Tax revolts and sovereign defaults

A ce, Fe nando; Mo gan, Jan; We quin, Nicolas
Wo king Pape
Tax e ol s and so e eign de aul s
IDB Wo king Pape Se ies, No. IDB-WP-1719
P o ided in Coope a ion wi h:
In e -Ame ican De elopmen Bank (IDB), Washing on, DC
Sugges ed Ci a ion: A ce, Fe nando; Mo gan, Jan; We quin, Nicolas (2025) : Tax e ol s and so e eign
de aul s, IDB Wo king Pape Se ies, No. IDB-WP-1719, In e -Ame ican De elopmen Bank (IDB),
Washing on, DC,
h ps://doi.o g/10.18235/0013572
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/324816
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/3.0/igo/
Tax Re ol s and So e eign De aul s
Fe nando A ce
Jan Mo gan
Nicolas We quin
WORKING PAPER No IDB-WP-1719
In e -
A
me ican De elopmen Bank
Depa men o Resea ch and Chie Economis
June 2025
* In e -
A
me ican De elopmen Bank
** Massachuse s Ins i u e o Technology
*** Fede al Rese e Bank o Chicago
Tax Re ol s and So e eign De aul s
Fe nando A ce*
Jan Mo gan**
Nicolas We quin***
In e -
A
me ican De elopmen Bank
Depa men o Resea ch and Chie Economis
June 2025
Ca aloging-in-Publica ion da a p o ided by he
In e -Ame ican De elopmen Bank
Felipe He e a Lib a y
A ce, Fe nando.
Tax e ol s and so e eign de aul s / Fe nando A ce, Jan Mo gan, Nicolas
We quin.
p. cm. — (IDB Wo king Pape Se ies ; 1719)
Includes bibliog aphical e e ences.
1. Financial c isis-Ma hema ical models-A gen ina. 2. Tax and expendi u e
limi a ions-Ma hema ical models-A gen ina. 3. Fiscal policy-Ma hema ical
models-A gen ina. I. Mo gan, Jan. II. We quin, Nicolas. III. In e -Ame ican
De elopmen Bank. Depa men o Resea ch and Chie Economis . IV. Ti le.
V. Se ies.
IDB-WP-1719
h p://www.iadb.o g
Copy igh © 2025 In e -Ame ican De elopmen Bank ("IDB"). This wo k is subjec o a C ea i e
Commons license CC BY 3.0 IGO (h ps://c ea i ecommons.o g/licenses/by/3.0/igo/legalcode). The
e ms and condi ions indica ed in he URL link mus be me and he espec i e ecogni ion mus be
g an ed o he IDB.
Fu he o sec ion 8 o he abo e license, any media ion ela ing o dispu es a ising unde such license
shall be conduc ed in acco dance wi h he WIPO Media ion Rules. Any dispu e ela ed o he use o
he wo ks o he IDB ha canno be se led amicably shall be submi ed o a bi a ion pu suan o he
Uni ed Na ions Commission on In e na ional T ade Law (UNCITRAL) ules. The use o he IDB's name
o any pu pose o he han o a ibu ion, and he use o IDB's logo shall be subjec o a sepa a e
w i en license ag eemen be ween he IDB and he use and is no au ho ized as pa o his license.
No e ha he URL link includes e ms and condi ions ha a e an in eg al pa o his license.
The opinions exp essed in his wo k a e hose o he au ho s and do no necessa ily e lec he iews o
he In e -Ame ican De elopmen Bank, i s Boa d o Di ec o s, o he coun ies hey ep esen .
Abs ac *
Poli ical c ises o en coincide wi h iscal c ises, wi h complex causal dynamics a play.
We examine he in e ac ion be ween ax e ol s and so e eign isk using a quan i a i e
s uc u al model calib a ed o A gen ina. In he model, he go e nmen can be con olled
by poli ical pa ies wi h di e en p e e ences o edis ibu ion. Households may op
o e ol in esponse o he iscal decisions o he ule . While e ol s en ail economic
cos s, hey also inc ease he likelihood o poli ical u no e . Ou model mi o s he da a
by gene a ing poli ical c ises concu en wi h iscal u moil. Speci ically, we ind ha
ou model aligns closely wi h he condi ions obse ed du ing he Mac i adminis a ion
(2015-2019). We ind ha le -leaning pa ies a e mo e p one o de aul upon en e ing
o ice, while igh -leaning pa ies issue mo e deb . Ou amewo k explains he high
de ici s obse ed du ing he Mac i adminis a ion as well as he so e eign de aul ha
occu ed immedia ely a e he le egained powe .
JEL classi ica ions: E32, E44, F41, G01, G28
Keywo ds: Ci il un es , Financial
c ises, So e eign de aul , Redis ibu ion
∗
We hank Manuel Amado , C is ina A ellano, Gadi Ba le y, Ja ie Bianchi, Emilio Espino, S elios
Fou akis, Mike Goloso , G ey Go don, Tim Kehoe, Ilenin Kondo, Andy Neumeye , Juan Pablo Nicolini,
Leona do Ma inez, Ma ias Mo e i, F ancisco Roldan, Cesa Sosa-Padilla, Monica T an-Xuan, and Jing
Zhang, as well as, semina pa icipan s a he NBER Eme ging Ma ke s Sp ing Mee ing 2024, he SED 2024,
SCIEA 2024, and he IMF Fiscal A ai s Wo kshop, o hei use ul commen s. We a e also hank ul o
Sophia Lansell and Caydn Hillie o hei excellen esea ch assis ance. The iews exp essed he ein a e hose
o he au ho s and no necessa ily hose o he Fede al Rese e Bank o Chicago, he Fede al Rese e Sys em,
o he In e -Ame ican De elopmen Bank, i s Execu i e Boa d, o i s Managemen . All e o s a e ou own.

1 In oduc ion
Fiscal decisions a e made by public o icials aligned wi h speci ic ac ions o he na ion. In
p ac ice, ax e o ms and iscal adjus men decisions o en p o oke ci il un es . While he
li e a u e on so e eign de aul has explo ed a ious ace s o poli ical isk since he seminal
wo k o Ha chondo e al. (2009) and Ha chondo and Ma inez (2010), he heo e ical insigh s
om Acemoglu and Robinson (2001) on how he isk o ci il con lic cons ains go e nmen al
decisions ha e ye o be in eg a ed in o a quan i a i e model o so e eign isk. This pape
aims o ill ha oid.
P o es s di e om o he o ms o poli ical backlash along c ucial dimensions ele an
o so e eign isk. Unlike elec ions, p o es s allow opposi ion membe s o exp ess hei
dissa is ac ion immedia ely, o en leading o subsequen declines in eelec ion a es o he
incumben (Mades am e al. (2013)). Mo eo e , p o es s dis up economic ac i i y o
bo h demons a o s and he wide economy, in oducing signi ican s a egic conside a ions
o all in ol ed pa ies. P o es e s mus be willing o bea he cos s, while go e nmen s
may adjus iscal policy o mi iga e un es . Fu he mo e, since p o es s can al e he
eelec ion p ospec s o incumben s, hey can also shape lende s’ expec a ions ega ding u u e
go e nmen p e e ences, di ec ly impac ing he p icing o public deb .
We highligh he signi icance o hese dynamics by examining A gen ina, pa icula ly
du ing he p esidency o Mau icio Mac i om Decembe 2015 o Decembe 2019, while also
showing ha he ela ionship be ween poli ical and so e eign isk holds o a c oss-sec ion o
coun ies. We show ha episodes o heigh ened ci il un es , which we e e o as e ol s in
his pape , a e associa ed wi h inc eases in so e eign sp eads, i.e., iscal c ises.
We de elop a quan i a i e so e eign deb model in he adi ion o Ea on and Ge so i z
(1981) and A ellano (2008), supplemen ing i wi h a non-linea ax amewo k inspi ed by
Hea hco e e al. (2017) and inco po a ing ci il con lic dynamics om Acemoglu and Robinson
(2001). He e ogeneous households pa icipa e in p oduc ion and a e uled by poli ical pa ies
wi h di e en p e e ences o edis ibu ion ha al e na e in o ice. Fu he mo e, households
can s a egically espond o he go e nmen ’s iscal choices by s aging e ol s. Re ol s lowe
p oduc i i y bu dec ease he p obabili y ha he incumben pa y will s ay in powe . As in
he da a, we no e ha poli ical and iscal c ises o en coincide.
Ou model unco e s wo channels linking e ol s and so e eign isk. Fi s , poli ical con lic
can ele a e de aul isk by inc easing he p obabili y o ansi ioning om a uling pa y wi h
a low de aul a e o one wi h a highe a e. This mechanism gene a es a es able p edic ion
ha coincides wi h he A gen inean expe ience in 2020. De aul s a e mo e likely o ollow
1
a ansi ion om igh -wing o le -wing go e nance, and p o es s make hese ansi ions
mo e likely. Second, e ol s can mi iga e de aul isk by sho ening he expec ed enu e o
incumben s who choose o de aul . Indeed, we ind ha in equilib ium, e ol s a e mo e
common du ing de aul s han unde epaymen , allowing hem o unc ion as an endogenous
de aul cos . The lende s an icipa e his addi ional cos , and o e a mo e a o able bond
p ice schedule o an incumben in ou baseline model ela i e o an en i onmen wi h poli ical
u no e bu wi hou e ol s.
Ou quan i a i e analysis shows ha , while he wo channels o se each o he in e ms o
de aul equency, e ol s lead o inc eased o e all indeb edness. In equilib ium, go e nmen s
u ilize he iscal space c ea ed by e ol s o ake on mo e deb . Al hough bo h pa ies bo ow
mo e, igh -wing incumben s—who ace a highe equency o p o es s—a e no ably mo e
p one o inc ease de ici s. As a esul , i becomes mo e likely ha a le -wing go e nmen
will inhe i a deb load misaligned wi h i s dis ibu ional p io i ies, p omp ing i o de aul .
Rela ed li e a u e This pape connec s o se e al s ands o he li e a u e on so e eign
de aul , poli ical economy, and public inance. Ou p ima y con ibu ion lies wi hin he
poli ical economy o so e eign de aul , ocusing on he isks associa ed wi h poli ical u no e .
1
Ou pape builds on he li e a u e, o igina ing wi h Cuad a and Sap iza (2008) and
Ha chondo e al. (2009), ha s udies he in e ac ion be ween poli ical economy and so e eign
de aul — o which we add a edis ibu i e mo i e o iscal policy. The exogenous componen
o poli ical u no e in ou model esembles Ha chondo e al. (2009) and Co oc, Joh i and
Sosa-Padilla (2021). In hese pape s, howe e , he asymme y in de aul beha io ac oss
pa ies s ems om exogenous di e ences in hei discoun ac o s o p obabili y o eelec ion.
2
By con as , in ou model, bo h pa ies sha e iden ical discoun ac o s, de aul cos s, and
exogenous p obabili y o s aying in o egaining powe —pa ies di e only in hei p e e ences
ega ding he edis ibu i e e ec s o iscal policy. Thus, absen endogenous e ol choices,
bo h pa ies in ou model would de aul a he same a es. In o he wo ds, ou pape
endogenizes di e ences in de aul a es ac oss pa ies wi hou ha d-wi ing his esul ia
di e ences in exogenous pa y cha ac e is ics.
The ecen wo ks o And easen e al. (2019), Azzimon i and Mi a (2023b), and Scholl
1
Ou model will, howe e , also d aw elemen s om he s and o he li e a u e ha ocuses on so e eign
epu a ion (Amado and Phelan (2021), Fou akis (2023), Mo elli and Mo e i (2023).
2
Asymme ic de aul s a e also p esen in he model o endogenous poli ical u no e o Cha e jee and
Eyigungo (2019), whe e i a ises om he abili y o misapp op ia e public unds o p i a e gain and om
in o ma ional ic ions conce ning he e ec o go e nmen policies. These issues a e ou side he scope o ou
pape .
2
(2024) a e he mos closely aligned wi h ou pape .
3
Simila o hese pape s, we p esen
a model ha in eg a es poli ical cons ain s, edis ibu i e policies, and he dynamics o
so e eign de aul . Ou model complemen s his li e a u e in wo main ways. Fi s , p o es s
endogenize he likelihood o poli ical u no e and simul aneously cons ain go e nmen
policies. They se e as an economically cos ly means o households o penalize incumben s.
Thus, he me e h ea o p o es s in luences he s a egic decisions o he go e nmen , e en
hough ou amewo k is lexible enough o gene a e e ol s along he equilib ium pa h a
a es consis en wi h empi ical da a. By con as , And easen e al. (2019) and Azzimon i
and Mi a (2023b) impose by cons uc ion ha go e nmen iscal policies sa is y poli ical
easibili y cons ain s (e.g., pa liamen a y app o al, o co up ion in he o m o inancial
incen i es channeled o ce ain g oup leade s). The e o e, coe ced poli ical ansi ions ne e
ac ually happen in equilib ium in hese pape s.
4
Ou poli ical economy se ing is close in
his sense o ha o Scholl (2024), in which he p obabili y o elec ions is exogenous bu
he elec ion ou come depends on he edis ibu i e choices made by he incumben pa y.
Ins ead, we endogenize u no e ia economically cos ly and s a egic e ol s, a he han
peace ul ans e s o powe h ough scheduled elec ions.
Second, And easen e al. (2019), Azzimon i and Mi a (2023b), and Scholl (2024) all assume
ha go e nmen deb has a sho (one pe iod) du a ion. By con as , we allow go e nmen s
o issue long- e m deb . Beyond me e ealism, inco po a ing long- e m ma u i y o analyze
de aul s in an en i onmen wi h poli ical ic ions yields h ee impo an implica ions. Fi s ,
i allows us o quan i a i ely app oxima e he le els and ola ili y o deb and sp eads
obse ed in empi ical da a.
5
Second, long ma u i ies imply ha he bo owing policies o he
opposi ion pa y a ec he p icing o he incumben ’s deb . In a one-pe iod deb model, only
he de aul policies (and no he bo owing choices) o he opposi ion in luence he p ice o
he incumben ’s deb . Poli ical ic ions hus in e ac wi h bo h de aul and deb -dilu ion
incen i es. Thi d, long deb ma u i ies educe exposu e o ollo e isk. A model whe e he
en i e s ock o deb is due each pe iod can o e es ima e he impac o poli ical u no e s on
de aul isk. In Sec ion 7.1, we demons a e ha igh - o-le ansi ions a e he p ima y
channel h ough which poli ical isk can escala e de aul isk, emphasizing he impo ance o
accu a ely es ima ing his channel.
Finally, ou pape builds on an ex ensi e heo e ical li e a u e ega ding he economic
3
See also hei ela ed wo k, Azzimon i and Mi a (2023a), Scholl (2017), and Scholl and He mann (2024).
4
In addi ion, edis ibu ion in ou model is subjec o an e iciency-equi y ade-o ha a ec s agg ega e
ou pu and, consequen ly, he go e nmen ’s epaymen capaci y. This a ia ion in epaymen capaci y leads
o he “poli ical de aul s” p oposed by Ha chondo e al. (2009), which a e de aul s ha occu only because o
poli ical u no e and a e, in ou se ing, acili a ed by p o es s.
5This esul is es ablished in Cha e jee and Eyigungo (2012) and Ha chondo and Ma inez (2009).
3
e ec s o poli ical isk and con lic , pa icula ly in ela ion o axa ion and edis ibu ion;
e.g., Alesina and Tabellini (1989), Acemoglu and Robinson (2001), Ba aglini and Coa e
(2008), Ya ed (2010), Acemoglu e al. (2011), Scheue and Woli zky (2016), Do is e al.
(2016), Ba be a and Jackson (2020), among o he s. In pa icula , ecen con ibu ions o
his ield ha e ocused on models o egime change o global coo dina ion games o s udy
p o es s. While hese issues a e ce ainly impo an , hey all ou side he scope o ou p esen
analysis. Fu he mo e, he analy ical insigh s p esen ed in hese s udies a e de i ed wi hin
delibe a ely simpli ied models. By con as , ou pape is quan i a i e in na u e and aims o
simul aneously ma ch he da a on in e es a e sp eads, de aul s, poli ical ansi ions, and
p o es s.
2 Empi ical Mo i a ion
2.1 C oss-coun y E idence
Following he li e a u e, we documen he posi i e ela ion be ween poli ical isk and sp eads
ac oss di e en coun ies. Ha chondo and Ma inez (2009) we e i s o highligh he impo -
ance o poli ical isk measu ed by he In e na ional Coun y Risk Guide (ICRG) indica o ,
which hey in e p e as cap u ing he e ec o go e nmen al u no e on so e eign sp eads.
They s udy he 2001 A gen ine de aul episode, poin ing o ou comes o high go e nmen
u no e d i en by popula dissa is ac ion. Simila ly, T ebesch (2019) uses ICRG and leng h
o enego ia ion o a gue ha in ense poli ical u moil makes es uc u ing mo e di icul .
We con i m he posi i e co ela ion be ween poli ical isk and sp eads in he c oss-coun y
panel eg ession o coun ies wi h a c edi a ing below A-, p esen ed in Table 1.6
The posi i e associa ion be ween poli ical isk and so e eign sp eads pe sis s e en a e
accoun ing o mac oeconomic undamen als (such as Cu en Accoun Balance, Rese es,
Real G oss Domes ic P oduc (GDP) g ow h, and P ima y Balance), as well as ime and
coun y ixed e ec s, as shown in Table 1. In Appendix G, we show ha his ela ionship
is also p esen ega dless o he pa y in powe , albei wi h a s onge e ec when he
incumben is a igh -wing pa y. Figu e 1shows he i ed alues o he eg ession wi h all
he a o emen ioned mac oeconomic con ols and he ixed e ec s.
Addi ionally, in Figu e 2, we u he in es iga e his ela ionship in he da a by ocusing
6
The sou ce o c edi de aul swaps (CDS) sp eads is CMAN ob ained om Bloombe g Pe Secu i y Da a
and he Bloombe g e minal, used wi h he pe mission o Bloombe g. This eg ession only includes coun ies
whose o eign-exchange deb was a ed below A- by S anda d and Poo ’s in July 2024. We conduc he same
exe cise while keeping in coun ies o all a ings in Appendix Gand ind simila esul s.
4
a es a e mono onically inc easing (dec easing) wi h income.
9
No ice ha i he ax schedule
is p og essi e, low-income households ecei e ans e s, i.e.,
τ
(
y
)
<
0, whene e
y < τ1/τ1
0
.
Thus, o a gi en a e o p og essi i y
τ1>
0, a highe pa ame e
τ0
maps in o bo h a highe
base and a highe le el o ans e s o low-income households.
Households a e hand- o-mou h. When making hei decisions, hey ake wages and all
o he agg ega e s a es o he economy—including he iscal package p oposed by he pa y in
powe —as gi en. A household o ype iwi h wage wisol es
Ui= max
C,N ui(C, N) (1)
s. . C=wiN−τ(wiN).
We ob ain ha he op imal choice o labo supply is gi en by
Ni= [ψi(1 −τ1)]ψ/(1+ψ).
Thus, labo supply is dec easing in he a e o p og essi i y
τ1
. I does no explici ly depend
on he wage a e
wi
because he income and subs i u ion e ec s on labo supply o se each
o he wi h a log u ili y o consump ion. We le
yi
=
wiNi
be he labo income and
Ui
he
indi ec u ili y o households o ype i.
3.2 Rep esen a i e Fi m
A ep esen a i e i m hi es bo h ypes o households o p oduce he inal good. The p oduc ion
unc ion is CES wi h an elas ici y o subs i u ion be ween he wo ypes o labo
η <
1.
The o al ac o p oduc i i y
α
(
A, D,R
) depends on an exogenous and pe sis en shock
A
,
he go e nmen ’s cu en s anding wi h i s c edi o s
D ∈ {
0
,
1
}
, and he households’ e ol
decision R∈ {0,1}. The i m p oduces
Y=α(A, D,R) [(θLNL)η+ (θRNR)η]1/η
We assume ha he unc ion
α
(
A, D,R
) is s ic ly dec easing in bo h
D
and
R
. Tha is, bo h
de aul s—which las un il he coun y eco e s i s good s anding wi h c edi o s—and e ol s
lead o a educ ion in he economy’s ou pu .
We assume he i m maximizes i s low p o i s. The labo ma ke is compe i i e. In
9
The pa ame e
−τ1
is hen equal o he elas ici y o he household’s “keep a e”, i.e., 1
−τ′
(
y
), wi h
espec o labo income y.
11

equilib ium, wages o households o ype ia e gi en by
wi=1 + θjNj
θiNiη(1−η)/η
α(A, D,R)θi.(2)
Exp ession (2) shows ha bo h de aul s and e ol s lead o income losses o he households.
No e also ha by in luencing he decision o households o e ol , he ax schedule a ec s
no only hei disposable incomes (“ edis ibu ion”) bu also hei p e- ax wages (“p e-
dis ibu ion”). Mo e gene ally, wages, labo supplies, and indi ec u ili ies depend on he
exogenous shock A, he iscal package D,τ, and he e ol decision R.
3.3 Pa ies
Pa y Ideology. The e a e wo poli ical pa ies, le and igh , indexed by
i∈ {L, R}
.
As o households, whene e we e e o pa y
i
, we deno e by
j
he opposi ion, and ice
e sa. Bo h pa ies a e bene olen and s a egic, bu hey di e in hei edis ibu i e as es.
Speci ically, hey bo h e alua e social wel a e acco ding o a weigh ed u ili a ian c i e ion, bu
hey assign di e en wel a e weigh s o he wo ypes o households. We deno e by Ω
i|i
and
Ω
j|i
he wel a e weigh s espec i ely assigned o households o ype
i
and
j
by pa y
i
, wi h
Ω
i|i>
1
/
2 and Ω
i|i
+ Ω
j|i
= 1. Thus, he igh -wing (le -wing) pa y places a highe Pa e o
weigh on he high-skilled (low-skilled) households—implying in u n ha households
R
(
L
)
op imally suppo he policies o pa y
R
(
L
). Besides hese he e ogeneous edis ibu i e
p e e ences, bo h pa ies a e iden ical; in pa icula , hey ha e he same discoun ac o
β
and ace he same exogenous de aul cos s.
Poli ical Tu no e . Each poli ical pa y is ei he he incumben (in-o ice) o he opposi ion
(ou -o -o ice). The p obabili y ha he incumben pa y
i
s ays in powe nex pe iod depends
on he households’ decision o e ol ,
πi|i
(
R
). The p obabili y ha he incumben pa y
i
is
ous ed and eplaced by pa y
j
is hen equal o
πj|i
(
R
) = 1
−πi|i
(
R
). We assume ha
πi|i
(
R
)
is s ic ly dec easing in
R
. Thus, he p obabili y o eelec ion, wi hou e ol s, is he same
o bo h pa ies; i is una ec ed by he de aul s a us, bu i is s ic ly lowe i a e ol akes
place in he cu en pe iod.10
10
This educed o m assump ion allows he model o ma ch he empi ical e idence showing ha incumben s
acing social un es a e eelec ed a lowe a es (Mades am e al. (2013)). A model whe e he median o e is
dissa is ied wi h he social un es and a leas pa ially blames he incumben o i could a ionalize his
assump ion.
12
Fiscal Policy. We deno e by
B
he cu en le el o deb ha he pa y in powe ,
i
, inhe i s
om he p e ious pe iod. I he coun y is cu en ly in good s anding wi h i s c edi o s,
pa y
i
ge s o choose whe he o de aul ,
D ∈ {
0
,
1
}
, an end-o -pe iod le el o deb
B′
, and a
ax-and- ans e policy
τ
consis ing o a le el o ans e s
τ0
and a a e o p og essi i y
τ1
.
11
I he pa y de aul s o has de aul ed in he pas and has no ye eco e ed i s good s anding
wi h c edi o s, hen
D
= 0 and
B′
= 0, i.e., he coun y canno bo ow on in e na ional
ma ke s. In his case, he ax schedule
τ
mus balance he cu en -pe iod budge . We use
he sho hand no a ion Fi≡ {D, B′, τ} o deno e he iscal package chosen by pa y i.
No e he so e eign decides how much deb o issue be o e knowing whe he households
will io —bu his ou come will in u n a ec he p ice o bo owing. As we explain below,
he le el o ans e s
τ0
needs o adjus ex pos in o de o ensu e ha he go e nmen budge
cons ain holds. Fo he sake o simplici y we omi he a gumen R om τ0in he sequel.
We assume ha deb is a long- e m con ac p omising a s eam o exponen ially declining
coupon paymen s. Speci ically, a uni o he bond issued a ime
p omises o pay (1
−
δ
)
+s−1
(
δ
+
z
) uni s o he consump ion good in pe iod
+
s
. The p ice o he newly issued
bonds aced by pa y
i
, deno ed by
Qi
(
A, B′,R
), is hen a unc ion o he exogenous o al
ac o p oduc i i y
A
, he le el o deb announced o nex pe iod
B′
(condi ional on which
he ini ial s ock o deb
B
and he ax policy
τ
a e i ele an ), and he households’ e ol
decision
R∈ {
0
,
1
}
. No e ha he ideology o he pa y in o ice
i∈ {L, R}
ma e s explici ly
o he bond p ice, since his a iable is pe sis en and he e o e in o ma i e abou u u e
bo owing and de aul choices. The go e nmen budge cons ain hen eads
0 = X
k∈{L,R}
τ(yk(A, D,R, τ1))+Qi(A, B′,R)[B′−(1 −δ)B]−(δ+z)B+κ(B, B′)(1−D).
(3)
The i s e m on he igh -hand side is he ax e enue, consis ing o he sum o axes
le ied on he labo incomes
yL, yR
o he wo ypes o households. Recall ha labo income
depends on he a iables
A, D,R
ia he o al ac o p oduc i i y
α
(
·
), and on he a e o
p og essi i y
τ1
ha a ec s labo supply decisions. The second e m on he igh -hand side
is he go e nmen ’s deb balance. I is non-ze o only i he go e nmen is able o bo ow,
i.e., i i is in good s anding wi h i s c edi o s so ha
D
= 0. The las e m in he cu ly
b acke s (
κ
(
B, B′
) is a con ex po olio adjus men cos ha penalizes he go e nmen o
la ge changes in he s ock o deb . We add his e m o a oid he well-known issue o ex eme
dilu ion immedia ely be o e a de aul ; a he calib a ed alues, less han 0.06% o ou pu is
11
The opposi ion pa y
j
does no make a iscal decision bu s ill ecei es a u ili y low acco ding o i s
own p e e ences.
13
spen on his adjus men cos . No e ha a e ol has he ex-pos e ec o al e ing he p ice
o bonds and ax e enue. Howe e , gi en a speci ic e ol choice, he budge cons ain (3)
ende s one o he iscal a iables edundan . The choices o end-o -pe iod deb
B′
(possibly
cons ained o ze o i
D
= 1) and ax p og essi i y
τ1
uniquely de e mine he le el o ans e s
τ0
. We assume ha he go e nmen commi s o he le els o bo owing and ax p og essi i y,
bu once he e ol decision is ealized, i adjus s he ans e s (
τ0
) o sa is y he budge
cons ain .
Following a epaymen o i s deb , he go e nmen keeps i s good s anding and can bo ow
in he nex pe iod. By con as , a e a de aul , he go e nmen is unable o bo ow un il i
eco e s i s good s anding and can e-en e c edi ma ke s. We assume ha his happens
wi h an exogenous posi i e p obabili y
γ∈
(0
,
1) in each pe iod. Le ing
G
(
D
) deno e he
p obabili y ha he go e nmen is in good s anding in he cu en pe iod, we hus ha e
G(D) = 1 i D= 0, and G(D) = γas long as D= 1.
Tas e Shocks. The inal ing edien o he model a e wo p i a ely obse ed as e shock
ec o s (
εg, εh
) ha a ec he poli ical pa ies and he households in each pe iod. The
go e nmen ’s shock
εg
(
B′,τ
) is a ec o con aining all he po en ial iscal packages ha a e
a ailable o he go e nmen .
12
The households’ shock
εh
(
R
)
∈R2
is a wo-dimensional ec o
associa ed wi h he cos s o e ol ing and o accep ing he iscal package. We d aw he as e
shocks om a gene alized ype-one ex eme alue dis ibu ion wi h scale pa ame e s
µg, µh
and co ela ion coe icien s
ρg, ρh
o he go e nmen and he households, espec i ely, wi h
ρh
= 0. The shocks a e independen ly and iden ically dis ibu ed (i.i.d.) o e ime and
unco ela ed o each o he .
3.4 Value Func ions
We w i e he model in ecu si e o m and “p imed” a iables (e.g.,
B′
) always ep esen he
nex -pe iod alues. We deno e by
W
he alue unc ions o he poli ical pa ies and by
V
he
alue unc ions o he households. Whene e a a iable has a supe sc ip o he o m
i|i
o
j|i
, he second a iable (
i
in hese wo examples) deno es he pa y ha is cu en ly in o ice.
12
Fo compu a ional simplici y, we assume ha he po en ial choices o end-o -pe iod deb
B′
and ax
p og essi i y
τ1
can only ake a ini e se o alues:
B′∈ B ≡ {B1, B2, ..., BNB}
and
τ1∈ T ≡ {τ1, τ2, ..., τNτ}
.
The e a e hus
NB×Nτ
possible iscal packages in epaymen and
Nτ
possible packages in de aul . Hence,
εg∈R(NB+1)×Nτ.
14
Pa ies. Suppose ha he coun y is in good s anding and ha pa y
i
is in powe ; hence,
pa y
j
is in he opposi ion. We le
Wi|i
and
Wj|i
deno e he alue unc ions o pa ies
i
and
j
espec i ely, gi en ha
i
is in o ice. Mo eo e , we le
Wi|i
D
and
Wj|i
D
be he alues ob ained
when pa y i epays he deb (D= 0) o de aul s (D= 1).
The pa y in o ice,
i
, chooses whe he o epay o de aul , so ha i s alue unc ion
sa is ies
Wi|i(A, B, εg) = max nWi|i
0(A, B, εg) ; Wi|i
1(A, εg)o,(4)
whe e he s a e a iables a e he exogenous le el o p oduc i i y
A
obse ed a he beginning
o he pe iod (and inhe i ed om he p e ious pe iod), he ini ial s ock o deb
B
, and he
ec o o as e shocks
εg
. The ini ial s ock o deb becomes i ele an once he go e nmen
has decided o de aul (
D
= 1). We deno e by
Di
(
A, B, εg
)
∈ {
0
,
1
}
he esul ing policy
unc ion ha de e mines he decision o de aul o epay by he pa y in powe
i
. Fo he
opposi ion pa y
j
, he e is no op imiza ion p oblem: I akes he policy unc ions o he
incumben iand o he households as gi en. I s alue unc ion is hen gi en by
Wj|i(A, B, εg) = (1 − Di(A, B, εg))Wj|i
0(A, B, εg) + Di(A, B, εg)Wj|i
1(A, εg).(5)
We now cha ac e ize he alues o epaymen and de aul . Fo simplici y’s sake, we only
de i e hose o he pa y in powe ,
Wi|i
0, Wi|i
1
, and omi hose o he opposi ion,
Wj|i
0, Wj|i
1
.
Pa y
i
makes i s iscal decisions aking in o accoun he households’ eac ion unc ion, i.e.,
whe he o e ol o no . Howe e , i canno pe ec ly p edic whe he a e ol will happen,
since i does no obse e he as e shocks
εh
. We deno e by
P
(
R|A, Fi
) he p obabili y ha
households o ype
j
make he decision
R∈ {
0
,
1
}
gi en he p oduc i i y shock
A
, which
a ec s hei wages, as well as he incumben pa y
i
’s choice o iscal package
Fi
=
{D, B′,τ}
.
In case o epaymen , we ha e
Wi|i
0(A, B, εg) = (6)
max
B′,τX
R∈{0,1}
P(R|A, Fi)

X
k∈{i,j}
Ωk|iUk(A, Fi,R) + βEhWi|i′(A′, B′, εg′)A, Ri


+εg(Fi)
subjec o he budge cons ain (3) wi h D= 0. In case o de aul , we ha e
Wi|i
1(A, εg) = (7)
max
τX
R∈{0,1}
P(R|A, Fi)

X
k∈{i,j}
Ωk|iUk(A, Fi,R) + βEhWi|i′(A′,0, εg′)A, Ri


+εg(Fi)
15
subjec o he budge cons ain (3) wi h
D
= 1. In hese exp essions,
εg
(
Fi
) is he alue
o he as e shock
εg
, gi en he chosen iscal package
Fi
=
{D, B′,τ}
, and he expec a ion
inside he cu ly b acke s is aken o e u u e p oduc i i y
A′
and he ideology o he nex
go e nmen
i′
, gi en he cu en p oduc i i y shock
A
and he household decision
R
. To
cha ac e ize his con inua ion alue, ecall ha pa y
i
is ous ed om o ice wi h p obabili y
πj|i
(
R
)
∈
(0
,
1); hus, when choosing he iscal package, he incumben pa y in e nalizes ha
i s p obabili y o s aying in powe is s ic ly smalle i ci izens decide o e ol . In addi ion,
ecall ha he go e nmen will be in good s anding wi h c edi o s in he nex pe iod wi h
p obabili y
G
(
D
), which is equal o 1 i i epays i s deb (
D
= 0), bu is s ic ly lowe han 1
i i de aul s (D= 1). We can hus w i e
EhWi|i′(A′, B′, εg′)i(8)
=EA′|A
X
i′∈{i,j}
πi′|i(R)nG(D)Wi|i′(A′, B′, εg′) + (1 − G(D))Wi|i′
1(A′, εg′)o,

whe e B′= 0 i D= 1.
Households. Gi en he s a e o he economy, he iscal package
Fi
=
{D, B′,τ}
chosen by
he incumben go e nmen o ype
i
and hei as e shock
εh
, households o he opposi e ype
j=imake he e ol decision.13 They sol e:
Vj|i(A, Fi, εh) = max
R∈{0,1}Vj|i
R(A, Fi, εh),(9)
whe e
Vj|i
0, V j|i
1
a e he alues o e ol ing and accep ing he iscal package, espec i ely. We
deno e by
Rj
(
A, Fi, εh
) he co esponding policy unc ion. By con as , he households ha
suppo he pa y in o ice do no make a e ol decision and ake all o he policy unc ions
as gi en. Thei alue unc ion is hen gi en by
Vi|i(A, Fi, εh) = (1 −Rj(A, Fi, εh))Vi|i
0(A, Fi) + Rj(A, Fi, εh)Vi|i
1(A, Fi).(10)
13
This assump ion is wi hou loss o gene ali y in a Ma ko equilib ium. Re ol ing agains a pa y o he
same ype as he households is no a c edible h ea , since i would dec ease cu en wages and inc ease he
odds o a u u e incumben who places a lowe wel a e weigh on he household.
16

In u n, he alues o e ol ing and accep ing he iscal package o household jsa is y
Vj|i
R(A, Fi, εh) = Uj(A, Fi,R) + βEA′|A
X
i′∈{i,j}
πi′|i(R)Vj|i′(A′,F′i′, εh′)
+εh(R),(11)
whe e
εh
(
R
) is he alue o he as e shock associa ed wi h he e ol o accep ance decision
R
. Re ol s he e o e lead o a decline in wages and hence in u ili y, in he cu en pe iod
bu inc ease he odds ha a household’s p e e ed pa y will be in powe nex pe iod. The
alue unc ions
Vi|i
0, V i|i
1
o households o ype
i
, who suppo he incumben , can be de i ed
analogously.
3.5 Fo eign Lende s
A con inuum o deep-pocke ed, isk-neu al, and compe i i e in e na ional lende s can buy
he go e nmen ’s bonds. Lende s ha e access o a one-pe iod isk- ee a e bond ha pays
in e es a e
. As is s anda d in he li e a u e, lende s a e o wa d-looking and p ice he
isk o de aul and deb dilu ion. Mo eo e , in ou en i onmen , lende s also in e nalize ha
he go e nmen ’s edis ibu i e p e e ences a y by pa y—and hence change o e ime—and
ha e ol s dec ease he odds o an incumben s aying in powe . When pa y
i
is in o ice,
he bond p ice ha sa is ies he lende s’ ze o-p o i condi ion is gi en by
Qi(A, B′,R) = 1
1 + EA′|A"1−X
i′∈{i,j}
πi′|i(R)Di′(A′, B′, εg′) (12)
×(δ+z+ (1 −δ)X
R′∈{0,1}
P(R′|A′,Fi′)Qi′A′, B′′ i′(A′, B′, εg′),R′)#,
whe e
B′′ i′
is he policy unc ion ha de e mines nex -pe iod’s bo owing o pa y
i′
. The
p ice o deb he e o e depends on he p obabili y o eelec ion o he incumben , no only
because o he de aul decision nex pe iod (as in, e.g., Scholl (2024)), bu also because
he ideology o he incumben changes he le el o u u e deb issuance and he e o e he
p obabili y o u u e de aul s. Long- e m deb also implies ha u u e poli ical ins abili y
( ep esen ed by he p obabili y o u u e e ol s) also has an e ec on he p ice o cu en
bonds. In a model wi h one-pe iod deb , bo h o hese channels would be absen .
17
3.6 Recu si e equilib ium de ini ion
De ini ion 1. Ma ko Pe ec Equilib ium (MPE). An MPE is de ined by alue unc-
ions
{Wi|i, Wj|i, Wi|i
D, Wj|i
D, V i|i, V j|i, V i|i
R, V j|i
R}
, policy unc ions
{Di, B′i,Fi, Ni,Ri}
, and
p ices {wi, Qi}, o all (i, j)∈ {L, R}2and (D,R)∈ {0,1}2, such ha :
1. Households’ policy unc ions sol e (1) and (9)-(11).
2. Pa ies sol e he dynamic p og amming p oblems (4)-(8).
3. Wages a e gi en by (2).
4. Bond p ices a e gi en by (12).
The logis ic shocks om D o kin e al. (2021) allow us o ind closed- o m solu ions o
all policy unc ions, and alue unc ions in expec a ion o he as e shocks. We sol e he
model nume ically using alue unc ion i e a ion.
4 Quan i a i e Analysis
The model is calib a ed a he qua e ly equency using A gen ine mac oeconomic da a.
A i s se o pa ame e s o alues ha a e ei he s anda d in he li e a u e o based on
his o ical A gen ine da a. We in e nally calib a e he emaining pa ame e s o ma ch ele an
momen s o A gen ina’s so e eign sp eads, poli ical u no e , equency o e ol s, and o he
business-cycle s a is ics. Table 2summa izes he pa ame e s se ou side he model.
We ake he i s se o pa ame e s om so e eign de aul models calib a ed o A gen ina.
The qua e ly isk- ee eal in e es a e
is se o 0
.
01, a s anda d alue o his pe iod. The
in e se F isch elas ici y is
ψ
=
.
5, in line wi h he alues used by A ellano e al. (2017) and
A ellano and Bai (2017) on so e eign deb models wi h labo . The ma u i y a e
δ
= 0
.
05
and i s coupon alue
z
= 0
.
03 a e se o he alues used by Cha e jee and Eyigungo (2012)
who also s udy A gen ina and ma ch he a e age ma u i y o he deb o 5 yea s and he deb
se ice. Simila ly, we assume ha he p oduc i i y shock ollows an AR(1) p ocess gi en by
ln
(
A
) =
ρAln
(
A −1
) +
ϵA
wi h
ϵA
∼N
(0
, σA
). Once again, we use Cha e jee and Eyigungo
(2012) es ima es o pa ame e s o an AR(1) endowmen income p ocess on de ended GDP
da a. We keep he pe sis ence a hei alues
ρA
= 0
.
95, and we adjus he ola ili y o
inno a ion such ha he simula ed ola ili y o ou pu ma ches ha o he da a:
σA
= 0
.
03.
The een y pa ame e is se o
γ
=
.
0385, which co esponds o an a e age exclusion pe iod
18
om c edi ma ke s a e de aul o 6 yea s and 6 mon hs.
14
We use Mo elli and Mo e i
(2023) es ima es o poli ical change in A gen ina as ou measu e o he a e age p obabili y o
eelec ion wi hou e ol s (i.e.,
πi|i
(
R
= 0) =
.
969). Wi hou e ol s, his co esponds o an
a e age enu e in o ice o 8 yea s o each poli ical pa y.
15
The eelec ion p obabili y unde
sus ained e ol s (
πi|i
(
R
= 1)) is a key de e minan o he bene i s o e ol ing and as such
will be in e nally calib a ed.
We ake a second se o pa ame e s om he li e a u e on skill p emia and inequali y in
La in Ame ica. Gallego (2006) analyzes 40 yea s o skill p emium da a in Chile ollowing
he same me hod as K usell e al. (2000). They measu e a labo elas ici y be ween skilled
and unskilled labo o 1
.
5 ha is consis en wi h
η
= 0
.
66, in line wi h es ima es o he
Uni ed S a es. We use da a on hou ly wages by educa ion g oup in A gen ina om he Socio-
Economic Da abase o La in Ame ica and he Ca ibbean (SEDLAC) da ase (SEDLAC
(2023)). The da a a e a ailable biannually om 2003 o 2021. The da ase spli s he
A gen inean labo o ce in o h ee g oups wi h di e en yea s o o mal educa ion ( ewe han
8 yea s, be ween 8 yea s and 13, and mo e han 13 yea s). Fo each g oup, we ha e hei size,
hou ly wages, labo hou s, and inally o al ne labo income. We use his o di ide he labo
o ce in o wo hal - iles o equal size. We ollow Hea hco e e al. (2017) and use equa ion (2)
o compu e he model p edic ion o he a io o hou s:
NR
NL=ψR
ψLψ
1+ψ
.
We use a e age hou s o each ype o hal - ile in he da a o es ima e
ψR, ψL
, inding
ha
NR
NL
= 1
.
15. No malizing by he mass o households and using ou es ima e o
ψ
, we
ob ain
ψR
= 0
.
60 and
ψL
= 0
.
40. Simila ly, we know om equa ion (2) ha he a io o
p e- ax wage in he model is:
wR
wL=(θR)η
(θL)ηNR
NL1−η
.
Using he p e ious esul o he a io o hou s, along wi h ou es ima e o
η
, we es ima e
θR
θL
= 2
.
3, and a e being no malized o he sum o one, his yields
θR
= 0
.
70 and
θL
= 0
.
30.
Table 3shows he pa ame e s o he model ha we calib a e in e nally. The s ochas ic
discoun ac o (
β
) is he same o bo h pa ies and he households. We ollow D o kin e al.
(2021) and assume a Gene alized Type One Ex eme Value dis ibu ion wi h scale pa ame e
14
This numbe co esponds o he a e age leng h o deb enego ia ion pe iod ac oss mul iple A gen ine
de aul s and is compu ed in Cha e jee and Eyigungo (2012) using da a om Benjamin and W igh (2009).
15
As in he Uni ed S a es, A gen ina’s p esiden ial elec ions a e held e e y ou yea s, and only one
eelec ion is pe mi ed.
19
Table 2: Pa ame e s Es ima ed ou side o he Model
Pa ame e Value Sou ce/T ansi ion
Risk- ee a e =.01 S anda d alue
In e se F isch elas ici y ψ=.5 S anda d alue
P oduc i i y shock ρA=.95 Cha e jee and Eyigungo (2012)
log(A ) = ρAlog(A −1) + ϵA
σA=.03 A gen ina’s GDP
Deb ma u i y δ=.05 A g. ma u i y o deb
Deb coupon z= 0.03 Deb se ice
Reen y p obabili y γ= 1/26 A e age enego ia ion leng h
Reelec ion odds unde s abili y πi|i(R= 0) = 1 −1/32 Mo elli and Mo e i (2023)
Elas ici y o subs i u ion η= 2/3Gallego (2006)
Labo p oduc i i y θR=.70, θL=.30 Hou ly wage p emia
Tas e o e o ψR=.60 Hou s op educa ion hal - ile
Tas e o e o ψL=.40 Hou s bo om educa ion hal - ile
σϵG
and co ela ion
ρϵG
o he iscal as e shock and scale pa ame e
σϵHH
o he e ol
decision o he households.
16
We also ake he unc ional o m o he po olio adjus men
cos o deb om D o kin e al. (2021):
ι(B′, B) = ι1exp(ι2|B′−B|)−ι1).
A he calib a ed pa ame e s, less han 6
e−
4 o ou pu is spen on hese cos s. To assess he
e ec o e ol s and de aul s on p oduc i i y, we bo ow he unc ional o m o he de aul
cos s om Cha e jee and Eyigungo (2012), and assume a simila ans o ma ion o he
e ol cos s:17
α(A, D= 0,R= 0) = A,
α(A, D= 0,R= 1) = A−max(ϕR
0A+ϕR
1A2,0),
α(A, D= 1,R= 0) = A−max(ϕD
0A+ϕD
1A2,0) ≡αD
R0,
α(A, D= 1,R= 1) = αD
R0−max(ϕR
0αD
R0+ϕR
1αD
R0
2,0).
(13)
These ans o ma ions add ou pa ame e s o calib a e in e nally. The i s wo,
ϕD
0, ϕD
1
,
co espond o he exogenous de aul cos s common in he so e eign de aul li e a u e. The
o he wo,
ϕR
0, ϕR
1
, ep esen he analogous penal y ha he economy su e s du ing a e ol .
16
The households’ as e shock is a scala and no a ec o ; hus, i has no co ela ion pa ame e . We allow
o he go e nmen ’s as e shocks o be co ela ed ac oss iscal packages (ρϵG= 1)
17
In Appendix Dwe e i y ha he main esul s o he pape a e obus o al e na i e speci ica ions o he
e ol cos s. In pa icula , we show ha ou esul s would hold quali a i ely unde u ili y cos s o e ol s.
20
0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2
0.7
0.75
0.8
0.85
0.9
0.95
1
1.05
1.1
1.15
1.2
(a) Bo owing Policy Func ions
0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2
-0.2
-0.1
0
0.1
0.2
0.3
0.4
0.5
(b) Tax p og essi i y.
Figu e 8: Expec ed iscal packages by pa y
No e: Panel (a) depic s he expec ed end-o -pe iod deb (
B′
) by pa y a ilia ion o he ule as a unc ion o
he ini ial le el o deb while keeping TFP ixed a mean alues. Since c edi ma ke s a e only a ailable in
epaymen , we only plo he bo owing choices when he ule ’s epaymen p obabili y is abo e 5%. Panel(b)
depic s he expec ed ax p og essi i y (
τ′
1
) by pa y a ilia ion o he ule as a unc ion o he ini ial le el o
deb while keeping TFP ixed a mean alues. Since he p og essi i y decision depends on he de aul decisions,
ax decisions a e depic ed wi h a solid line when he ule ’s epaymen p obabili y is abo e 5% and wi h a
do ed line when he p obabili y o de aul is abo e 5%.
Re ol p obabili ies: Panel (a) o Figu e 9depic s he p obabili ies o e ol agains each
pa y as a unc ion o he ini ial le el o deb . In line wi h p e ious igu es, we di e en ia e
be ween policy unc ions when he p obabili y o epaymen exceeds 5% (dashed lines) and
when he p obabili y o de aul exceeds 5% (do ed lines). This igu e p o ides in ui ion
o he posi i e associa ion be ween poli ical and iscal c ises, highligh ing why his e ec is
pa icula ly p onounced unde a igh -wing ule .
Fi s , e ol p obabili ies show a discon inuous inc ease o bo h pa ies when ansi ioning
om epaymen o de aul , as he oppo uni y cos s o e ol a e signi ican ly lowe in de aul
condi ions. Second, e en condi ional on epaymen , he p obabili y o e ol agains a
igh -wing ule —when he e is a posi i e p obabili y o de aul — ises wi h he ini ial le el
o deb . This inc eased isk o e ol occu s because he deb le els a which a igh -wing
ule is a isk o de aul all en i ely wi hin he de aul se s o he opposi ion. As a esul ,
he iscal policies enac ed by he igh -wing ule o sus ain epaymen a hese deb le els
a e pa icula ly in ole able o he opposi ion. In con as , we do no obse e a simila ise in
e ol p obabili ies agains a le -wing ule nea hei de aul h eshold, as hei epaymen
decision is consis en wi h wha he igh -wing opposi ion would do in he same scena io.
Righ -wing go e nmen s ace a o able schedules, bu e ol s wo sen hem: Panel
(b) in Figu e 9shows he p ice schedules o go e nmen deb , a he a e age TFP, by pa y
27

0.7 0.75 0.8 0.85 0.9 0.95 1 1.05 1.1 1.15 1.2
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
(a) P obabili y o Re ol
0 0.2 0.4 0.6 0.8 1 1.2
0
500
1000
1500
Sp ead (BPS)
(b) P ice o deb .
Figu e 9: Re ol p obabili ies and sp eads
No e: Panel (a) depic s he p obabili y o de aul and e ol by pa y a ilia ion o he ule as a unc ion o
he ini ial le el o deb while keeping TFP ixed a mean alues. Re ol p obabili ies depend on he de aul
decisions; as such, we deno e wi h dashed-line e ol p obabili ies when he ule ’s epaymen odds a e abo e
5% and wi h do ed lines he e ol p obabili ies when he de aul odds a e abo e 5%. Panel (b) plo s he
policy-implied sp eads as a unc ion o u u e deb o bo h pa ies in he baseline model unde s abili y
(do ed lines) and unde e ol (solid lines). We ix TFP a i s mean alue, and we ix ini ial deb o he mean
le el in he e godic dis ibu ion.
a ilia ion o he ule and by e ol s a us o he opposi ion. Consis en wi h he de aul se s
p e iously discussed, igh -wing go e nmen s bene i om a mo e a o able p ice schedule
compa ed wi h hei le -wing coun e pa s. Tha is, o any gi en end-o -pe iod deb choice,
a igh -wing ule pays a lowe in e es a e sp ead ela i e o he isk- ee a e han a
le -wing ule .24
Fu he mo e, when end-o -pe iod deb is high, igh -wing ule s ace highe sp eads when
he households e ol (
R
= 1) a he han op o poli ical s abili y (
R
= 0). This occu s
because o wa d-looking lende s an icipa e ha e ol s heigh en he likelihood o poli ical
u no e , and he obse ed deb le els all wi hin he de aul se s o he opposi ion. As a
esul , lende s demand highe sp eads o compensa e o his inc eased poli ical de aul isk.
25
24
Co oc e al. (2021) ind ha a simila esul holds a he c oss-coun y le el on a e age when a coun y
leans owa ds elec ing le -wing leade s.
25
A simila mechanism is p esen in e e se o he le -wing pa y in Appendix B. Tha is, e ol s lowe
bo owing cos s o le -wing go e nmen s. In he Appendix, we also show ha , as in he da a, mos e ol s
ha e a e y small e ec on he sp eads ou side o iscal c ises.
28
7 Equilib ium E godic Dis ibu ion
A e p esen ing he policy unc ions o each po en ial incumben , we now u n o he
desc ip ion o he e godic equilib ium. Table 5p o ides a de ailed b eakdown o e ol s
and de aul s in he e godic dis ibu ion by pa y. We calcula e he p opo ion o ime
spen in epaymen (
D
= 0) and de aul (
D
= 1), bo h in o al and when disagg ega ed by
incumben ype. Addi ionally, we epo he equency o e ol s (
R
= 1) uncondi ionally and
condi ionally depending on he de aul s a us and he ideology o he ule . Two asymme ies
be ween he pa ies become appa en once again.
Le -wing go e nmen s de aul mo e: A he e godic dis ibu ion, 27
.
7% o pe iods
a e spen in de aul ; speci ically, 15
.
9% o pe iods a e spen in de aul unde a le -wing
incumben and only 11
.
8% o hem a e spen in de aul unde a igh -wing incumben . Thus
72
.
3% o pe iods a e spen in epaymen , wi h his sha e almos e enly spli be ween he wo
ypes o ule s. The ac ha le -wing go e nmen s a e mo e associa ed wi h de aul s a ises
om he in e ac ion be ween asymme ies in de aul s se s (Figu e 7) and poli ical ansi ions.
P[D= 0] P[D= 1] P[R= 1] P[R= 1|D = 0] P[R= 1|D = 1]
Righ -wing ule 36.1 11.8 20.5 17.2 29.3
Le -wing ule 36.2 15.9 16.1 11.0 29.3
To al 72.3 27.7 36.6 28.2 58.7
Table 5: F equency o Each S a e in Pe cen a he E godic Dis ibu ion
No e: The i s and second columns epo he equency o epaymen and de aul s a es, a he e godic
dis ibu ion. The hi d column shows he uncondi ional equency o e ol s, while he ou h and i h columns
epo he condi ional equency o e ol s wi hin epaymen and de aul s a es, espec i ely. The bo om
line ow he o al equencies a he e godic, while he i s and second ows show he decomposi ion by pa y
ideology o he ule .
Re ol s a e mo e common du ing de aul s: When c edi ma ke s a e accessible, e ol s
occu 28% o he ime. In con as , du ing de aul pe iods, he equency o e ol s ises o
59%. Addi ionally, igh -wing incumben s ace a highe likelihood o e ol s in all s a es,
wi h his dispa i y being pa icula ly p onounced du ing epaymen pe iods. This asymme y
in e ol equency ul ima ely leads o di e ences in he ime each pa y spends in o ice.
Al hough he exogenous p oduc i i y cos o e ol s is independen o he incumben ’s
ideology, he oppo uni y cos o e ol ing is highe o mo e p oduc i e ( igh -leaning)
households, especially du ing epaymen pe iods when p oduc i i y is ela i ely high. Du ing
de aul pe iods, a e age p oduc i i y is lowe , which na ows he di e ence in he oppo uni y
cos o e ol ing ac oss all pa ies.
29
7.1 Poli ical De aul s: Righ - o-le - o-de aul T ansi ions
Poli ical ansi ions om igh o le a e associa ed wi h a heigh ened isk o de aul , as
igh -wing ule s can sus ain highe le els o deb compa ed wi h hei le -wing coun e pa s.
To illus a e his, we examine de aul pa e ns ollowing poli ical u no e s wi hin he e godic
dis ibu ion. Figu e 10 aces he du a ion o each pa y’s enu e om he momen i assumes
powe un il ei he a de aul occu s o powe is ans e ed o he opposi ion. By his de ini ion,
mos enu es las no longe han 60 pe iods.
10 20 30 40 50 60+
Pe iods since pa y change
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Cumula i e ac ion
De aul
Poli ical De aul Le
Tu no e
(a) Le Takeo e
10 20 30 40 50 60+
Pe iods since pa y change
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Cumula i e ac ion
De aul
Poli ical De aul Righ
Tu no e
(b) Righ akeo e
Figu e 10: Dynamics o de aul ollowing a swi ch in he ideology o he uling pa y
No e: Plo ed a e he cumula i e ac ions o le -wing( igh -wing) go e nmen s ha ha e de aul ed ac oss 60
pe iods since coming in o powe . Each enu e is s a ed by he le ( igh ) coming in o powe in epaymen .
The cumula i e ac ion is compu ed by isola ing e en s om a 100,000-pe iod simula ion, yielding 1,783
(1,586) akeo e s by he le ( igh ). A eas in pu ple co espond o s a es in which bo h ypes o incumben s
choose o de aul . A eas in blue ( ed) a e poli ical u no e s whe e only le -wing ( igh -wing) incumben s
de aul .
We plo he cumula i e numbe o enu es ending in u no e (g ay shaded a eas) and
de aul s o e ime o le -wing (panel (a)) and igh -wing (panel (b)) go e nmen s. A
dis inc ion is made be ween poli ical de aul s (blue and ed) and egula de aul s (pu ple).
Following Ha chondo e al. (2009), we de ine a poli ical de aul as a de aul ha occu s in
he i s pe iod a e a poli ical u no e and in a s a e whe e he p e ious incumben would
ha e op ed o epay.
This exe cise e eals a s a k asymme y: o e 15% o ansi ions om igh o le esul
in immedia e de aul s, almos all o which a e poli ical. These de aul s accoun o nea ly a
hi d o all de aul s obse ed unde le -wing go e nmen s.
26
In con as , no poli ical de aul s
a e obse ed when a igh -wing pa y akes powe . The simila i y in he size o he pu ple
26
When a le -wing go e nmen inhe i s a subs an ial deb bu den, i s al e na i e o de aul is dele e aging
h ough aus e i y measu es. The dynamics o such policies unde le -wing ule a e explo ed in Appendix C.
30
a eas in bo h panels sugges s ha poli ical de aul s a e he p ima y d i e o he asymme y
in de aul equencies be ween le -wing and igh -wing go e nmen s ou lined in Table 5.
Poli ical de aul s allow ou model o con ibu e o he policy deba e (IMF (2013) and
Guzman e al. (2016)) ha a gues ha coun ies de aul “ oo la e.”
27
Poli ical de aul s occu
when se icing he deb equi es eg essi e policies ha a e accep able o he igh bu no
o he le . In he lead-up o such a de aul , a planne wi h he le ’s p e e ences would
ha e de aul ed be o e he poli ical ansi ion akes place. In he A gen inean con ex his is
eminiscen o he 2020 de aul , bu also o he he ap u ous applause in Cong ess a e he
mo ion o de aul passed in 2001(La-Nacion (2001)).
7.2 Re ol s as an Endogenous De aul Cos
In ou model, e ol s a e mo e han wice as likely o occu du ing de aul pe iods han
du ing epaymen pe iods. This ea u e e ec i ely ans o ms e ol s in o an endogenous cos
o de aul . Incumben s who de aul should expec a sho e enu e. To assess he s eng h o
his de e en , we compa e he de aul dynamics o ou baseline model wi h hose o a model
wi h only exogenous poli ical u no e . To simpli y he compa ison, we se he exogenous
p obabili y o u no e o he alues we use in ou baseline unde s abili y.
Figu e 11, panel (a) compa es he de aul se s o a le -wing go e nmen in he baseline
model (da k blue) and in he coun e ac ual model wi h exogenous u no e (ligh blue). The
esul s indica e ha allowing o e ol s educes he size o he de aul se . In o he wo ds,
o ce ain combina ions o p oduc i i y and ini ial deb , a le -wing incumben would choose
o de aul in he model wi h exogenous ansi ions bu would op o epay in ou baseline
model. In panel (b), we examine his mechanism in e en mo e de ail by ocusing on one such
ini ial s a e: we ix he p oduc i i y a i s mean le el and he ini ial deb o 96% o mean
ou pu a he e godic dis ibu ion.
The i s pai o ba s shows he di e ence be ween he alue o epaymen and he alue o
de aul o he baseline model (da k blue) and he model wi h only exogenous u no e (ligh
blue). As expec ed, his di e ence is posi i e in he baseline (indica ing ha epaymen is
p e e ed) and nega i e in he exogenous u no e model (indica ing ha de aul is p e e ed).
The second and hi d pai s o ba s decompose his di e ence in o he con ibu ions o
cu en -pe iod u ili y and expec ed con inua ion alues in he u u e, espec i ely.
The esul s show ha e ol s ha e a limi ed e ec on cu en -pe iod alues bu signi ican ly
27
A ela ed b anch o he academic li e a u e has ocused on de aul enego ia ions ha a e ine icien ly
long; see, o ins ance, Bai and Zhang (2012), Benjamin and W igh (2009), and Asonuma and T ebesch
(2016).
31
(a) De aul Se s Le -wing Rule
To al alue U ili y Con inua ion
-0.06
-0.04
-0.02
0
0.02
0.04
0.06
(b) Repaymen minus de aul
Figu e 11: De aul decisions: Baseline and model wi h exogenous u no e
No e: In panel (a) we plo he de aul se s o he baseline model (da k blue) alongside he de aul se s (ligh
blue) o he model wi h exogenous u no e . In panel (b), we plo he di e ence be ween he alue o epaymen
and de aul while ixing he as e shocks a ze o, he TFP shock a i s mean alue, and he ini ial le el o
deb a 96% o mean ou pu in he baseline. In his s a e he alue o epaymen is highe han he alue
o de aul in he baseline (da k blue)) economy, while he opposi e is ue o he economy wi h exogenous
poli ical u no e (ligh blue). We u he decompose he o al di e ence in alues be ween he di e ences in
cu en u ili ies and he di e ences in con inua ion alues.
impac con inua ion alues. In he model wi h only exogenous ansi ions, de aul yields a
highe alue in bo h he cu en pe iod and he u u e. By con as , in he model wi h e ol s,
de aul ing s ill p o ides highe u ili y in he cu en pe iod, bu hese gains a e ou weighed
by he losses in con inua ion alues. This is because de aul ing aises he likelihood o acing
a e ol , which in u n educes he p obabili y o emaining in o ice.28
7.3
Implica ions o Re ol s and Tu no e o Deb and De aul
Risk
To e alua e he quan i a i e impac o e ol s on so e eign isk, we examine a ious speci ica-
ions o he baseline model. Table 6p o ides a summa y o he agg ega e momen s ac oss six
di e en model con igu a ions.
The i s wo speci ica ions a e models wi hou poli ical u no e , ep esen ing ei he a
pe manen ule by he igh -wing pa y o he le -wing pa y. These models e eal ha
28A po en ial conce n is ha in ou model, eelec ion p obabili ies depend no only on de aul s a us bu
also on pa y a ilia ion. Mo eo e , e ol s lead o a lowe a e age enu e in o ice a he e godic dis ibu ion
compa ed wi h he model wi h pu ely exogenous u no e conside ed he e. We add ess hese issues in
Appendix E, whe e we explo e al e na i e speci ica ions o he exogenous u no e model. Ou analysis
con i ms ha he esul s a e p ima ily d i en by di e ences in eelec ion a es based on de aul s a us.
32

in he absence o poli ical u no e , he incumben ’s ideology does no signi ican ly al e
he a e age le els o deb , sp eads, and de aul s. This is because bo h pa ies ace iden ical
exogenous de aul cos s and discoun ac o s.
Table 6: Momen s Compa ison be ween Models
Model speci ica ion Deb Sp ead F eq. de aul Re ol s Sha e in powe (Righ )
Pe manen le -wing 85.0 6.5 4.0 - 0.0
Pe manen igh -wing 86.7 6.3 3.9 - 100.0
Exogenous u no e 81.4 7.1 4.3 - 50.0
Re ol s only in epaymen 55.2 10.1 5.1 19.7 47.5
Re ol s only in de aul 118.4 6.1 3.6 13.9 50.9
Baseline 89.6 7.3 4.4 36.6 47.9
Da a 88.8 8.4 4.9 39.7 46.4
No e: We compa e key momen s o he da a and he baseline model wi h he i e al e na i e model speci ica ions.
The momen s a e compu ed using 100,000 simula ions o each model speci ica ion. Re ol s a e no possible in
he i s h ee speci ica ions and a e hus no epo ed.
We hen e u n o he model wi h only exogenous poli ical u no e discussed in he
p e ious subsec ion. The eelec ion p obabili y is kep a he alue unde s abili y in ou
baseline, and he e o e each pa y a e ages an eigh -yea e m in o ice. Rela i e o he
models wi h pe manen ule by one pa y, his speci ica ion shows lowe le els o deb , highe
sp eads, and mo e equen de aul s. As in Ha chondo e al. (2009) we ind ha poli ical
u no e on i s own can inc ease he equency o so e eign de aul s.
29
These addi ional
“poli ical de aul s” happen in hei amewo k when an impa ien go e nmen akes o e om
a mo e pa ien one. We add o hei insigh s by showing ha poli ical de aul s can a ise
because o ideological di e ences in he e iciency-equi y ade-o . Mo eo e , since in ou
calib a ion e ol s agains a igh -wing ule a e mo e common, e ol s can exace ba e de aul
isk by inc easing he likelihood o a poli ical de aul .30
The inal wo speci ica ions in ol e models whe e e ol s a e es ic ed o ei he epaymen
o de aul s a es. When e ol s a e allowed only du ing epaymen pe iods, sp eads inc ease,
de aul s become mo e equen , and he economy main ains signi ican ly lowe le els o
deb . In con as , when e ol s a e allowed solely du ing de aul pe iods, highe le els o
deb a e sus ained wi h lowe de aul equencies. In he i s scena io, de aul ing becomes
mo e a ac i e as i can ex end he incumben ’s enu e. In he second scena io, e ol s
29
This esul is also ound in he so e eign de aul li e a u e wi h epu a ion. In his li e a u e, one ype o
go e nmen is assumed o be mo e myopic han he o he . See, o ins ance, Mo elli and Mo e i (2023),
Amado and Phelan (2021), Al a o and Kanczuk (2005), and Cole e al. (1995).
30
Besides he di e ences in discoun ac o s o Ha chondo e al. (2009), asymme ic de aul a es by pa y
can eme ge because o a ia ions in exogenous eelec ion p obabili ies (as in Co oc e al. (2021)) o in deb
es uc u ing (as in C uces and T ebesch (2013)).
33
se e only as an addi ional cos o de aul , as hey p esen a new h ea o he incumben ’s
enu e. This h ea enhances commi men , allowing he economy o sus ain mo e deb . In
ou model, e ol s a e always possible bu endogenously obse ed mo e equen ly du ing
de aul s, e ec i ely ac ing as an endogenous de aul cos . Consequen ly, ela i e o he model
wi h only exogenous u no e , ou baseline speci ica ion sus ains signi ican ly highe le els o
deb a he cos o a modes inc ease in sp eads and de aul equency.
These quan i a i e esul s indica e ha e ol s in ou baseline model c ea e addi ional
iscal space ela i e o he model wi h only exogenous u no e . Howe e , his addi ional
iscal space is p ima ily u ilized o inc ease he deb by a ound 5% o ou pu . The nex
sec ion explo es he ole o each pa y in con ibu ing o his addi ional bo owing.
7.4 Re ol s and De ici s: A Theo y o “G adualism”
(a) De aul Se s o he Righ -wing Pa y
0 0.5 1 1.5
0
500
1000
1500
Sp ead
Righ -wing go e nmen , baseline
Righ -wing go e nmen , model wi h exogenous u no e
Righ -wing go e nmen , model wi h e ol s only in epaymen
Righ -wing go e nmen , model wi h pe manen pa ies
(b) Righ -wing p ice schedule.
Figu e 12: De aul se s and compa ison o policy-implied sp eads ac oss model a ian s.
No e: In panel (a) we plo he de aul se s implied by he policy unc ions co esponding o ou baseline and o
he model wi h exogenous u no e o he igh -wing pa y exclusi ely. This is done by adop ing he p e ious
de ini ion unde which he shaded a ea ep esen s an ex-an e p obabili y o de aul ha is g ea e han 0.5,
condi ional on being in good s anding ini ially. In panel (b). we plo he schedule o he baseline model(solid
line), he model wi h exogenous u no e (dashed line), he model wi hou e ol s in de aul (dashed line wi h
as e isks), and he model wi h pe manen ypes. Fo his pu pose, we ix p oduc i i y and he ini ial deb a
hei mean alue.
Figu e 12, panel (a), illus a es a compa ison o he de aul se s o a igh -wing ule in
ou baseline model e sus he model wi h only exogenous u no e . As be o e, e ol s a e
associa ed wi h a smalle de aul se . This implies ha , o a gi en deb le el and p oduc i i y,
a model accoun ing o e ol s p edic s lowe de aul isk, which should in luence lende
34
p icing. Panel (b) alida es his by showing he p ice schedules aced by a igh -wing ule
unde he di e en model speci ica ions ou lined in sec ion 7.3. Including e ol s in he model
leads o mo e a o able bo owing e ms.
Howe e , as Table 6shows, he a e age le el o sp eads and he de aul equency a he
e godic dis ibu ion in he baseline model a e sligh ly highe han hose in he model wi h
only exogenous u no e s. This di e ence can be a ibu ed o he ac ha he a e age
deb le el in he baseline model is signi ican ly highe han in he economy wi h exogenous
u no e s. Indeed, bo h pa ies—bu pa icula ly igh -wing ones— ake ad an age o he
imp o ed bond p ice schedules enabled by he e ol s o issue mo e deb ela i e o an
economy wi h only exogenous u no e s.
0 10 20 30 40 50 60
Pe iods since een y
0
0.2
0.4
0.6
0.8
1
1.2
Le Baseline
Le Exogenous Tu no e
(a) Deb unde a Le -wing Pa y
0 10 20 30 40 50 60
Pe iods since een y
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
1.1
Righ Baseline
Righ Exogenous Tu no e
(b) Deb unde a Righ -wing pa y.
Figu e 13: Deb le els a he e godic dis ibu ion a e een y.
No e: Plo ed in panel (a) (panel (b)) a e he a e age le els o deb o le -wing ( igh -wing) go e nmen s ha
een e wi h ze o deb a pe iod ze o and emain in powe and epaymen up o 60 pe iods la e a he e godic
dis ibu ion o ou baseline model. The a e ages a e compu ed by isola ing e en s om a 100,000-pe iod
simula ion. In bo h panels, we also compu e he hypo he ical bo owing decisions ha he same pa y would
ha e made in he model wi h only exogenous u no e gi en he same p oduc i i y shock.
Figu e 13 illus a es he con ibu ions o each pa y o he addi ional bo owing obse ed
in ou baseline model. I shows he e olu ion o deb a he e godic dis ibu ion ollowing
een y o c edi ma ke s (s a ing wi h ze o deb ) o bo h le -wing (panel (a)) and igh -
wing (panel (b)) go e nmen s. As a benchma k, we compu e coun e ac ual deb le els using
he bo owing ules om he model wi h only exogenous u no e , applied unde iden ical
shocks and pa y condi ions. The igu e e eals ha bo h pa ies issue mo e deb , hough
he inc ease is mo e p onounced o igh -wing go e nmen s. Two mechanisms d i e his
beha io .
Fi s , as discussed in Sec ion 7, igh -wing go e nmen s in he baseline model ace a
35
highe equency o e ol s du ing epaymen pe iods. As a esul , Figu e 13 sugges s ha
hese go e nmen s bo ow mo e o align iscal policies wi h opposi ion p e e ences and
mi iga e un es . This dynamic mi o s A gen ina’ss “G adualism” app oach unde he Mac i
adminis a ion, whe e conce ns abou social un es led o iscal concessions ha expanded
he budge de ici , when bo owing cos s we e s ill low.
Second, he igu e aligns wi h he poli ical economy heo y o “S a ing he Beas .”
31
This
heo y posi s ha igh -wing go e nmen s delibe a ely c ea e la ge budge de ici s o compel
u u e le -wing adminis a ions o implemen spending cu s a e a u no e . Al hough he
o iginal heo y does no add ess de aul s, ou model sugges s ha a le -wing pa y inhe i ing
a la ge deb bu den mus choose be ween poli ical de aul o adop aus e i y measu es o
es o e sus ainabili y. In Appendix C, we con i m his by examining le -wing go e nmen s
ha inhe i high deb le els wi hou eso ing o de aul .
Finally, in Appendix Fempi ically es s wo key p edic ions de i ed om ou model. Fi s ,
we con i m ha poli ical ansi ions om igh -wing o le -wing go e nmen s a e associa ed
wi h highe de aul isk, as e idenced by mo emen s in in e es a e sp eads du ing hese
ansi ions. Second, we demons a e ha igh -wing go e nmen s end o issue mo e deb
compa ed o hei le -wing coun e pa s.
8 Conclusion
We de elop a quan i a i e model o so e eign deb ha inco po a es pa ies wi h dis inc
p e e ences o edis ibu ion, poli ical ansi ions, and social un es . In ou amewo k,
e ol s impose economic cos s bu inc ease he p obabili y o poli ical u no e . Go e nmen s
wi h di e en ideologies s a egically choose iscal policies o sa is y hei cons i uen s and
ex end hei ime in o ice, while households s a egically decide when o e ol . We calib a e
he model o he economic and poli ical con ex o A gen ina. Speci ically, we ind ha ou
model aligns closely wi h he condi ions obse ed du ing he Mac i adminis a ion (2015-2019),
a pe iod cha ac e ized by social un es unde a igh -wing go e nmen , a subsequen poli ical
ansi ion om igh o le , and a so e eign de aul .
Ou model is quan i a i ely consis en wi h se e al empi ical pa e ns. Fi s , i cap u es
he posi i e ela ionship be ween poli ical c ises and so e eign isk, wi h poli ical ins abili y
leading o highe in e es a e sp eads, especially unde igh -wing leade ship. Second, we
ind ha igh -wing ule s ace mo e equen p o es s and ha p o es s a e mo e common
31
Fo a poli ical economy pe spec i e, see Ba le (2007), and o an economic applica ion, see Ma in
(2019).
36
O he da a sou ces : Ou c oss-coun y eg essions use da a on Ex e nal Deb , G oss
Domes ic P oduc (GDP), Cu en Accoun Balance, Rese es, and P ima y iscal balance.
The ex e nal deb da a is om he Join Ex e nal Deb Hub o he Wo ld Bank, In e -
na ional Mone a y Fund (IMF), and Bank o In e na ional Se lemen s (BIS). GDP da a
in na ional cu ency and U.S. dolla s a e om he Wo ld Bank’s Global Economic Mon-
i o , Ha e /Na ional Accoun sou ces, and Na ional Accoun sou ces in Global Financial
Da a. Cu en Accoun Balance, Rese es, and P ima y iscal balance a e om he IMF
In e na ional Financial S a is ics da a se . Pa y a ilia ion da a is om he In e -Ame ican
De elopmen Bank’s (IADB) 2020 Da abase o Poli ical Ins i u ions. De aul da a is om
Asonuma and T ebesch (2016) Mon hly De aul and Res uc u ing Da abase. Finally, he
c edi a ings used o selec coun ies o he eg ession and e en analysis, a e om S anda d
and Poo ’s coun y a ings o o eign exchange deb ex ac ed in July 2024. We use da a on
hou ly wages om he SEDLAC (2023) da ase , ex ac ed in Ap il 2023.
43

B Re ol s and Sp eads: Da a and model
1995-1 1995-7 1996-1 1996-7 1997-1 1997-7 1998-1 1998-7 1999-1 1999-7 1999-12
0
250
500
750
1000
1250
1500
1750
2000
Sp ead (bp)
22
24
26
28
30
32
Po i ica Risk
E)en Type
P o es
S ike
Bo h
P o es e s
( 100
( 1, 000
( 10, 000
( 100, 000
2nd e m
E ec ion
EMBI≳ Po i ica Risk
(a) 1995-1999.
2000-1 2000-7 2001-1 2001-7 2002-1 2002-7 2003-1 2003-6
0
1000
2000
3000
4000
5000
6000
7000
8000
Sp ead (bp)
De aul)
20
25
30
35
40
45
50
Poli)ical Ri(k
E en) T-pe
P o)e()
S) ike
Bo)
P o)e()e (
≳ 100
≳ 1
, 000
, 10, 000
, 100, 000
De la
R
≳
ua
Pue a
Rod
≳
iguez
De aul
Cama

no
Du alde
EMBI+ Poli ical Risk
(b) 2000-2003.
2003-6 2004-1 2004-7 2005-1 2005-7 2006-1 2006-7 2007-1 2007-7 2007-12
0
1000
2000
3000
4000
5000
6000
7000
Sp ead
(bp)
De aul
28
30
32
34
36
38
40
Poli ical Ri)k
E en Type
P(o e)
S (ike
Bo h
P(o e) e()
≳ 100
≳ 1
, 000
≳ 10, 000
≳ 100, 000
Inaugu a ion
Res uc u ing
begins
Deal
wi h IMF
Elec ion
EMBI+ Poli ical Risk
(c) 2003-2007.
2008 2009 2010 2011 2012 2013 2014 2015 2016
0
250
500
750
1000
1250
1500
1750
2000
Sp ead (bp)
De a)l(
30
32
34
36
38
40
Poli(ical Ris
E en( Type
P o(es(
S( i e
Bo(h
P o(es(e s
≳ 100
≳ 1
, 000
≳ 10, 000
≳ 100, 000
1s ≳Te m
Elec ion
2nd≳Te m
1s ≳missed
pa,men
Elec ion
EMBI
+ Poli ical Risk
(d) 2008-2015.
Figu e 14: Sp eads, poli ical isk, and daily p o es s
No e: Le e ical axis measu es he in e es a e sp ead using he EMBI+ in basis poin s. The igh e ical
axis co esponds o one hund ed minus coun y isk om he ICRG da abase. This is an index o poli ical
isk wi h high alues ep esen ing highe le els o poli ical ins abili y. Blue and o ange ci cles co espond
espec i ely o p o es s and gene al s ikes men ioned in he Dow Jones Fac i a da ase and a e associa ed
wi h iscal e o ms. The size o he do co esponds o he highes p o es size eco ded.
We ex end he daily analysis conduc ed o he Mac i adminis a ion, shown in Figu e 4,
o he adminis a ions ha go e ned A gen ina be ween 1995 and 2015. The esul s, plo ed
in Figu e 14, e eal ha he 2001 de aul also expe ienced simila equen episodes o social
un es , go e nmen u no e s, and apid inc eases in sp eads. Howe e , i is impo an o
no e ha ou side o hese de aul c isis pe iods, mos p o es s seem o only ha e a small
impac on sp eads.
In he model, p o es s in luence he eelec ion p obabili y o he incumben . Consequen ly,
hey can induce an inc ease in sp ead when a igh -wing go e nmen is in o ice, as illus a ed
in Figu e 9in he main body o he pape . Con e sely, p o es s can lead o a decline in
44
sp eads when a le -wing go e nmen is in o ice, as depic ed in panel (a) o Figu e 15, whe e
he y-axis is chosen o highligh his gap.
Al hough a signi ican change in sp eads exis s o ce ain po ions o he p ice schedule,
he sp eads paid unde s abili y and e ol also coincide o subs an ial po ions o he deb
suppo . A he e godic dis ibu ion, we ind e idence ha e ol s a ec sp eads a a simila
a e as he one obse ed in he da a. In panel (b) o Figu e 15, we plo he di e ence in
sp eads be ween s abili y and e ol o e 300 pe iods o he e godic dis ibu ion. The esul s
align wi h he da a, showing ha mos e ol s change he in e es a e sp ead by less han
30 basis poin s, wi h a ew a e ins ances ha ing a signi ican impac . On a e age, e ol s
inc ease sp eads by 8.6 basis poin s, wi h a s anda d de ia ion o 29.8 basis poin s.
0 0.2 0.4 0.6 0.8 1 1.2
0
0.5
1
1.5
2
2.5
3
3.5
4
4.5
5
Sp ead (BPS)
104
(a) Le -wing gap in sp eads. (b) Re ol s and sp eads a he e godic.
Figu e 15: E ec o e ol s on sp eads in he baseline model.
No e: Panel (a) plo s he policy-implied sp eads as a unc ion o u u e deb o bo h pa ies in he baseline
model unde s abili y(do ed lines), and unde e ol (solid lines). Fo his pu pose, we ix TFP o he mean
le el, and we ix ini ial deb o he mean le el in he e godic. This igu e is iden ical o Figu e 9wi h he
excep ion o he y-axis.
Panel (b) plo s a s ylized simula ion ha is mean o be an example o he median e ec o e ol s on sp eads.
The simula ion s a s in powe and las s o 300 pe iods. The ini ial deb le el is ixed o
B
= 1
.
23, which
co esponds o he median le el o deb o he igh -wing pa y in he e godic dis ibu ion. The ini ial
p oduc i i y le el is ixed o he mean le el, which is in he middle o he p oduc i i y g id.
We also e i y ha ou measu e o p o es s is consis en wi h he al e na i e measu e
ob ained by Ba e e al. (2022) wi h he Dow Jones Fac i a Da a o A gen ina. 16 p esen s
ou measu e o p o es s o A gen ina a he qua e ly le el desc ibed in 14 alongside he
Index o Social Un es compu ed by Ba e e al. (2022). The esul s o he main body o
he pape a e alida ed by his al e na i e measu e.
45
1995 1997 2000 2003 2006 2009 2012 2015 2018 2021
0
200
400
600
800
1000
Ba e e al
. (2022) Social Un es Index
De aul
0
2
4
6
8
10
12
14
16
E2en0s pe q1a 0e
E2en0 Type
P o0es0
Bo0h
S0 ike
P o0es0e s
≳ 100
≳ 1
, 000
≳ 10, 000
≳ 100, 000
Le 0
Righ0
Cen0e
Menem
2nd Te m
De la
R
≳
ua
3 Tu no e s
Duhalde
N
. Ki chne
C. Ki chne
1s Te m
C. Ki chne
2nd Te m
Mac i
Fe nandez
Figu e 16: Index o social un es and equency o p o es s in A gen ina 1995-2021
No e: We agg ega e a he qua e ly le el he index o Social Un es o Ba e e al. (2022) o
A gen ina, using he upda ed numbe s om Sep embe 2024.
C Aus e i y policies unde a le -wing ule
When a le -wing go e nmen akes powe and op s agains a poli ical de aul , i mus s ill
na iga e he subs an ial deb bu den le by i s igh -wing p edecesso . The only way o
es o e he deb o a le el mo e aligned wi h i s ideology is h ough aus e i y measu es.
In Figu e 17, we p esen he a e age e olu ion o a ious agg ega es a ound a poli ical
u no e om a igh -wing o a le -wing go e nmen ha does no in ol e a de aul .
35
Panel
(a) depic s he e olu ion o he deb - o-a e age-ou pu a io, panel (b) shows he e olu ion o
agg ega e consump ion (
CH
+
CL
) ela i e o a e age ou pu , and panel (c) illus a es he
e olu ion o he consump ion sha e o ype-
L
households as a pe cen age o o al consump ion
(CL
CL+CH).
Panel (a) shows ha he le -wing go e nmen begins i s e m by educing he deb by
app oxima ely 4% o a e age ou pu . The subsequen ise in deb issuance a e he i s
i e pe iods is pa ially explained by a selec ion bias, as economies expe iencing posi i e
p oduc i i y shocks a e o e ep esen ed in ou sample. Panels (b) and (c) illus a e how
aus e i y is implemen ed. The simul aneous decline in agg ega e consump ion and he inc ease
in he consump ion sha e o low-income households a e consis en wi h highe p og essi i y
(
τ1
) and inc eased o e all ax le els (
τ0
). In o he wo ds, aus e i y is achie ed h ough
disp opo iona ely highe axes on high-income households. Gi en he e iciency-equi y
ade-o in ou model, hese policies also esul in lowe o al ou pu , leading o a poli ical
35
The e en analysis ocuses on episodes wi hou de aul s o 10 pe iods be o e and a e he poli ical
u no e . Addi ionally, we exclude episodes whe e an addi ional u no e occu ed wi hin hese 20 pe iods.
46
-10 0 10
Pe iod
0.92
0.93
0.94
0.95
0.96
0.97
0.98
0.99
1
A e age deb le el
Le -wing akeo e
(a) Deb o mean ou pu .
-10 0 10
Pe iod
0.9
0.92
0.94
0.96
0.98
1
1.02
1.04
1.06
1.08
1.1
A e age o al consump ion
Le -wing akeo e
(b) Agg ega e consump ion
-10 0 10
Pe iod
0.24
0.25
0.26
0.27
0.28
0.29
0.3
0.31
0.32
0.33
A e age ax p og essi i y
Le -wing akeo e
(c) Redis ibu ion
Figu e 17: Aus e i y policies unde a le -wing go e nmen
No e: Plo ed a e episodes o poli ical ansi ion om a igh -wing go e nmen o a le -wing go e nmen in
he e godic dis ibu ion o he baseline model. The ansi ion happens a he pe iod labeled as 0. The episodes
a e de ined as chains in which he igh -wing pa y has been in epaymen o a leas 10 pe iods, o be hen
eplaced by a le -wing go e nmen ha emains in epaymen o a leas 10 pe iods a e he ansi ion.
Plo ed a e he associa ed se ies o deb ela i e o a e age GDP in he economy, o al consump ion ela i e o
a e age GDP, and consump ion o he L- ype household as a sha e o o al consump ion.
ecession.
D Al e na i e e ol cos s
We show he e ha ou esul s a e b oadly obus o di e en speci ica ions o he e ol cos s
(13). In all he speci ica ions we keep he alues o
ϕR
0
and
ϕR
1
a hei calib a ed alues om
sec ion 4. We compa e he momen s unde di e en speci ica ions o he e ol cos s o he
baseline, and he model wi h only exogenous ansi ions in Table 7.
In he i s al e na i e speci ica ion, we al e only he cos o e ol s du ing de aul s.
47
Table 7: Compa ison be ween di e en e ol cos speci ica ions.
Model speci ica ion Deb Sp ead F eq. de aul Re ol s Sha e in powe (Righ )
Baseline 89.6 7.3 4.4 36.6 47.9
Exogenous u no e 81.4 7.1 4.3 - 50.0
Re ol -De aul cos 89.6 7.2 4.4 36.6 47.9
U ili y cos 82.5 5.5 3.6 38.4 47.4
No e: We compa e key momen s o he baseline model and he model wi h only exogenous u no e o wo
al e na i e e ol cos speci ica ions. The momen s a e compu ed using 100,000 simula ions o each model
speci ica ion. The Re ol -De aul cos s a e s ill p oduc i i y cos s as in he baseline model, bu he addi ional
cos o e ol unde de aul is a unc ion o o al p oduc i i y (
A
) as opposed o de aul p oduc i i y (
αD
R
).
The u ili y cos speci ica ion imposes addi i e u ili y cos s p opo ional o p oduc i i y. These cos s a e no a
unc ion o he de aul s a us and a ec only he p o es e s, bu no he households o he o he ype.
Ins ead o he ou h exp ession in (13) we use :
α(A, D= 1,R= 1) = αD
R0−max(ϕR
0A+ϕR
1A2,0),
This speci ica ion inc eases e ol cos s du ing de aul s ela i e o he baseline non-
linea ly wi h p oduc i i y. Howe e , gi en ha de aul s will endogenously occu only when
p oduc i i y is low, he agg ega e momen s a e e y close o he ones in he baseline model.
In he second al e na i e speci ica ion, we use ins ead a u ili y cos o de aul , as opposed
o a p oduc i i y cos . Howe e , o keep hings compa able, we s ill make his u ili y cos
p opo ional o p oduc i i y.
36
A household o ype
i
will he e o e ha e he ollowing s a ic
u ili y in bo h de aul and epaymen s a es :
ui(C, N)−R×max(ϕR
0A+ϕR
1A2,0)
Compa ed o ou baseline model, his speci ica ion exhibi s a simila equency o e ol s
and igh -wing go e nmen s in powe . While he deb le el a he e godic dis ibu ion is
lowe , bo h he equency o de aul s and he sp eads a e also educed. When we compa e
his speci ica ion o he model wi h only exogenous u no e , we s ill obse e a highe le el o
deb issuance, albei a lowe sp eads and wi h ewe de aul s. The dec ease in deb ela i e
o ou baseline model is expec ed, as ou pu is highe when e ol s do no impac o al ac o
p oduc i i y (TFP). Re ol s a e mo e e ec i e as an endogenous de aul cos when hei
cos s a e modeled in e ms o u ili y losses o he p o es o s.
36
U ili y cos s o his o m a e now common in he so e eign deb li e a u e. See o ins ance Aguia and
Amado (2013), Bianchi and Sosa-Padilla (2023), and Roch and Uhlig (2018).
48

E Al e na i e models wi h only exogenous u no e
We show he e ha he di e ences be ween ou baseline model and he model wi h only
exogenous ansi ion in oduced in sec ion 7.2 a e p ima ily d i en by he asymme ic
eelec ion odds ha de aul and epaymen en ail o he ule . Speci ically, we sol e
al e na i e speci ica ions o he model wi h only exogenous u no e and compu e he
momen s o Table 6 o each o hem. We show ha speci ica ions ha do no allow o
eelec ion odds ha depend on he de aul s a us sus ain subs an ially less deb han ou
baseline model.
Table 8, summa izes he e godic dis ibu ion momen s o he baseline model, he model
wi h only exogenous u no e , and a ious al e na i e speci ica ions o he la e . Fo cla i y,
we e e o he exogenous u no e model p esen ed in he main pape as he No-Re ol s
model. In his model, he p obabili y o eelec ion is se o he alue unde s abili y in he
baseline model, esul ing in a highe a e age eelec ion p obabili y han he a e age eelec ion
odds a he e godic dis ibu ion o he baseline model.
In he i s al e na i e speci ica ion, we keep eelec ion p obabili ies cons an ac oss s a es
bu ma ch he a e age p obabili y o eelec ion a he e godic dis ibu ion o he baseline.
The esul ing momen s a e e y simila o hose in he No-Re ol s model. In he second
speci ica ion, we allow eelec ion p obabili ies o di e by pa y bu keep hem cons an ac oss
de aul o epaymen s a es. Again, he a e age deb le el and sp eads emain close o hose
in he No-Re ol s model. In he hi d e sion, we se equal exogenous eelec ion p obabili ies
o each pa y, bu le hem depend on he de aul o epaymen s a us, using condi ional
means om ou baseline. The esul ing momen s mo e closely esemble hose o he baseline
model han he No-Re ol s model. Finally, in he ou h speci ica ion, we allow eelec ion
p obabili ies o a y by pa y and by de aul s a us. The momen s o his speci ica ion a e
he closes o ou baseline bu s ill unde es ima e he le el o deb issuances by 3% o mean
ou pu . This sugges s ha he eelec ion odds channel is no he only way e ol s lead o
highe bo owing. Fo addi ional discussion abou he ole o e ol s o bo owing see sec ion
7.4.
F P edic ions o he model in he da a
We e i y se e al model p edic ions using A gen inean and c oss-coun y da a. In Sec ion 7we
explain ha igh -wing go e nmen s sus ain highe le els o deb , while le -wing go e nmen s
de aul mo e equen ly upon en e ing o ice. One implica ion o he i s obse a ion is
49
Table 8: Compa ison o al e na i e speci ica ions o he model wi h only exogenous u no e
Model speci ica ion Deb Sp ead F eq. de aul Sha e in powe (Righ )
Baseline 89.6 7.3 4.4 47.9
No- e ol s model 81.4 7.1 4.3 50.0
A e age eelec ion odds 81.4 6.7 4.1 51.0
Odds a y by pa y only 81.3 6.7 4.2 48.4
Odds a y by de aul s a us only 85.1 7.1 4.3 50.1
Odds a y by pa y and de aul s a us 86.43 6.9 4.1 47.7
No e: We compa e momen s o he baseline model and he model wi h exogenous u no e (No- e ol s) o wo
al e na i e speci ica ions o he model wi h only exogenous ansi ions. The momen s a e compu ed using
100,000 simula ions o each model speci ica ion. The a e age eelec ion odds model keeps he p obabili y o
u no e cons an a he a e age o he e godic dis ibu ion o he baseline model. The ou h line speci ies a
model whe e he p obabili y o u no e a ies by pa y bu is exogenously ixed o he condi ional means
measu ed a he e godic. The i h line speci ies a model whe e he p obabili y o u no e a ies by de aul
s a us bu is exogenously ixed o he condi ional means measu ed a he e godic. The six h line speci ies a
model whe e he p obabili y o u no e a ies by de aul s a us and pa y bu is exogenously ixed o he
condi ional means measu ed a he e godic.
ha we should expec mo e deb issuances unde igh -wing adminis a ions compa ed o
le -wing ones. Addi ionally, he second obse a ion sugges s ha , in he absence o de aul s,
we should an icipa e la ge inc eases in sp eads ollowing a Righ - o-Le ansi ion han a e
a Le - o-Righ . He e, we p o ide sugges i e e idence suppo ing he alidi y o hese wo
p edic ions in he da a.
Da a Model
−2
−1
0
1
2
3
4
5
A e age Yea ly Change in Le el o Deb o GDP
So ce
Le
Righ
(a) A gen ina changes in deb - o-GDP.
Da a Model
−3
−2
−1
0
1
2
3
4
5
6
A e age
Yea ly Change in Le el o Deb o GDP
Sou ce
Le
Righ
(b) C oss-coun y changes in deb - o-GDP.
Figu e 18: Fi s di e ence in annual deb - o-GDP by ideology o he execu i e.
No e: In panel (a) We plo he a e age annual change in deb as a pe cen age o GDP in A gen ina so ed
by ideology o he execu i e. We use annual cen al go e nmen deb as a pe cen age o GDP om he IMF
Global Deb Da abase. We combine his da a wi h poli ical in o ma ion om he DPI. This es ic s ou
s udy o yea s be ween 1975 and 2020. We use he pa y o ien a ion o he execu i e o assign each yea o a
poli ical leaning. In panel (b). We plo he a e age annual change in deb as a pe cen age o GDP o ou
c oss-coun y sample g ouped by which pa y was in powe . We use he same da a and me hods as panel (a).
We exclude yea s in which a coun y was in de aul as de e mined by Asonuma and T ebesch (2016) Mon hly
De aul and Res uc u ing Da abase.
50
Figu e 18 depic s he e olu ion o he deb - o-GDP a io by pa y a ilia ion o he
incumben execu i e om 1975 o 2020, excluding de aul s. The analysis is conduc ed a an
annual equency, using pa y a ilia ion da a om he DPI and annual cen al go e nmen
deb as a pe cen age o GDP om he his o ical IMF Global Deb Da abase. Panel (a)
displays he i s di e ence in le els o A gen ina and he baseline model, while panel (b)
p esen s he same s a is ics ac oss di e en coun ies. The esul s a e quali a i ely consis en
wi h he model p edic ions. In he model, le -wing go e nmen s ha do no de aul mus
ini ia e hei enu es wi h aus e i y measu es (see Appendix C), esul ing in sligh ly nega i e
changes in he s ock o deb an obse a ion ha aligns wi h he da a om A gen ina. In he
c oss-coun y da a, we ind inc eases in he deb - o-GDP a io o bo h ypes o go e nmen s,
bu he inc eases a e less p onounced unde le -wing adminis a ions.
3 Mon hs 6 Mon hs 12 Mon hs
Leng h Be o e and A e
0
50
100
150
200
250
Change in Sp eads A ound Tu no e (Basis Poin s)
Sou ce
Righ o Le
Le o Righ
(a) Da a changes in sp eads a e u no e .
3 Mon hs 6 Mon hs 12 Mon hs
Leng h Be o e and A e
0
50
100
150
200
250
Change in Sp eads A ound Tu no e (Basis Poin s)
Sou ce
Righ o Le
Le o Righ
(b) Model changes in sp eads a e u no e .
Figu e 19: Changes in he a e age sp ead a ound a poli ical u no e .
No e: In panel (a) We plo he a e age change in sp eads a ound poli ical u no e . We use DPI da a o ind
poli ical u no e s. We use he elec ion da e as he u no e momen and conduc a mon hly analysis. Fo
each u no e , we exclude he h ee mon hs immedia ely p eceding and immedia ely ollowing he elec ion
mon h. We hen ook he a e age sp ead o he gi en pe iod be o e and a e and ook a di e ence. We hen
ook an a e age o hese di e ences g ouped by u no e di ec ion. In panel (b). We plo he a e age sp ead
changes a ound poli ical u no e gene a ed by ou model, excluding pe iods o de aul and one qua e be o e
and a e he ansi ion.
Figu e 19 shows he a e age change in he CDS sp ead a ound a poli ical u no e . We
iden i y Righ - o-Le and Le - o-Righ u no e s using he DPI da ase , wi h he elec ion
day ma ked as he day o u no e . To mi iga e an icipa ion e ec s in he da a, we calcula e
he a e age sp ead while excluding he qua e be o e and a e he u no e in bo h he
model and he ac ual da a. Ou side o his exclusion pe iod, we de e mine he di e ence
be ween he a e age sp ead h ee mon hs (as well as six mon hs and wel e mon hs) a e
he elec ion and he a e age sp ead h ee mon hs (six mon hs, and wel e mon hs) p io o
51
he elec ion.
37
Consis en wi h he model, we obse e ha a e age sp eads end o inc ease
mo e ollowing Righ - o-Le u no e s han a e Le - o-Righ ansi ions.
37
We only pe o m his analysis on A gen inean da a o he 2020 u no e , which we examine in de ail in
he pape . The o he wo u no e s o which we ha e sp ead da a (2003 and 2015) occu ed while A gen ina
was in de aul and engaged in li iga ion wi h i s c edi o s.
52