–649. [1718]
RUH,PHILIPPE,AND STEFAN STAUBLI (2019): “Financial Incen i es and Ea nings o Disabili y Insu ance Re-
cipien s: E idence F om a No ch Design,” Ame ican Economic Jou nal: Economic Policy, 11 (2), 269–300.
[1700]
SCHMIDHEINY,KURT,AND SEBASTIAN SIEGLOCH (2023): “On E en S udy Designs and Dis ibu ed-Lag
Models: Equi alence, Gene aliza ion and P ac ical Implica ions,” Jou nal o Applied Econome ics,38(5),
695–713. [1721]
SEIBOLD,ARTHUR (2021): “Re e ence Poin s o Re i emen Beha io : E idence F om Ge man Pension Dis-
con inui ies,” Ame ican Economic Re iew, 111 (4), 1126–1165. [1706]
SEIBOLD,ARTHUR,SEBASTIAN SEITZ,AND SEBASTIAN SIEGLOCH (2022): “P i a izing Disabili y Insu ance,”
ZEW Discussion Pape No. 22-010 / Ap il 2022. [1733]
(2025): “Supplemen o ‘P i a izing Disabili y Insu ance’,” Econome ica Supplemen al Ma e ial, 93,
h ps://doi.o g/10.3982/ECTA22113.[1701]
SEITZ,SEBASTIAN (2021): “The Mo al Haza d Cos o P i a e Disabili y Insu ance and I s Wel a e Conse-
quences,” Wo king Pape . h p://dx.doi.o g/10.2139/ss n.4323015.[1700,1706,1728]
STATISTISCHES BUNDESAMT (DESTATIS) (2019a): “Einkommens- und Ve b auchss ichp obe (EVS) 2013,” Re-
sea ch Da a Cen e o he Ge man Fede al S a is ical O ice, accessed 03-08-2022, DOI: h ps://doi.o g/10.
21242/63221.2013.00.04.3.1.1.[1706,1707]
STATISTISCHES BUNDESAMT (DESTATIS) (2019b): “On-Si e Ve sion Ve diens s uk u e hebung 2014,” Re-
sea ch Da a Cen e o he Ge man Fede al S a is ical O ice, accessed 14-02-2020, DOI: h ps://doi.o g/
10.21242/62111.2014.00.00.1.1.1.[1705,1706]