Mae z, Se aphine F.; Edgell, Amanda B.; Wilson, Ma hew C.; Hellmeie , Sebas ian;
Lindbe g, S a an I.
A icle — Published Ve sion
Episodes o egime ans o ma ion
Jou nal o Peace Resea ch
P o ided in Coope a ion wi h:
WZB Be lin Social Science Cen e
Sugges ed Ci a ion: Mae z, Se aphine F.; Edgell, Amanda B.; Wilson, Ma hew C.; Hellmeie ,
Sebas ian; Lindbe g, S a an I. (2024) : Episodes o egime ans o ma ion, Jou nal o Peace
Resea ch, ISSN 1460-3578, Sage, Thousand Oaks, CA, Vol. 61, Iss. 6, pp. 967-984,
h ps://doi.o g/10.1177/00223433231168192
This Ve sion is a ailable a :
h ps://hdl.handle.ne /10419/310004
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/
h ps://doi.o g/10.1177/00223433231168192
Jou nal o Peace Resea ch
2024, Vol. 61(6) 967 –984
© The Au ho (s) 2023
A icle euse guidelines:
sagepub.com/jou nals-pe missions
DOI: 10.1177/00223433231168192
jou nals.sagepub.com/home/jp
1225162JPR0010.1177/00223433231168192Jou nal o Peace Resea chMae z e al.
esea ch-a icle2023
Regula A icle
Episodes o egime ans o ma ion
Se aphine F Mae z
Goe he Uni e si y F ank u , Depa men
o Poli ical Science
Amanda B Edgell
Uni e si y o Alabama
Ma hew C Wilson
Uni e si y o Sou h Ca olina, Depa men
o Poli ical Science
Sebas ian Hellmeie
WZB Be lin Social Science Cen e ,
T ans o ma ions o Democ acy Uni
S a an I Lindbe g
V-Dem Ins i u e, Uni e si y o Go henbu g,
Poli ical Science
Abs ac
This a icle p o ides a new concep ualiza ion o egime ans o ma ion ha allows schola s o add ess democ a iza-
ion and au oc a iza ion as ela ed bu ob e se p ocesses. We in oduce a da ase ha cap u es 680 episodes o egime
ans o ma ion (ERT) om 1900 o 2019 and o e s no el insigh s in o egime change o e he pas 120 yea s. The
ERT has h ee main ad an ages o e o he app oaches. Fi s , i a oids p oblema ic assump ions o uni homogenei y
and cons an as well as symme ic e ec s. Second, i in eg a es key insigh s om quali a i e s udies by ea ing egime
change as a g adual and unce ain p ocess. Thi d, he ERT is based on a uni ied amewo k o s udying egime
ans o ma ion in ei he di ec ion. The da ase di e en ia es be ween ou b oad ypes o egime ans o ma ion:
libe aliza ion in au oc acies, democ a ic deepening in democ acies, and au oc a iza ion in bo h democ acies and
au oc acies (democ a ic and au oc a ic eg ession). I u he dis inguishes en pa e ns wi h dis inc ou comes,
including s anda d depic ions o egime change (i.e. democ a ic ansi ion o b eakdown). A mino i y (32%) o
ERTs p oduce a egime ansi ion, wi h he majo i y o episodes ei he ending be o e a ansi ion akes place o no
ha ing he po en ial o such a ansi ion (i.e. u he democ a iza ion in democ a ic egimes o u he au oc a iza-
ion in au oc a ic egimes). We also p o ide compa isons o o he da ase s, illus a i e case s udies o demons a e
ace alidi y, and a discussion abou how he ERT amewo k can be applied in peace esea ch.
Keywo ds
au oc a iza ion, democ a iza ion, episodes, ans o ma ion, ansi ions
In oduc ion
Why do some dic a o ships ansi ion o democ acy,
whe eas o he s esis p essu es o libe alize o espond
wi h only minimal e o ms? Why do some democ acies
exhibi esilience, whe eas o he s expe ience backslid-
ing o b eakdown? Such ques ions abou egime change
ea u e p ominen ly in esea ch on con lic p ocesses
and cons i u e one o he mos in ensely esea ched
a eas in poli ical science. Ye , he dominan quan i a-
i e app oaches o s udying egime change equi e
imp obable assump ions, use deba able uni s o analy-
sis, and pu sue ques ions abou democ a ic b eakdown
and democ a ic ansi ion as sepa a e esea ch agendas,
Co esponding au ho :
[email p o ec ed]
968 jou nal o Peace Resea ch 61(6)
which hinde s a join and cohe en s udy o egime
change.
This a icle in oduces he Episodes o Regime T ans-
o ma ion (ERT) da ase . Using an inno a i e concep-
ual amewo k o o e come limi a ions in he li e a u e,
he ERT da ase cap u es p ocesses o egime ans o ma-
ion in ei he di ec ion along he democ a ic–au oc a ic
con inuum. I allows o esea ch on ou b oad ypes o
egime ans o ma ion, including libe aliza ion in au o-
c acies, democ a ic deepening in democ acies and
au oc a iza ion in bo h democ acies and au oc acies. The
ERT da ase dis inguishes en pa e ns wi h dis inc ou -
comes, including s anda d depic ions o egime change
(i.e. democ a ic ansi ion o b eakdown). The da ase
(V2.2) includes s a and end da es, as well as he ype
and ou come, o 680 episodes obse ed in he Va ie ies
o Democ acy (V-Dem) da ase om 1900 o 2019
(Coppedge e al., 2020). I is upda ed e e y yea wi h
a new elease o he V-Dem da ase .
1
The ERT da ase has h ee main ad an ages o e
o he app oaches. Fi s , i a oids p oblema ic assump-
ions o uni homogenei y and cons an and symme ic
e ec s. Second, i in eg a es key insigh s om quali a i e
s udies by ea ing egime change as a g adual and unce -
ain p ocess. Finally, he ERT da ase allows schola s o
s udy democ a iza ion and au oc a iza ion wi hin he
same sys ema ic amewo k. Fo quan i a i e esea che s,
he ERT da ase p o ides oppo uni ies o model he
causes and consequences o democ a iza ion and au o-
c a iza ion simul aneously while aking he e ogenei y in
he sample in o accoun . Fo quali a i e esea che s, i
p o ides key insigh s o single and compa a i e case
selec ion. In sho , he ERT da ase enables schola s o
analyze p ocesses, mechanisms and ou comes wi hin and
be ween de ined ERTs.
The ques ions aised he e a e also highly ele an o
he policy-p ac i ione communi y. Democ acy is asso-
cia ed wi h in e na ional and domes ic peace (Al man,
Rojas-de Gala e a & U dinez, 2021; Fjelde, Knu sen &
Nygå d, 2021), economic de elopmen (Acemoglu
e al., 2019) and en i onmen al p o ec ions (Fa zin &
Bond, 2006). Democ a ic ins i u ions p omo e in es -
men s in human de elopmen (Ge ing, Thacke &
Al a o, 2012) ha bene i o dina y ci izens h ough
imp o ed educa ion (Ansell & Lind all, 2013), heal h
(Bollyky e al., 2019) and gende equali y (Sunds o
¨m
e al., 2017). The e o e, knowledge abou how democ-
acy eme ges, declines, and dies is no me ely an aca-
demic exe cise.
A di ided li e a u e on egime change
Resea ch on egime change usually akes ei he a ansi-
ologis o inc emen alis pe spec i e (Table I). The an-
si ologis app oach ocuses on democ a ic ansi ions and
b eakdowns as disc e e e en s.
2
Fo example, classic
case-based wo ks on democ a ic ansi ions ocus on
ounding elec ions as momen s o disc e e egime change
(O’Donnell & Schmi e , 1986; Diamond e al., 1989).
While his li e a u e acknowledges complex p ocesses
and mul iple pa hways o unce ain ou comes, he goal
is o explain he ansi ion momen . Meanwhile, quan i-
a i e wo ks in his gen e use a dicho omous measu e o
Table I. Two dominan app oaches o he s udy o egime change
T ansi ologis Inc emen alis
On ological assump ions Democ acy and au oc acy a e dicho omous;
obse able ansi ion momen exis s
Democ acy-au oc acy con inuum; inc emen al
changes in ei he di ec ion a e meaning ul
equi alen s
Guiding ques ions Wha explains democ a ic ansi ion, su i al, and
b eakdown?
Wha explains changes in le els o democ acy?
Da a sou ces, key s udies Al a ez e al. (1996); Bjø nsko & Rode (2020);
Boix, Mille & Rosa o (2013); Cheibub,
Gandhi & V eeland (2010)
Acemoglu & Robinson (2006); Claassen
(2020); Jackman & Bollen (1989); Teo ell
(2010); Coppedge e al. (2022)
Limi a ions Assump ions o uni homogenei y; omi s
unsuccess ul a emp s; ansi ions as disc e e
e en s; democ a iza ion/au oc a iza ion as
sepa a e inqui ies
Assump ions o cons an and symme ic e ec s;
sho - un changes as disc e e e en s;
democ a iza ion/au oc a iza ion as empi ical
equi alen s
1
Fo he la es e sion isi h ps://gi hub.com/ demins i u e/ERT
2
While ‘ ansi ology’ usually e e s o a case-based li e a u e (e.g.
O’Donnell & Schmi e , 1986), quan i a i e s udies in es iga ing
disc e e egime ansi ions ha e simila on ological assump ions.
2jou nal o PEACE RESEARCH XX(X)
Mae z e al. 969
which hinde s a join and cohe en s udy o egime
change.
This a icle in oduces he Episodes o Regime T ans-
o ma ion (ERT) da ase . Using an inno a i e concep-
ual amewo k o o e come limi a ions in he li e a u e,
he ERT da ase cap u es p ocesses o egime ans o ma-
ion in ei he di ec ion along he democ a ic–au oc a ic
con inuum. I allows o esea ch on ou b oad ypes o
egime ans o ma ion, including libe aliza ion in au o-
c acies, democ a ic deepening in democ acies and
au oc a iza ion in bo h democ acies and au oc acies. The
ERT da ase dis inguishes en pa e ns wi h dis inc ou -
comes, including s anda d depic ions o egime change
(i.e. democ a ic ansi ion o b eakdown). The da ase
(V2.2) includes s a and end da es, as well as he ype
and ou come, o 680 episodes obse ed in he Va ie ies
o Democ acy (V-Dem) da ase om 1900 o 2019
(Coppedge e al., 2020). I is upda ed e e y yea wi h
a new elease o he V-Dem da ase .
1
The ERT da ase has h ee main ad an ages o e
o he app oaches. Fi s , i a oids p oblema ic assump-
ions o uni homogenei y and cons an and symme ic
e ec s. Second, i in eg a es key insigh s om quali a i e
s udies by ea ing egime change as a g adual and unce -
ain p ocess. Finally, he ERT da ase allows schola s o
s udy democ a iza ion and au oc a iza ion wi hin he
same sys ema ic amewo k. Fo quan i a i e esea che s,
he ERT da ase p o ides oppo uni ies o model he
causes and consequences o democ a iza ion and au o-
c a iza ion simul aneously while aking he e ogenei y in
he sample in o accoun . Fo quali a i e esea che s, i
p o ides key insigh s o single and compa a i e case
selec ion. In sho , he ERT da ase enables schola s o
analyze p ocesses, mechanisms and ou comes wi hin and
be ween de ined ERTs.
The ques ions aised he e a e also highly ele an o
he policy-p ac i ione communi y. Democ acy is asso-
cia ed wi h in e na ional and domes ic peace (Al man,
Rojas-de Gala e a & U dinez, 2021; Fjelde, Knu sen &
Nygå d, 2021), economic de elopmen (Acemoglu
e al., 2019) and en i onmen al p o ec ions (Fa zin &
Bond, 2006). Democ a ic ins i u ions p omo e in es -
men s in human de elopmen (Ge ing, Thacke &
Al a o, 2012) ha bene i o dina y ci izens h ough
imp o ed educa ion (Ansell & Lind all, 2013), heal h
(Bollyky e al., 2019) and gende equali y (Sunds o
¨m
e al., 2017). The e o e, knowledge abou how democ-
acy eme ges, declines, and dies is no me ely an aca-
demic exe cise.
A di ided li e a u e on egime change
Resea ch on egime change usually akes ei he a ansi-
ologis o inc emen alis pe spec i e (Table I). The an-
si ologis app oach ocuses on democ a ic ansi ions and
b eakdowns as disc e e e en s.
2
Fo example, classic
case-based wo ks on democ a ic ansi ions ocus on
ounding elec ions as momen s o disc e e egime change
(O’Donnell & Schmi e , 1986; Diamond e al., 1989).
While his li e a u e acknowledges complex p ocesses
and mul iple pa hways o unce ain ou comes, he goal
is o explain he ansi ion momen . Meanwhile, quan i-
a i e wo ks in his gen e use a dicho omous measu e o
Table I. Two dominan app oaches o he s udy o egime change
T ansi ologis Inc emen alis
On ological assump ions Democ acy and au oc acy a e dicho omous;
obse able ansi ion momen exis s
Democ acy-au oc acy con inuum; inc emen al
changes in ei he di ec ion a e meaning ul
equi alen s
Guiding ques ions Wha explains democ a ic ansi ion, su i al, and
b eakdown?
Wha explains changes in le els o democ acy?
Da a sou ces, key s udies Al a ez e al. (1996); Bjø nsko & Rode (2020);
Boix, Mille & Rosa o (2013); Cheibub,
Gandhi & V eeland (2010)
Acemoglu & Robinson (2006); Claassen
(2020); Jackman & Bollen (1989); Teo ell
(2010); Coppedge e al. (2022)
Limi a ions Assump ions o uni homogenei y; omi s
unsuccess ul a emp s; ansi ions as disc e e
e en s; democ a iza ion/au oc a iza ion as
sepa a e inqui ies
Assump ions o cons an and symme ic e ec s;
sho - un changes as disc e e e en s;
democ a iza ion/au oc a iza ion as empi ical
equi alen s
1
Fo he la es e sion isi h ps://gi hub.com/ demins i u e/ERT
2
While ‘ ansi ology’ usually e e s o a case-based li e a u e (e.g.
O’Donnell & Schmi e , 1986), quan i a i e s udies in es iga ing
disc e e egime ansi ions ha e simila on ological assump ions.
2jou nal o PEACE RESEARCH XX(X)
democ acy, eg essing disc e e changes in egime classi-
ica ion on explana o y ac o s o in e es (e.g. Bjø nsko
& Rode, 2020; Boix & S okes, 2003; B ownlee, 2009;
Eps ein e al., 2006; Hagga d & Kau man, 2012; Mille ,
2015; P zewo ski e al., 2000). As a esul , he ansi ion
momen becomes isola ed om he complex p ocesses
discussed in case s udies, e en i ecen wo k on he
‘s ickiness’ o au ho i a ian ins i u ions highligh s poli -
ical legacies (Albe us & Menaldo, 2018). Rega dless o
me hodology, he ansi ologis app oach makes wo
co e on ological assump ions: (1) egimes can be dicho -
omized in o he concep ual ca ego ies o democ acies
and au oc acies; and (2) he e is a dis inc , obse able
momen o ansi ion be ween democ acy and au oc acy.
By con as , he inc emen alis app oach explo es g a-
dual changes in le els o democ acy, usually wi h quan-
i a i e me hods, al hough hey migh be pai ed wi h
case s udies (e.g. Acemoglu & Robinson, 2006; Claas-
sen, 2020; Coppedge & Reinicke, 1990; Jackman &
Bollen, 1989; Le i z & Pop-Eleches, 2010; Teo ell,
2010; Coppedge e al., 2022).
3
Fo example, Teo ell
(2010) and Coppedge e al. (2022) p o ide an empi ical
o e iew o he de e minan s o democ a iza ion based
on annual changes, as well as annual up u ns and down-
u ns. These s udies a oid on ological assump ions abou
he dicho omous na u e o egimes o ansi ions as
e en s (Jackman & Bollen, 1989). Ins ead, hey ely
on wo di e en on ological assump ions: (1) democ acy
and au oc acy lie a opposi e ends o a con inuum; and
(2) equidis an changes in one di ec ion o ano he , and
a any place along he scale, a e meaning ul equi alen s.
Th ee co e limi a ions
The di ided li e a u e on egime change impedes e o s
a knowledge accumula ion. Each app oach in ol es a
‘bundle’ o assump ions and choices wi h ce ain ade-
o s. While a ew ecen wo ks a emp o combine
inc emen alism wi h ansi ology (e.g. Mainwa ing &
Bizza o, 2019; Hagga d & Kau man, 2021), such in e-
g a ed app oaches o s udying egime change a e in e-
quen and o e shadowed in he li e a u e domina ed by
he ansi ologis -inc emen alis di ide. As a esul , h ee
undamen al limi a ions in he li e a u e unde mine
e o s a knowledge accumula ion and p ac ical ele-
ance, which we aim o o e come wi h a uni ied ame-
wo k o s udying egime ans o ma ion.
Fi s , he ansi ologis app oach assumes uni homo-
genei y by ea ing all obse a ions wi hin he same
egime class as equi alen , e en hough he cases and
unde lying p ocesses o en di e . Ope a ionalizing
democ a iza ion as a disc e e ansi ion om non-
democ acy o democ acy based on a nominal-le el mea-
su e such as ha o Boix, Mille & Rosa o (2013) o
some h eshold on a con inuous indica o assumes ha
all au oc acies ha e an equal likelihood o ansi ioning
o democ acy, ce e is pa ibus. This igno es he e ogenei y
among au oc acies ha would be di icul o ully con ol
o , and ea s ully closed coun ies as equally likely o
ansi ion o democ acy as mo e libe al cases. Mo e c i-
ically, i ails o accoun o hose cases whe e democ a-
iza ion (o au oc a iza ion) s a ed bu a ansi ion ne e
occu ed. This means ha he p obabili y o democ a i-
za ion is based on compa ing hose ha success ully
democ a ized wi h a sample ha includes bo h hose ha
ne e libe alized and hose ha did. Igno ing he e oge-
nei y among he null uni s in he sample means o e -
looking ‘po en ially ele an and heo e ically e ealing
cases’ (Zibla , 2006: 24).
4
The inc emen alis app oach o e comes he assump-
ion o uni homogenei y by ope a ionalizing democ a-
iza ion as changes in le els o democ acy using a
con inuous measu e. Ye , his in oduces an equally ex-
ing assump ion o cons an and symme ic e ec s, which
ea s he same amoun o change as equi alen , and he
e ec s o causes as he same, ac oss he ull ange o
alues. I seems un ealis ic, o example, o assume ha
an annual change o 0.05 on a scale o 0–1 means he
same hing o Saudi A abia, which sco ed 0.02 in 2019,
as i does o Denma k, which sco ed 0.90. A e ages
based on a sample ha includes bo h coun ies ha
inc eased in he democ acy sco e and hose ha
dec eased also ea s hem as ha ing he same unde lying
causes. Ye , esea ch sugges s ha he ac o s causing
annual inc eases (democ a iza ion) and annual dec eases
(au oc a iza ion) in democ acy sco es a e o en di e en
(Teo ell, 2010; Coppedge e al., 2022).
Second, bo h app oaches ampli y sho - e m changes
when using quan i a i e me hods. Whe he measu ed as
a dicho omy o in e al, egime change is ypically
ea ed as an annual e en . Reg essing he p obabili y
o egime change (dicho omous o inc emen al) on an e-
ceden ac o s wi hou conside ing he p eceding g adual
changes isks misa ibu ing causes by in e p e ing he
3
We use he e m ‘inc emen alis ’ because hese s udies end o
ope a ionalize egime change in inc emen s, usually annual changes
in he alue o a con inuous measu e.
4
This p oblem is o en e e ed o as Simpson’s Pa adox (Wagne ,
1982).
Mae z e al. 3
970 jou nal o Peace Resea ch 61(6)
e ec s o a long- e m p ocess h ough sho - e m
changes in co ela ed independen a iables. This con-
as s wi h he longe , g adual, and highly unce ain
p ocesses o egime ans o ma ion desc ibed in he
case-based li e a u e (e.g. Rus ow, 1970; O’Donnell &
Schmi e , 1986; Acemoglu & Robinson, 2006; Boix,
2003).
Thi d, he wo app oaches ei he ea democ a iza-
ion and au oc a iza ion as sepa a e ields o inqui y o as
meaning ul equi alen s. Fo example, whe eas Linz
(1978) discusses he b eakdown o democ a ic egimes,
Linz & S epan (1996) ocus exclusi ely on democ a ic
ansi ions (and consolida ion), wi h li le b idging
be ween he heo ies. This end ca ies o e in o quan-
i a i e esea ch ha ypically heo izes and models dem-
oc a ic ansi ion and democ a ic b eakdown in sepa a e
publica ions. E en ecen wo k combining elemen s
om bo h app oaches only conside s backsliding o e o-
sion in democ acies (e.g. Mainwa ing & Bizza o, 2019;
Hagga d & Kau man, 2021). One excep ion is Ga es
e al. (2006), la e co obo a ed by Knu sen & Nygå d
(2015), showing ha hyb id egimes a e mo e likely o
expe ience egime ansi ions in ei he di ec ion. By con-
as , he inc emen alis app oach usually makes no dis-
inc ion be ween democ a iza ion and au oc a iza ion.
Equidis an annual changes on democ acy sco es a e
ea ed as empi ical equi alen s, ega dless o whe he
hose changes a e posi i e o nega i e. Two excep ions
a e Teo ell (2010) and Coppedge e al. (2022), who
show ha ac o s associa ed wi h annual inc eases and
dec eases a e o en dis inc , calling in o ques ion he
common app oach o ea ing he de e minan s o demo-
c a iza ion and au oc a iza ion as pu ely symme ical.
In sho , he li e a u e p esen s pa allel se s o explana-
ions o ela ed p ocesses, wi h a p oli e a ion o ja gon
(e.g. ‘democ a ic backsliding’ e sus ‘au oc a iza ion’) and
incomple e heo y building. The ansi ologis app oach
ea s egimes axonomically by dicho omizing hem and
he inc emen alis app oach iews egimes as a single class
o phenomenon whose a ibu es can be quan i ied along
a unidimensional con inuum.
5
As a esul , we know e y
li le abou whe he and how ansi ions in ei he di ec-
ion a e simila (o complemen s) o e ime, bo h in p o-
cess and hei de e minan s.
Episodes o egime ans o ma ion
We aim o o e come some key challenges in he li e a-
u e by de eloping a uni ying amewo k ha le e ages
he s eng hs o bo h he ansi ologis and inc emen al-
is pe spec i es. We b idge hese wo app oaches by con-
cep ualizing egime ans o ma ion as an inc emen al
p ocess ha may o may no yield a disc e e egime
ansi ion. Mo e speci ically, we concep ualize ERTs as
pe iods when a coun y unde goes sus ained and sub-
s an ial changes along a democ acy–au oc acy con i-
nuum. These episodes subs an i ely ans o m he
egime ( i ing wi h he inc emen alis app oach) bu
may no necessa ily yield a egime ansi ion ( om he
ansi ologis app oach).
6
Thus, we apply a ‘di ec ional’
de ini ion o egime ans o ma ion whe eby democ a i-
za ion and au oc a iza ion occu e en i he case does no
c oss some quali a i e h eshold be ween democ acy and
au oc acy (T eisman, 2020: 6).
As illus a ed in Figu e 1, we dis inguish episodes
based on hei di ec ion o mo emen along a con inuum
om libe al democ acy o closed au oc acy (Schedle ,
2001). We ea egimes as he same class o phenomena
ha can exhibi a ying deg ees o con o mi y o libe al
democ acy as an ideal ype (simila o he inc emen alis
app oach), while also acknowledging he impo an
di iding line be ween egimes ha ul ill he minimal
c i e ia o democ acy and hose ha do no (simila o
he ansi ologis app oach). We base hese minimal
c i e ia on he six ins i u ional gua an ees o
Figu e 1. Concep ualizing ERTs
5
This is possible because au oc acy is a ‘ esidual ca ego y’ (S olik,
2012) de ined by ‘wha i is no ’ (Linz, 1975), namely no democ acy.
6
While we conside changes be ween closed and elec o al au oc acy o
libe al and elec o al democ acy as ou comes o egime ans o ma ion,
we do no e e o hese as egime ansi ions.
4jou nal o PEACE RESEARCH XX(X)
Mae z e al. 971
e ec s o a long- e m p ocess h ough sho - e m
changes in co ela ed independen a iables. This con-
as s wi h he longe , g adual, and highly unce ain
p ocesses o egime ans o ma ion desc ibed in he
case-based li e a u e (e.g. Rus ow, 1970; O’Donnell &
Schmi e , 1986; Acemoglu & Robinson, 2006; Boix,
2003).
Thi d, he wo app oaches ei he ea democ a iza-
ion and au oc a iza ion as sepa a e ields o inqui y o as
meaning ul equi alen s. Fo example, whe eas Linz
(1978) discusses he b eakdown o democ a ic egimes,
Linz & S epan (1996) ocus exclusi ely on democ a ic
ansi ions (and consolida ion), wi h li le b idging
be ween he heo ies. This end ca ies o e in o quan-
i a i e esea ch ha ypically heo izes and models dem-
oc a ic ansi ion and democ a ic b eakdown in sepa a e
publica ions. E en ecen wo k combining elemen s
om bo h app oaches only conside s backsliding o e o-
sion in democ acies (e.g. Mainwa ing & Bizza o, 2019;
Hagga d & Kau man, 2021). One excep ion is Ga es
e al. (2006), la e co obo a ed by Knu sen & Nygå d
(2015), showing ha hyb id egimes a e mo e likely o
expe ience egime ansi ions in ei he di ec ion. By con-
as , he inc emen alis app oach usually makes no dis-
inc ion be ween democ a iza ion and au oc a iza ion.
Equidis an annual changes on democ acy sco es a e
ea ed as empi ical equi alen s, ega dless o whe he
hose changes a e posi i e o nega i e. Two excep ions
a e Teo ell (2010) and Coppedge e al. (2022), who
show ha ac o s associa ed wi h annual inc eases and
dec eases a e o en dis inc , calling in o ques ion he
common app oach o ea ing he de e minan s o demo-
c a iza ion and au oc a iza ion as pu ely symme ical.
In sho , he li e a u e p esen s pa allel se s o explana-
ions o ela ed p ocesses, wi h a p oli e a ion o ja gon
(e.g. ‘democ a ic backsliding’ e sus ‘au oc a iza ion’) and
incomple e heo y building. The ansi ologis app oach
ea s egimes axonomically by dicho omizing hem and
he inc emen alis app oach iews egimes as a single class
o phenomenon whose a ibu es can be quan i ied along
a unidimensional con inuum.
5
As a esul , we know e y
li le abou whe he and how ansi ions in ei he di ec-
ion a e simila (o complemen s) o e ime, bo h in p o-
cess and hei de e minan s.
Episodes o egime ans o ma ion
We aim o o e come some key challenges in he li e a-
u e by de eloping a uni ying amewo k ha le e ages
he s eng hs o bo h he ansi ologis and inc emen al-
is pe spec i es. We b idge hese wo app oaches by con-
cep ualizing egime ans o ma ion as an inc emen al
p ocess ha may o may no yield a disc e e egime
ansi ion. Mo e speci ically, we concep ualize ERTs as
pe iods when a coun y unde goes sus ained and sub-
s an ial changes along a democ acy–au oc acy con i-
nuum. These episodes subs an i ely ans o m he
egime ( i ing wi h he inc emen alis app oach) bu
may no necessa ily yield a egime ansi ion ( om he
ansi ologis app oach).
6
Thus, we apply a ‘di ec ional’
de ini ion o egime ans o ma ion whe eby democ a i-
za ion and au oc a iza ion occu e en i he case does no
c oss some quali a i e h eshold be ween democ acy and
au oc acy (T eisman, 2020: 6).
As illus a ed in Figu e 1, we dis inguish episodes
based on hei di ec ion o mo emen along a con inuum
om libe al democ acy o closed au oc acy (Schedle ,
2001). We ea egimes as he same class o phenomena
ha can exhibi a ying deg ees o con o mi y o libe al
democ acy as an ideal ype (simila o he inc emen alis
app oach), while also acknowledging he impo an
di iding line be ween egimes ha ul ill he minimal
c i e ia o democ acy and hose ha do no (simila o
he ansi ologis app oach). We base hese minimal
c i e ia on he six ins i u ional gua an ees o
Figu e 1. Concep ualizing ERTs
5
This is possible because au oc acy is a ‘ esidual ca ego y’ (S olik,
2012) de ined by ‘wha i is no ’ (Linz, 1975), namely no democ acy.
6
While we conside changes be ween closed and elec o al au oc acy o
libe al and elec o al democ acy as ou comes o egime ans o ma ion,
we do no e e o hese as egime ansi ions.
4jou nal o PEACE RESEARCH XX(X)
pa icipa ion and con es a ion (Dahl, 1971). The uppe
pa o Figu e 1 illus a es democ a iza ion as an o e -
a ching concep o episodes ha exhibi subs an ial and
sus ained imp o emen o democ a ic ins i u ions and
p ac ices (Wilson e al., o hcoming). Con e sely, he
lowe pa o Figu e 1 depic s au oc a iza ion as episodes
ha esul in a sus ained and subs an ial decline o dem-
oc a ic a ibu es (Lu
¨h mann & Lindbe g, 2019). Thus,
we conside au oc a iza ion and democ a iza ion as
ob e se egime ans o ma ion p ocesses.
We u he dis inguish episodes ha ha e he po en ial
o p oduce a egime ansi ion om hose ha en ich
quali ies cong uen wi h he cu en egime ype.
7
The
o me , ep esen ed by he dashed lines in Figu e 1,
include episodes o democ a iza ion in au oc acies (lib-
e alizing au oc acy) and episodes o au oc a iza ion in
democ acies (democ a ic eg ession). The la e , ep e-
sen ed by he solid lines in Figu e 1, include episodes
o democ a iza ion in democ acies (democ a ic deepening)
and episodes o au oc a iza ion in au oc acies (au oc a ic
eg ession).
Regime ans o ma ion p ocesses a e highly unce ain
and a ansi ion is nei he ine i able no he only
possible ou come (Schedle , 2001, 2013; T eisman,
2020). Figu e 2 depic s en possible pa e ns and ou -
comes o ERTs, all based on an in ui ion abou a non-
linea ela ionship be ween egime ans o ma ion and
measu es o democ acy. The do ed line illus a es he
bounda y be ween democ acy (abo e) and au oc acy
(below). Figu e 2(a) p o ides an o e iew o ou comes
o democ a iza ion episodes. A democ a ic ansi ion
occu s when an au oc a ic egime sees su icien e o ms
o c oss a minimal h eshold o democ acy and hen
holds a ounding democ a ic elec ion. We de ine a
ounding democ a ic elec ion as he i s ee and ai
elec ion held unde minimally democ a ic condi ions
a e which he elec ed o icials assumed o con inued
o ice in ei he he na ional legisla u e, execu i e, o con-
s i uen assembly. Libe alizing au oc acies can ail o
p oduce a democ a ic ansi ion in h ee ways. Fi s , he
egime could encoun e a p eemp ed democ a ic ansi ion
by achie ing minimally democ a ic condi ions bu ail-
ing o hold a ounding elec ion be o e e e ing back o
au oc acy. Second, au oc a ic egimes may unde go sub-
s an ial libe aliza ion be o e becoming a s abilized elec-
o al au oc acy. Thi d, a e expe iencing subs an ial
libe aliza ion, he egime could e e back o lowe le -
els o democ acy (i.e. e e ed libe aliza ion, Wilson
e al., o hcoming). Finally, o episodes o democ a ic
deepening, we conside he ou come a o egone conclu-
sion – e e ing o his as deepened democ acy.
Figu e 2(b) illus a es ou comes in au oc a iza ion epi-
sodes, which mi o he pa hways shown in Figu e 2(a). A
democ a ic b eakdown occu s when a democ a ic egime
d ops below he minimal h eshold o democ acy (indi-
ca ed by he dashed line in Figu e 2) and one o he
ollowing condi ions holds: (a) i becomes a closed au o-
c acy (i.e. no longe holds mul ipa y elec ions o he
execu i e o he legisla u e); (b) i holds a ounding
au ho i a ian elec ion o he execu i e, legisla u e, o a
cons i uen assembly; o (c) i emains au oc a ic o
Figu e 2. Rep esen a ion o ou comes o democ a iza ion (a, le ) and au oc a iza ion (b, igh ) episodes
7
In Figu e 1, ansi ions o democ acy and democ a ic b eakdowns
a e ep esen ed by he space be ween elec o al au oc acy and elec o al
democ acy bu a e included unde he dashed line because hey ha e
he po en ial o e e se.
Mae z e al. 5
972 jou nal o Peace Resea ch 61(6)
a su icien pe iod o ime o no longe be conside ed a
democ acy. Episodes o democ a ic eg ession may a oid
b eakdown in h ee ways. Fi s , a p eemp ed democ a ic
b eakdown occu s when a democ acy alls below he min-
imal h eshold o democ acy bu hen c osses back abo e
he h eshold be o e mee ing any o he c i e ia su icien
o b eakdown lis ed abo e. Second, a egime can decline
in democ a ic quali y be o e s abilizing as a diminished
democ acy. Thi d, democ acies may expe ience a e ed
eg ession by e e ing o some highe democ a ic s a e
a e a pe iod o subs an ial declines in democ a ic quali y.
Finally, we also conside he ou come a o egone conclu-
sion o au oc a ic eg ession, e e ing o his simply as
eg essed au oc acy.
Ope a ionalizing ERTs
Table II summa izes he de aul pa ame e s o iden i y-
ing episodes in he ERT da ase , which we de i ed om
a h ee-s ep alida ion p ocess ou lined below. We use
V-Dem’s Elec o al Democ acy Index (EDI) o ep esen
he con inuum om au oc acy o democ acy on a ange
o 0–1 (Coppedge e al., 2020). We code ERTs based on
an ini ial annual change o a leas +0.01 (s a inclu-
sion), ollowed by an o e all change o a leas +0.10
o e he du a ion o he episode (cumula i e inclusion).
ERTs a e conside ed ongoing as long as he EDI sco e
(i) changes a leas once e e y i e consecu i e yea s
( ole ance), (ii) does no ha e a e e se annual change o
0.03 o g ea e (annual u n), and (iii) does no expe i-
ence a cumula i e e e se change o 0.10 o e a i e-yea
pe iod (cumula i e u n). The inal yea o all episodes is
coded as he yea he case expe ienced a change o a leas
+0.01 a e episode onse and immedia ely p io o
expe iencing one o hese h ee condi ions o e mina-
ion. The episode is censo ed i he end da e co esponds
wi h he inal yea o he yea be o e a gap s a s in he
V-Dem coding o he coun y uni .
We coded he ou come o each episode based on ou
s ylized ep esen a ion in Figu e 2. We use he Regimes
o he Wo ld (RoW) (Lu
¨h mann, Tannenbe g & Lind-
be g, 2018) o es ima e a po en ial democ a ic ansi ion
o democ a ic b eakdown wi hin he episode. Ra he
han using a single a bi a y cu o on an in e al scale,
RoW iden i ies democ acies as egimes wi h su icien ly
ee and ai mul ipa y elec ions and whe e o he ins i-
u ions and p ac ices iden i ied by Dahl (1971) a e su i-
cien ly de eloped. RoW dis inguishes be ween closed
au oc acies wi h no mul ipa y elec ions and elec o al
au oc acies ha hold lawed mul ipa y elec ions. And
i di e en ia es elec o al democ acies – o hose egimes
wi h ee and ai mul ipa y elec ions – om libe al
democ acies ha also ha e su icien mino i y p o ec-
ions and ule o law.
When a case c osses he h eshold om he demo-
c a ic end o he spec um (i.e. libe al/elec o al democ-
acy) o closed au oc acy on he RoW measu e, we
au oma ically code his as a democ a ic b eakdown
because mul ipa y elec ions no longe exis . When a case
goes om being democ a ic o elec o al au oc acy, we
i s check o see ha i ei he (a) held elec ions ha we e
no ee and ai o (b) emained classi ied as an elec o al
au oc acy o he ole ance pe iod ( i e yea s). Fo dem-
oc a ic ansi ions, simila ly, we look o a change in
classi ica ion om au oc acy (ei he closed o elec o al)
o democ acy (ei he elec o al o libe al). We also check
o see whe he ee and ai elec ions occu ed and
whe he he winne s we e allowed o ake o ice.
8
Finally,
we code he o he (non- ansi ion) ou comes o bo h
democ a iza ion and au oc a iza ion based on he c i e ia
o de e mining episode e mina ion ollowing he con-
cep ualiza ions ou lined abo e and illus a ed in Figu e 2.
The ou come is censo ed o episodes ha ha e he
po en ial o a egime ansi ion bu a e ongoing in he
inal obse a ion yea o he da ase o be o e a gap in
coding.
Classi ying obse a ions as we ha e done wi h he
ERT da ase aces se e al common challenges, including
he ques ion o whe he concep ual ca ego ies a e
mu ually exclusi e and wha app op ia e h esholds
should be used o classi ica ion. We a i ed a ou
de aul pa ame e s h ough an ex ensi e alida ion p o-
cess, which we de ail in ou ‘s a ic’ Online appendix. In
sho , we aimed o ensu e he g ea es ace- alidi y based
on case e idence, while also cap u ing meaning ul
changes om he ea lies momen and minimizing o e -
lap be ween episodes. Fo addi ional anspa ency, we
p o ide a ‘dynamic’ Shiny app(endix) and an R package,
bo h o which allow use s o lexibly adjus he pa a-
me e s and o es how changes o he de aul h esholds
Table II. Ope a ionaliza ion o episodes
EDI pa ame e s Democ a iza ion Au oc a iza ion
S a inclusion 0.01 –0.01
Cumula i e inclusion 0.10 –0.10
Annual u n –0.03 0.03
Cumula i e u n –0.10 0.10
Tole ance 5.00 5.00
8
Using he V-Dem a iables 2el ype and 2elasmo .
6jou nal o PEACE RESEARCH XX(X)
Mae z e al. 973
a su icien pe iod o ime o no longe be conside ed a
democ acy. Episodes o democ a ic eg ession may a oid
b eakdown in h ee ways. Fi s , a p eemp ed democ a ic
b eakdown occu s when a democ acy alls below he min-
imal h eshold o democ acy bu hen c osses back abo e
he h eshold be o e mee ing any o he c i e ia su icien
o b eakdown lis ed abo e. Second, a egime can decline
in democ a ic quali y be o e s abilizing as a diminished
democ acy. Thi d, democ acies may expe ience a e ed
eg ession by e e ing o some highe democ a ic s a e
a e a pe iod o subs an ial declines in democ a ic quali y.
Finally, we also conside he ou come a o egone conclu-
sion o au oc a ic eg ession, e e ing o his simply as
eg essed au oc acy.
Ope a ionalizing ERTs
Table II summa izes he de aul pa ame e s o iden i y-
ing episodes in he ERT da ase , which we de i ed om
a h ee-s ep alida ion p ocess ou lined below. We use
V-Dem’s Elec o al Democ acy Index (EDI) o ep esen
he con inuum om au oc acy o democ acy on a ange
o 0–1 (Coppedge e al., 2020). We code ERTs based on
an ini ial annual change o a leas +0.01 (s a inclu-
sion), ollowed by an o e all change o a leas +0.10
o e he du a ion o he episode (cumula i e inclusion).
ERTs a e conside ed ongoing as long as he EDI sco e
(i) changes a leas once e e y i e consecu i e yea s
( ole ance), (ii) does no ha e a e e se annual change o
0.03 o g ea e (annual u n), and (iii) does no expe i-
ence a cumula i e e e se change o 0.10 o e a i e-yea
pe iod (cumula i e u n). The inal yea o all episodes is
coded as he yea he case expe ienced a change o a leas
+0.01 a e episode onse and immedia ely p io o
expe iencing one o hese h ee condi ions o e mina-
ion. The episode is censo ed i he end da e co esponds
wi h he inal yea o he yea be o e a gap s a s in he
V-Dem coding o he coun y uni .
We coded he ou come o each episode based on ou
s ylized ep esen a ion in Figu e 2. We use he Regimes
o he Wo ld (RoW) (Lu
¨h mann, Tannenbe g & Lind-
be g, 2018) o es ima e a po en ial democ a ic ansi ion
o democ a ic b eakdown wi hin he episode. Ra he
han using a single a bi a y cu o on an in e al scale,
RoW iden i ies democ acies as egimes wi h su icien ly
ee and ai mul ipa y elec ions and whe e o he ins i-
u ions and p ac ices iden i ied by Dahl (1971) a e su i-
cien ly de eloped. RoW dis inguishes be ween closed
au oc acies wi h no mul ipa y elec ions and elec o al
au oc acies ha hold lawed mul ipa y elec ions. And
i di e en ia es elec o al democ acies – o hose egimes
wi h ee and ai mul ipa y elec ions – om libe al
democ acies ha also ha e su icien mino i y p o ec-
ions and ule o law.
When a case c osses he h eshold om he demo-
c a ic end o he spec um (i.e. libe al/elec o al democ-
acy) o closed au oc acy on he RoW measu e, we
au oma ically code his as a democ a ic b eakdown
because mul ipa y elec ions no longe exis . When a case
goes om being democ a ic o elec o al au oc acy, we
i s check o see ha i ei he (a) held elec ions ha we e
no ee and ai o (b) emained classi ied as an elec o al
au oc acy o he ole ance pe iod ( i e yea s). Fo dem-
oc a ic ansi ions, simila ly, we look o a change in
classi ica ion om au oc acy (ei he closed o elec o al)
o democ acy (ei he elec o al o libe al). We also check
o see whe he ee and ai elec ions occu ed and
whe he he winne s we e allowed o ake o ice.
8
Finally,
we code he o he (non- ansi ion) ou comes o bo h
democ a iza ion and au oc a iza ion based on he c i e ia
o de e mining episode e mina ion ollowing he con-
cep ualiza ions ou lined abo e and illus a ed in Figu e 2.
The ou come is censo ed o episodes ha ha e he
po en ial o a egime ansi ion bu a e ongoing in he
inal obse a ion yea o he da ase o be o e a gap in
coding.
Classi ying obse a ions as we ha e done wi h he
ERT da ase aces se e al common challenges, including
he ques ion o whe he concep ual ca ego ies a e
mu ually exclusi e and wha app op ia e h esholds
should be used o classi ica ion. We a i ed a ou
de aul pa ame e s h ough an ex ensi e alida ion p o-
cess, which we de ail in ou ‘s a ic’ Online appendix. In
sho , we aimed o ensu e he g ea es ace- alidi y based
on case e idence, while also cap u ing meaning ul
changes om he ea lies momen and minimizing o e -
lap be ween episodes. Fo addi ional anspa ency, we
p o ide a ‘dynamic’ Shiny app(endix) and an R package,
bo h o which allow use s o lexibly adjus he pa a-
me e s and o es how changes o he de aul h esholds
Table II. Ope a ionaliza ion o episodes
EDI pa ame e s Democ a iza ion Au oc a iza ion
S a inclusion 0.01 –0.01
Cumula i e inclusion 0.10 –0.10
Annual u n –0.03 0.03
Cumula i e u n –0.10 0.10
Tole ance 5.00 5.00
8
Using he V-Dem a iables 2el ype and 2elasmo .
6jou nal o PEACE RESEARCH XX(X)
a ec he episodes.
9
While we a e con inced ha ou
ex ensi e alida ion p ocess based on case e idence is
bes sui ed o iden i y ERTs, al e na i e app oaches a e
concei able. Fo example, u u e esea ch could use
machine lea ning o iden i y common pa e ns in he
da a and classi y episodes acco dingly.
10
Based on ou coding ules, he ERT da ase (V2.2)
p o ides in o ma ion on he s a and end yea ,
ype, and ou come o 680 ERTs om 1900 o 2019.
Figu e 3 p o ides a summa y o hese episodes and
hei ou comes. Democ a iza ion accoun s o 63%
o ERTs (n¼427), wi h libe aliza ion in au oc acies
being a mo e common (n¼383) han deepening in
democ acies (n¼44). Ne e heless, democ a ic ansi-
ions appea o be he excep ion a he han he ule.
O e 60% o he ime (n¼226 ou o 371 uncenso ed
episodes) libe aliza ion does no yield a democ acy.
While democ a ic ansi ion occu s in only 39% o
he episodes whe e he ou come is known (n¼145 ou
o 371 uncenso ed episodes), a sizeable majo i y (77%,
n¼112) go on o expe ience u he democ a ic
deepening.
Figu e 3. Desc ip ion o ou sample o ERTs (1900–2019)
9
The ERT da ase , R package, and codebook a e a ailable he e:
h ps://gi hub.com/ demins i u e/ERT. The Shiny app(endix) is
a ailable he e: h ps://episodes.shinyapps.io/ alida ion/. Fo mo e
in o ma ion on ou coding ules, see he codebook (Edgell e al., 2020).
10
We hank one o he anonymous e iewe s o his sugges ion.
Mae z e al. 7
974 jou nal o Peace Resea ch 61(6)
In cases whe e libe aliza ion did no p oduce a
democ a ic ansi ion, we ind a highe equency
o e e ed libe aliza ion (33%, n¼123), in which
e o ms – whe he s a egic o genuine – ab up ly e e se
cou se o e a one o i e yea pe iod. Meanwhile, 87
episodes s all unde au oc acy and 16 o he episodes
come close o a democ a ic ansi ion, only o be p e-
emp ed. These cases demons a e he high le el o unce -
ain y o libe aliza ion in au oc acies, cons i u ing ‘nea
misses’ ha dominan app oaches ha e o e looked. The
ou come emains censo ed o 11 episodes as o 2019,
and he Ge man Democ a ic Republic is censo ed by
Ge man euni ica ion in 1990.
The lowe hal o Figu e 3 desc ibes episodes o au o-
c a iza ion, ep esen ing 37% o he ERTs (n¼253). A
clea majo i y o hese (62%, n¼157) in ol e u he
eg ession in al eady au oc a ic egimes. By con as ,
only 96 (38%) a ec democ acies. Among he 84 uncen-
so ed episodes o democ a ic eg ession, 65 (77%) lead
o a democ a ic b eakdown, ollowed by u he au oc a-
iza ion abou 79% o he ime (n¼51 ou o 65 b eak-
downs). No ably, only 19 democ acies su i ed
au oc a iza ion. A e ed eg ession is he mos common
way (74%, n¼14). Cases o p eemp ed democ a ic
b eakdown appea jus i e imes in he ERT da ase –
Mali (1997–1998), India (1971–1976), Geo gia (2006–
2010), Finland (1937–1940) and No h Macedonia
(2000) – and we obse e no cases o diminished democ-
acy. Fo 11 democ acies – including he Uni ed S a es
(since 2015) and India (since 2002) – he ou come
emains unde e mined as o 2019. Aus ia om 1931–
1938 is also censo ed by Ge man occupa ion, which
esul s in a gap in he V-Dem da a.
O e coming h ee co e limi a ions
The ERT da ase add esses h ee limi a ions o dominan
ansi ologis and inc emen alis app oaches o s udying
egime change. Fi s , he ERT da ase a oids assump-
ions o uni homogenei y and cons an and symme ic
e ec s. I suppo s s udying g adual p ocesses o egime
ans o ma ion by d awing on con inuous da a while also
enabling di e en ia ion o p ocesses and ou comes in a
ca ego ical way, allowing o he e ogenei y. The delinea-
ion o episodes based on hei ajec o ies encou ages
schola s o e alua e he di e ences in de elopmen pa -
e ns exhibi ed by he shapes in Figu e 2. In pa icula ,
ou app oach p o ides in o ma ion abou ‘nea misses’
whe e, despi e conside able po en ial, a egime ansi ion
did no occu , allowing us o compa e ‘success ul’ and
a ious ypes o ‘unsuccess ul’ cases.
Second, he ERT da ase allows us o s udy egime
change as an inhe en ly unce ain p ocess ha is some-
imes d ama ic and o he imes inc emen al. I ecog-
nizes bo h he ans o ma ion p ocess and ansi ion
e en as key elemen s o egime change. While we a e
no he i s o concep ualize egime changes wi hin ‘epi-
sodes’ (see, o example, Cassani & Tomini, 2020; D es-
den & Howa d, 2016; Gu ses, 2011; Lu
¨h mann &
Lindbe g, 2019; Papaioannou & Siou ounis, 2008;
Tilly, 2001), pas ea men s use he e m in he con ex
o c ea ing egime ypologies o disc e e obse a ions o
egime change.
Finally, ou app oach cap u es ERTs in ei he di ec-
ion (bo h democ a iza ion and au oc a iza ion) wi hin
one amewo k. This uni ies he li e a u e, while a oid-
ing assump ions abou he empi ical equi alence o
uni changes in opposi e di ec ions on he democ acy–
au oc acy con inuum. We see oppo uni ies o heo y
building abou whe he democ a iza ion and au oc a i-
za ion ha e simila causes (and e ec s) and new esea ch
ques ions, such as whe he sequen ially ob e se episodes
a e legacies o one ano he . In sum, es ablishing eplic-
able ules o iden i ying democ a iza ion and au oc a i-
za ion episodes and summa izing he ways ha hey
begin and end akes se iously calls o imp o ing esea ch
on egime change, bo h uni ying and expanding on he
li e a u e.
Compa isons wi h o he da ase s
We compa e he ou comes obse ed in he ERT da ase
wi h egime ansi ions ound in Boix, Mille & Rosa o
(BMR, 2013), Cheibub, Gandhi & V eeland (CGV,
2010), and when dicho omizing he con inuous Poli y
IV index a a sco e o 6 (Ma shall, Gu & Jagge s, 2019)
in Table III. The le column lis s he ERT da ase ou -
comes and hei equencies. The o he columns show
he numbe o democ a ic ansi ions o b eakdowns by
each o he bina y measu es ha all wi hin ERTs by
ou come. Many o he democ a ic ansi ions and dem-
oc a ic b eakdowns ound in hese commonly used da a-
se s o e lap wi h simila ou comes coded in he ERT
da ase , p o iding e idence o con e gen alidi y. Fo
democ a ic ansi ions, we see he g ea es o e lap wi h
he BMR measu e, accoun ing o 62 (43%) ou o 145
episodes in he ERT da ase , ollowed by Poli y (57,
39%) and CGV (46, 32%). Poli y shows sligh ly g ea e
o e lap when i comes o democ a ic b eakdown, wi h
30 episodes (46%) as compa ed wi h BMR wi h 26
episodes (40%). By con as , CGV only co esponds o
11 (17%) o democ a ic b eakdowns in ou sample. In
8jou nal o PEACE RESEARCH XX(X)
Mae z e al. 981
10% ( espec i ely) o he episodes ha p oduced a dem-
oc a ic ansi ion. By con as , a emp ed (14%) o suc-
cess ul coups (13%) a e mo e likely when libe aliza ion
ails o p oduce a democ a ic ansi ion. O e hal (51%)
and h ee-qua e s (78%), espec i ely, o hese
a emp ed and success ul coups occu in episodes esul -
ing in e e ed libe aliza ion, and no success ul coups
occu in episodes whe e s abilized elec o al au oc acy is
he ou come. This sugges s ha mili a y in e en ions
pose a h ea o libe aliza ion, bu a e less likely when
eli es manage he p ocess as a su i al ac ic. Fu he -
mo e, we ind ha he onse o democ a ic eg ession
is no associa ed wi h coups; a he , mili a y in e en-
ions ypically occu owa d he end o such episodes.
Mo e han one- hi d o episodes ha p oduced a demo-
c a ic b eakdown expe ienced a leas one coup, while
no a single success ul coup is obse ed du ing episodes
ha a oided democ a ic b eakdown. This sugges s ha
mili a y in e en ions o aid a ailing democ acy (almos )
ne e help and may ac ually con ibu e o democ a ic
b eakdown.
Figu e 7 illus a es how he ERT da ase can help
expand exis ing wo k on he ela ionship be ween con-
lic and poli ical egimes (see Heg e e al., 2001). We
ind e idence ha in e s a e con lic is associa ed wi h
ailed libe aliza ion. Almos 9% o episodes o libe al-
izing au oc acy ha did no p oduce a democ a ic ansi-
ion had one o mo e in e s a e con lic s e sus only 4%
o episodes wi h a democ a ic ansi ion. Fo ci il con-
lic , he di e ences a e less p onounced. Libe alizing
au oc acies expe ience simila a es o in as a e con lic
ega dless o whe he a ansi ion o democ acy occu s
(26% o ansi ions and 27% o non- ansi ion ou -
comes). Meanwhile, he a i y o in e na ional con lic
du ing democ a ic eg ession sugges s ha domes ic ac-
o s may be mo e salien o he e osion and b eak-
down o democ acy. We ind only one case whe e
in e na ional con lic occu ed du ing an episode o
democ a ic eg ession – he Indo-Pakis ani Wa o
1971, coinciding wi h India’s episode om 1971 o
1976, which managed o a oid b eakdown. By con as ,
we ind ha in as a e con lic occu s in abou 30%
o episodes esul ing in a democ a ic b eakdown, while
only wo episodes whe e democ acy su i ed (Venezuela
in 1992 and India 1971–1976) expe ienced in as a e
con lic . These insigh s also speak o ques ions abou
egime ype and con lic . The ERT da ase con ibu es
o peace esea ch by allowing o a mo e ine-g ained
empi ical analysis o he p opensi y o con lic in
coun ies expe iencing egime ans o ma ion and he
e ec o con lic on ansi ion ou comes.
Conclusion
The ERT da ase uni ies he bi u ca ed li e a u e on
egime change while also add essing i s p eca ious limi a-
ions. We de elop a amewo k ha eschews assump-
ions o uni homogenei y and cons an and symme ic
e ec s, and allows analysis o g adual egime ans o ma-
ion, while simul aneously ca ego izing ou comes and
iden i ying equi inali y. The ERT da ase acili a es
analyses o egime change as an inhe en ly unce ain
p ocess – some imes d ama ic, o he imes inc emen al.
I cap u es mo emen s in ei he di ec ion (democ a iza-
ion and au oc a iza ion), while a oiding assump ions
abou he empi ical equi alence o uni changes.
This a icle d aws se e al ini ial conclusions om he
ERT da ase (V2.2). Fi s , only some ERTs ha e he
po en ial o a egime ansi ion, and he e is no gua an-
ee ha such a ansi ion will occu . We ind ha only
abou 40% o au oc acies ha libe alize ac ually ansi-
ion in o democ acy. By con as , 77% o democ acies
expe iencing au oc a iza ion b eak down by he end o
he episode. The ERT da ase no only iden i ies epi-
sodes wi h a po en ial o ansi ion, i also includes
episodes o ‘democ a ic deepening’ and ‘au oc a ic
eg ession’, which a e o en ea ed as a sepa a e domain
(see, o example, he li e a u e on democ a ic consolida-
ion). In eg a ing hem alongside episodes wi h (po en-
ial) egime ansi ions is a aluable poin o compa ison.
Second, when a democ a ic ansi ion o b eakdown
occu s, in mos cases he coun y con inues o expe ience
u he democ a iza ion o au oc a iza ion, espec i ely.
The ansi ion e en is one s ep in a longe p ocess
a he han i s culmina ion. Thi d, democ a ic egimes
a e less p one o expe iencing egime ans o ma ion, in
ei he di ec ion, when compa ed wi h au oc acies.
Roughly 80% o he obse ed ERTs since 1900 ha e
occu ed in au oc acies. Thus, au ho i a ian egimes a e
gene ally less s able han democ acies. Fou h, democ a-
iza ion is much mo e common han au oc a iza ion.
This inding i s wi h he mode n expansion o democ-
acy h ough se e al global wa es o democ a iza ion
(Hun ing on, 1993). Ye , ou indings also suppo
he a gumen ha he wo ld is cu en ly in a wa e o
au oc a iza ion – nea ly wo- hi ds o coun ies unde -
going egime ans o ma ion a he end o 2019 we e
au oc a izing.
The ac ha we obse e many di e en ou comes –
no jus democ a ic ansi ion o b eakdown – shows
oom o g ow h in s udying egime ans o ma ion.
This is exempli ied by he s ep-wise de e io a ion o
democ acies like p esen -day Tu key and he bumpy
14 jou nal o PEACE RESEARCH XX(X)
oad o democ acy punc ua ed by ailed libe aliza ion
and eg ession in A gen ina. By emb acing an episodes
app oach, we suppo a esea ch agenda ha encou ages
bounded gene aliza ions abou a complex and inde e -
mina e phenomenon. Such conclusions may be p esen
in he li e a u e, bu he condi ions unde which ce ain
heo ies hold a e unclea and he explana ions o ela ed
p ocesses emain dispa a e. A p ocess-o ien ed app oach
o iden i ying and explaining egime ans o ma ion may
he e o e help kni oge he exis ing conclusions and
expand schola ly unde s anding o an impo an se o
ou comes.
Se e al a eas o esea ch awai explo a ion. Fo peace
s udies, schola s can use he ERT da ase o conduc
ine-g ained empi ical analysis o how con lic a ec s
egime ans o ma ion and ice e sa. As illus a ed he e,
coups, in e s a e and in as a e con lic may explain bo h
he onse and ou come o episodes. Regime ans o ma-
ion may also ac as he independen a iable, explaining
why some coun ies a e mo e bellige en han o he s.
Quali a i e esea che s may bene i om using he da a-
se o selec app op ia e cases. Fo example, he ERT
da ase allows esea che s o isola e cases ha ha e simila
s a ing alues o pai ed compa isons, wi h con lic as a
possible explana ion o di e gen ou comes o egime
ans o ma ion. In sho , by iden i ying s a es ha a e
in he p ocess o egime ans o ma ion, we can be e
unde s and he eme gence and decline o democ acy, as
well as whe e s a es a e mo e likely o see coups and
con lic , all o which ha e impo an academic and policy
implica ions.
Replica ion da a
Replica ion iles o he igu es in his a icle can be
ound a h p://www.p io.no/jp /da ase s. The ERT
da ase , R package, and codebook a e a ailable he e:
h ps://gi hub.com/ demins i u e/ERT. The Shiny
app(endix) is a ailable he e: h ps://episodes.shinyapps.
io/ alida ion/
Acknowledgmen s
The ERT amewo k and da ase is he ou come o se -
e al yea s o collabo a i e wo k. The o iginal concep ual
ounda ion was c ea ed by Lindbe g in he ERC g an
applica ion unding he p ojec . We ecognize (in alpha-
be ical o de ) Vanessa Boese-Schlosse , Pa ik Linden-
o s, Jean Lachapelle, Anna Lu
¨h mann, Joshua K usell,
Lau a Maxwell, Ju aj Medziho sky, and Richa d Mo -
gan, all o whom ha e ac i ely con ibu ed o yea s o
ial and e o ha pu us in a place o au ho his piece
and o launch he ERT da ase . We also hank Emilie
Segu a o skill ul esea ch assis ance.
Funding
The au ho (s) ecei ed inancial suppo o he esea ch,
au ho ship, and/o publica ion o his a icle: This
esea ch p ojec was p incipally suppo ed by Eu opean
Resea ch Council, Consolida o G an 724191, PI: S a -
an I Lindbe g; bu also by Knu and Alice Wallenbe g
Founda ion o Wallenbe g Academy Fellow S a an I
Lindbe g, G an 2018.0144; as well as by co- unding
om he Vice-Chancello ’s o ice, he Dean o he Col-
lege o Social Sciences, and he Depa men o Poli ical
Science a Uni e si y o Go henbu g. Se aphine F Mae z
ecei ed unding by he Ge man Resea ch Founda ion,
p ojec no. 421517935.
ORCID iDs
Se aphine F Mae z h ps://o cid.o g/0000-0002-
7173-9617
Amanda B Edgell h ps://o cid.o g/0000-0002-3029-
691X
S a an I Lindbe g h ps://o cid.o g/0000-0003-03
86-7390
Re e ences
Acemoglu, Da on & James A Robinson (2006) Economic O i-
gins o Dic a o ship and Democ acy. New Yo k: Camb idge
Uni e si y P ess.
Acemoglu, Da on; Su esh Naidu, Pascual Res epo & James A
Robinson (2019) Democ acy does cause g ow h. Jou nal o
Poli ical Economy 127(1): 47–100.
Adcock, Robe & Da id Collie (2001) Measu emen alid-
i y: A sha ed s anda d o quali a i e and quan i a i e
esea ch. Ame ican Poli ical Science Re iew 95(3): 529–546.
Albe us, Michael & Vic o Menaldo (2018) Au ho i a ianism
and he Eli e O igins o Democ acy. New Yo k: Camb idge
Uni e si y P ess.
Als on, Lee J & And es A Gallo (2010) Elec o al aud, he ise
o Pe on and demise o checks and balances in A gen ina.
Explo a ions in Economic His o y 47(2): 179–197.
Al man, Da id; Fede ico Rojas de Gala e a & F ancisco U di-
nez (2021) An in e ac i e model o democ a ic peace. Jou -
nal o Peace Resea ch 58(3): 384–398.
Al a ez, Mike; Jose
´An onio Cheibub, Fe nando Limongi &
Adam P zewo ski (1996) Classi ying poli ical egimes.
S udies in Compa a i e In e na ional De elopmen 31(2):
3–36.
Mae z e al. 15
982 jou nal o Peace Resea ch 61(6)
Ansell, Ben & Johannes Lind all (2013) The poli ical o igins
o p ima y educa ion sys ems: Ideology, ins i u ions, and
in e denomina ional con lic in an e a o na ion-building.
Ame ican Poli ical Science Re iew 107(3): 505–522.
Bashi o , Galib & Ca oline Lancas e (2018) End o mode a-
ion: The adicaliza ion o AKP in Tu key. Democ a iza ion
25(7): 1210–1230.
Bjø nsko , Ch is ian & Ma in Rode (2020) Regime ypes
and egime change: A new da ase on democ acy, coups,
and poli ical ins i u ions. Re iew o In e na ional O ganiza-
ions 15(2): 531–551.
Boix, Ca les (2003) Democ acy and Redis ibu ion. New Yo k:
Camb idge Uni e si y P ess.
Boix, Ca les & Susan C S okes (2003) Endogenous democ a-
iza ion. Wo ld Poli ics 55(4): 517–549.
Boix, Ca les; Michael K Mille & Sebas ian Rosa o (2013) A
comple e da a se o poli ical egimes, 1800-2007. Com-
pa a i e Poli ical S udies 46(12): 1523–1554.
Bollyky, Thomas J; Ta a Templin, Ma hew Cohen, Diana
Schode , Joseph L Dieleman & Simon Wigley (2019) The
ela ionships be ween democ a ic expe ience, adul heal h,
and cause-speci ic mo ali y in 170 coun ies be ween 1980
and 2016: An obse a ional analysis. Lance 393(393):
1628–1640.
B ownlee, Jason (2009) Po en s o plu alism: How hyb id
egimes a ec democ a ic ansi ions. Ame ican Jou nal o
Poli ical Science 53(3): 515–532.
Cassani, And ea & Luca Tomini (2020) Re e sing egimes
and concep s: F om democ a iza ion o au oc a iza ion.
Eu opean Poli ical Science 19(2): 272–287.
Cheibub, Jose An onio; Jenni e Gandhi & James Raymond
V eeland (2010) Democ acy and dic a o ship e isi ed.
Public Choice 143: 67–101.
Chen, Linda (2007) A gen ina in he wen y- i s cen u y. In:
Howa d J Wia da & Ha ey F Kline (eds.) La in Ame ican
Poli ics and De elopmen . Boulde , CO: Wes iew:
103–122.
Claassen, Ch is ophe (2020) Does public suppo help
democ acy su i e? Ame ican Jou nal o Poli ical Science
64(1): 118–134.
Coppedge, Michael; Amanda B Edgell, Ca l Hen ik Knu sen
& S a an I Lindbe g (2022) Why Democ acies De elop and
Decline. New Yo k: Camb idge Uni e si y P ess.
Coppedge, Michael; John Ge ing, Ca l Hen ik Knu sen, S a -
an I Lindbe g, Jan Teo ell, Da id Al man, Michael Be n-
ha d, M. S e en Fish, Adam Glynn, Allen Hicken, Anna
Lu
¨h mann, Kyle L Ma qua d , Kelly McMann, Pamela
Pax on, Daniel Pems ein, B igi e Seim, Rachel Sigman,
S end-E ik Skaaning, Je ey S a on, S e en Wilson, Agnes
Co nell, Nazi a Alizada, Lisa Gas aldi, Haakon Gje løw,
Ga y Hindle, Nina Ilchenko, Lau a Maxwell, Vale iya
Mechko a, Ju aj Medziho sky, Johannes on Ro
¨me , Aksel
Sunds o
¨m, Ei an Tzelgo , Yi- ing Wang, To e Wig &
Daniel Zibla (2020) V-Dem [coun y–yea /coun y–da e]
da ase 10 (h ps://doi.o g/10.23696/ demds20).
Coppedge, Michael & Wol gang H Reinicke (1990) Measu -
ing polya chy. S udies in Compa a i e In e na ional De el-
opmen 25(1): 51–72.
C uz, A u o San a (1998) Mexico’s changing elec o al o de ,
1977-1996: I s meaning o s a e-socie y ela ions. Jou nal
o In e na ional and A ea S udies 5(1): 95–115.
Dahl, Robe A (1971) Polya chy: Pa icipa ion and Opposi ion.
New Ha en, CT: Yale Uni e si y P ess.
Diamond, La y Jay; Juan Jose
´Linz & Seymou Ma in
Lipse , e al. (1989) Democ acy in De eloping Coun ies:
La in Ame ica ( olume 4). Boulde , CO: Lynne Rienne .
D esden, Jenni e Raymond & Ma c Mo je
´Howa d (2016)
Au ho i a ian backsliding and he concen a ion o poli ical
powe . Democ a iza ion 23(7): 1122–1143.
Edgell, Amanda B; Ma hew C Wilson, Vanessa A Boese &
Sand a G ahn (2020) Democ a ic legacies: Using demo-
c a ic s ock o assess no ms, g ow h, and egime ajec o ies.
Wo king pape . V-Dem Wo king Pape Se ies, No 100
(h ps:// -dem.ne /wp.h ml).
Edgell, Amanda B; Se aphine F Mae z, Lau a Maxwell,
Richa d Mo gan, Ju aj Medziho sky, Ma hew C Wilson,
Vanessa A Boese, Sebas ian Hellmeie , Jean Lachapelle,
Pa ik Linden o s, Anna Lu
¨h mann & S a an I Lindbe g
(2020) Episodes o egime ans o ma ion da ase and
codebook, V2.2 (h ps://gi hub.com/ demins i u e/ERT).
Eps ein, Da id L; Robe Ba es, Jack Golds one, Ida K is ensen
& Sha yn O’Hallo an (2006) Democ a ic ansi ions. Ame -
ican Jou nal o Poli ical Science 50(3): 551–569.
Esen, Be k & Sebnem Gumuscu (2016) Rising compe i i e
au ho i a ianism in Tu key. Thi d Wo ld Qua e ly 37(9):
1581–1606.
Fa zin, Y Hossein & C aig A Bond (2006) Democ acy and
en i onmen al quali y. Jou nal o De elopmen Economics
81(1): 213–235.
Fjelde, Hanne; Ca l Hen ik Knu sen & Hå a d Moklei
Nygå d (2021) Which ins i u ions ma e ? Re-conside ing
he democ a ic ci il peace. In e na ional S udies Qua e ly
65(1): 223–237.
Ga es, Sco ; Hå a d Heg e, Ma k P Jones & Hå a d S and
(2006) Ins i u ional inconsis ency and poli ical ins abili y:
Poli y du a ion, 1800–2000. Ame ican Jou nal o Poli ical
Science 50(4): 893–908.
Ge ing, John; S om C Thacke & Rod igo Al a o (2012)
Democ acy and human de elopmen . Jou nal o Poli ics
74(1): 1–17.
Goe z, Ga y (2020) Social Science Concep s and Measu emen :
New and Comple ely Re ised Edi ion. P ince on, NJ: P in-
ce on Uni e si y P ess.
Gu ses, Mehme (2011) Eli es, oil, and democ a iza ion: A su -
i al analysis. Social Science Qua e ly 92(1): 164–184.
Hagga d, S ephan & Robe Kau man (2021) Backsliding:
Democ a ic Reg ess in he Con empo a y Wo ld. Camb idge:
Camb idge Uni e si y P ess.
Hagga d, S ephan & Robe R Kau man (2012) Inequali y
and egime change: Democ a ic ansi ions and he s abili y
16 jou nal o PEACE RESEARCH XX(X)
Mae z e al. 983
Ansell, Ben & Johannes Lind all (2013) The poli ical o igins
o p ima y educa ion sys ems: Ideology, ins i u ions, and
in e denomina ional con lic in an e a o na ion-building.
Ame ican Poli ical Science Re iew 107(3): 505–522.
Bashi o , Galib & Ca oline Lancas e (2018) End o mode a-
ion: The adicaliza ion o AKP in Tu key. Democ a iza ion
25(7): 1210–1230.
Bjø nsko , Ch is ian & Ma in Rode (2020) Regime ypes
and egime change: A new da ase on democ acy, coups,
and poli ical ins i u ions. Re iew o In e na ional O ganiza-
ions 15(2): 531–551.
Boix, Ca les (2003) Democ acy and Redis ibu ion. New Yo k:
Camb idge Uni e si y P ess.
Boix, Ca les & Susan C S okes (2003) Endogenous democ a-
iza ion. Wo ld Poli ics 55(4): 517–549.
Boix, Ca les; Michael K Mille & Sebas ian Rosa o (2013) A
comple e da a se o poli ical egimes, 1800-2007. Com-
pa a i e Poli ical S udies 46(12): 1523–1554.
Bollyky, Thomas J; Ta a Templin, Ma hew Cohen, Diana
Schode , Joseph L Dieleman & Simon Wigley (2019) The
ela ionships be ween democ a ic expe ience, adul heal h,
and cause-speci ic mo ali y in 170 coun ies be ween 1980
and 2016: An obse a ional analysis. Lance 393(393):
1628–1640.
B ownlee, Jason (2009) Po en s o plu alism: How hyb id
egimes a ec democ a ic ansi ions. Ame ican Jou nal o
Poli ical Science 53(3): 515–532.
Cassani, And ea & Luca Tomini (2020) Re e sing egimes
and concep s: F om democ a iza ion o au oc a iza ion.
Eu opean Poli ical Science 19(2): 272–287.
Cheibub, Jose An onio; Jenni e Gandhi & James Raymond
V eeland (2010) Democ acy and dic a o ship e isi ed.
Public Choice 143: 67–101.
Chen, Linda (2007) A gen ina in he wen y- i s cen u y. In:
Howa d J Wia da & Ha ey F Kline (eds.) La in Ame ican
Poli ics and De elopmen . Boulde , CO: Wes iew:
103–122.
Claassen, Ch is ophe (2020) Does public suppo help
democ acy su i e? Ame ican Jou nal o Poli ical Science
64(1): 118–134.
Coppedge, Michael; Amanda B Edgell, Ca l Hen ik Knu sen
& S a an I Lindbe g (2022) Why Democ acies De elop and
Decline. New Yo k: Camb idge Uni e si y P ess.
Coppedge, Michael; John Ge ing, Ca l Hen ik Knu sen, S a -
an I Lindbe g, Jan Teo ell, Da id Al man, Michael Be n-
ha d, M. S e en Fish, Adam Glynn, Allen Hicken, Anna
Lu
¨h mann, Kyle L Ma qua d , Kelly McMann, Pamela
Pax on, Daniel Pems ein, B igi e Seim, Rachel Sigman,
S end-E ik Skaaning, Je ey S a on, S e en Wilson, Agnes
Co nell, Nazi a Alizada, Lisa Gas aldi, Haakon Gje løw,
Ga y Hindle, Nina Ilchenko, Lau a Maxwell, Vale iya
Mechko a, Ju aj Medziho sky, Johannes on Ro
¨me , Aksel
Sunds o
¨m, Ei an Tzelgo , Yi- ing Wang, To e Wig &
Daniel Zibla (2020) V-Dem [coun y–yea /coun y–da e]
da ase 10 (h ps://doi.o g/10.23696/ demds20).
Coppedge, Michael & Wol gang H Reinicke (1990) Measu -
ing polya chy. S udies in Compa a i e In e na ional De el-
opmen 25(1): 51–72.
C uz, A u o San a (1998) Mexico’s changing elec o al o de ,
1977-1996: I s meaning o s a e-socie y ela ions. Jou nal
o In e na ional and A ea S udies 5(1): 95–115.
Dahl, Robe A (1971) Polya chy: Pa icipa ion and Opposi ion.
New Ha en, CT: Yale Uni e si y P ess.
Diamond, La y Jay; Juan Jose
´Linz & Seymou Ma in
Lipse , e al. (1989) Democ acy in De eloping Coun ies:
La in Ame ica ( olume 4). Boulde , CO: Lynne Rienne .
D esden, Jenni e Raymond & Ma c Mo je
´Howa d (2016)
Au ho i a ian backsliding and he concen a ion o poli ical
powe . Democ a iza ion 23(7): 1122–1143.
Edgell, Amanda B; Ma hew C Wilson, Vanessa A Boese &
Sand a G ahn (2020) Democ a ic legacies: Using demo-
c a ic s ock o assess no ms, g ow h, and egime ajec o ies.
Wo king pape . V-Dem Wo king Pape Se ies, No 100
(h ps:// -dem.ne /wp.h ml).
Edgell, Amanda B; Se aphine F Mae z, Lau a Maxwell,
Richa d Mo gan, Ju aj Medziho sky, Ma hew C Wilson,
Vanessa A Boese, Sebas ian Hellmeie , Jean Lachapelle,
Pa ik Linden o s, Anna Lu
¨h mann & S a an I Lindbe g
(2020) Episodes o egime ans o ma ion da ase and
codebook, V2.2 (h ps://gi hub.com/ demins i u e/ERT).
Eps ein, Da id L; Robe Ba es, Jack Golds one, Ida K is ensen
& Sha yn O’Hallo an (2006) Democ a ic ansi ions. Ame -
ican Jou nal o Poli ical Science 50(3): 551–569.
Esen, Be k & Sebnem Gumuscu (2016) Rising compe i i e
au ho i a ianism in Tu key. Thi d Wo ld Qua e ly 37(9):
1581–1606.
Fa zin, Y Hossein & C aig A Bond (2006) Democ acy and
en i onmen al quali y. Jou nal o De elopmen Economics
81(1): 213–235.
Fjelde, Hanne; Ca l Hen ik Knu sen & Hå a d Moklei
Nygå d (2021) Which ins i u ions ma e ? Re-conside ing
he democ a ic ci il peace. In e na ional S udies Qua e ly
65(1): 223–237.
Ga es, Sco ; Hå a d Heg e, Ma k P Jones & Hå a d S and
(2006) Ins i u ional inconsis ency and poli ical ins abili y:
Poli y du a ion, 1800–2000. Ame ican Jou nal o Poli ical
Science 50(4): 893–908.
Ge ing, John; S om C Thacke & Rod igo Al a o (2012)
Democ acy and human de elopmen . Jou nal o Poli ics
74(1): 1–17.
Goe z, Ga y (2020) Social Science Concep s and Measu emen :
New and Comple ely Re ised Edi ion. P ince on, NJ: P in-
ce on Uni e si y P ess.
Gu ses, Mehme (2011) Eli es, oil, and democ a iza ion: A su -
i al analysis. Social Science Qua e ly 92(1): 164–184.
Hagga d, S ephan & Robe Kau man (2021) Backsliding:
Democ a ic Reg ess in he Con empo a y Wo ld. Camb idge:
Camb idge Uni e si y P ess.
Hagga d, S ephan & Robe R Kau man (2012) Inequali y
and egime change: Democ a ic ansi ions and he s abili y
16 jou nal o PEACE RESEARCH XX(X)
o democ a ic ule. Ame ican Poli ical Science Re iew
106(3): 495–516.
Heg e, Hå a d; Tanja Ellingsen, Sco Ga es & Nils Pe e
Gledi sch (2001) Towa d a democ a ic ci il peace? Democ-
acy, poli ical change, and ci il wa , 1816–1992. Ame ican
Poli ical Science Re iew 95(1): 33–48.
Hun ing on, Samuel P (1993) The Thi d Wa e: Democ a iza-
ion in he La e Twen ie h Cen u y. No man, OK: Uni e -
si y o Oklahoma P ess.
Jackman, Robe W & Kenne h A Bollen (1989) Democ acy,
s abili y and dicho omies. Ame ican Sociological Re iew
54(4): 612–621.
Ka l, Te y Lynn (1987) Pe oleum and poli ical pac s: The
ansi ion o democ acy in Venezuela. La in Ame ican
Resea ch Re iew 22(1): 63–94.
Knu sen, Ca l Hen ik & Hå a d Moklei Nygå d (2015)
Ins i u ional cha ac e is ics and egime su i al: Why a e
semi-democ acies less du able han au oc acies and
democ acies? Ame ican Jou nal o Poli ical Science 59(3):
656–670.
Kubicek, Paul (2011) Poli ical condi ionali y and Eu opean
Union’s cul i a ion o democ acy in Tu key. Democ a iza-
ion 18(4): 910–931.
Le i z, Philip & G igo e Pop-Eleches (2010) Why no back-
sliding? The Eu opean Union’s impac on democ acy and
go e nance be o e and a e accession. Compa a i e Poli ical
S udies 43(4): 457–485.
Lewis, S ephen E (2020) Mexico’s democ a ic ansi ion and
consolida ion since 1977. In: William Beezley (ed.) The
Ox o d Handbook o Mexican His o y. New Yo k: Ox o d
Uni e si y P ess. (h ps://doi.o g/10.1093/ox o dhb/
9780190699192.013.35).
Linz, Juan J (1975) To ali a ian and au ho i a ian egimes.
In: F ed I G eens ein & Nelson W Polsby (eds) Handbook
o Poli ical Science. Reading, MA: Addison-Wesley:
175–411
Linz, Juan J & Al ed S epan (1996) P oblems o Democ a ic
T ansi ion and Consolida ion: Sou he n Eu ope, Sou h Ame -
ica, and Pos -communis Eu ope. Bal imo e, MD: Johns
Hopkins Uni e si y P ess.
Linz, Juan Juan Jose
´(1978) The B eakdown o Democ a ic
Regimes. Bal imo e, MD: Johns Hopkins Uni e si y P ess.
Lu
¨h mann, Anna & S a an I Lindbe g (2019) A hi d wa e o
au oc a iza ion is he e: Wha is new abou i ? Democ a iza-
ion 26(7): 1095–1113.
Lu
¨h mann, Anna; Ma cus Tannenbe g & S a an I Lindbe g
(2018) Regimes o he Wo ld (RoW): Opening new a e-
nues o he compa a i e s udy o poli ical egimes. Poli ics
& Go e nance 6(1): 60–77.
Mainwa ing, Sco & Fe nando Bizza o (2019) The a es
o hi d-wa e democ acies. Jou nal o Democ acy 30(1):
99–113.
Ma shall, Mon y G; Ted Robe Gu & Kei h Jagge s (2019)
Poli y IV P ojec : Poli ical Regime Cha ac e is ics and
T ansi ions, 1800-2018. Da ase Use s’ Manual (h ps://
www.sys emicpeace.o g/insc /p4manual 2018.pd ).
Mille , Michael K (2015) Democ a ic pieces: Au oc a ic elec-
ions and democ a ic de elopmen since 1815. B i ish
Jou nal o Poli ical Science 45(3): 501–530.
Mousseau, Deme Yalcin (2001) Democ a izing wi h e hnic
di isions: a sou ce o con lic ? Jou nal o Peace Resea ch
38(5): 547–567.
O’Donnell, Guille mo & Philippe C Schmi e (1986) T an-
si ions om Au ho i a ian Rule: Ten a i e Conclusions abou
Unce ain Democ acies. Bal imo e, MD: John Hopkins
Uni e si y P ess.
Papaioannou, Elias & G ego ios Siou ounis (2008) Demo-
c a isa ion and g ow h. Economic Jou nal 118(532):
1520–1551.
Powell, Jona han M & Clay on L Thyne (2011) Global
ins ances o coups om 1950 o 2010: A new da ase .
Jou nal o Peace Resea ch 48(2): 249–259.
P zewo ski, A; S Newman, S K Pa k, D Que al , G Ri e o &
K J Shin (2013) Poli ical Ins i u ions and Poli ical E en s
(PIPE) Da a Se , (h ps://si es.google.com/a/nyu.edu/
adam-p zewo ski/home/da a).
P zewo ski, Adam; Michael E Al a ez, Jose An onio Cheibub
& Fe nando Limongi (2000) Democ acy and De elopmen :
Poli ical Ins i u ions and Well-being in he Wo ld,
1950-1990. New Yo k: Camb idge Uni e si y P ess.
Rus ow, Dankwa A (1970) T ansi ions o Democ acy:Towa d
a Dynamic Model. Compa a i e Poli ics 2(3): 337–363.
Sa o i, Gio anni (1970) Concep mis o ma ion in compa a i e
poli ics. Ame ican Poli ical Science Re iew 64(4): 1033–1053.
Schedle , And eas (2001) Taking unce ain y se iously: The
blu ed bounda ies o democ a ic ansi ion and consolida-
ion. Democ a iza ion 8(4): 1–22.
Schedle , And eas (2013) The Poli ics o Unce ain y: Sus ain-
ing and Sub e ing Elec o al Au ho i a ianism. Ox o d:
Ox o d Uni e si y P ess.
Some , Mu a (2017) Conque ing e sus democ a izing
he s a e: Poli ical Islamis s and ou h wa e democ a-
iza ion in Tu key and Tunisia. Democ a iza ion 24(6):
1025–1043.
Sundbe g, Ralph & E ik Melande (2013) In oducing he
UCDP Geo e e enced E en Da ase . Jou nal o Peace
Resea ch 50(4): 523–532.
Sunds o
¨m, Aksel; Pamela Pax on, ing Wang Yi & S a an I
Lindbe g (2017) Women’s poli ical empowe men : A new
global index, 1900–2012. Wo ld De elopmen 94(June):
321–335.
S olik, Milan W (2012) The Poli ics o Au ho i a ian Rule.
New Yo k: Camb idge Uni e si y P ess.
Teo ell, Jan (2010) De e minan s o Democ a iza ion: Explain-
ing Regime Change in he Wo ld, 1972–2006. New Yo k:
Camb idge Uni e si y P ess.
Tilly, Cha les (2001) Mechanisms in poli ical p ocesses.
Annual Re iew o Poli ical Science 4(1): 21–41.
Mae z e al. 17
984 jou nal o Peace Resea ch 61(6)
T eisman, Daniel (2020) Democ acy by mis ake: How he
e o s o au oc a s igge ansi ions o ee go e nmen .
Ame ican Poli ical Science Re iew 114: 792–810.
Wagne , Cli o d H (1982) Simpson’s pa adox in eal li e.
The Ame ican S a is ician 36(1): 46–48.
Wa d, Michael D & K is ian S Gledi sch (1998) Democ a izing
o peace. Ame ican Poli ical Science Re iew 92(1): 51–61.
Wilson, Ma hew C; Ju aj Medziho sky, Se aphine F Mae z,
Pa ik Linden o s, Amanda B Edgell, Vanessa Boese &
S a an I Lindbe g ( o hcoming) Episodes o libe aliza ion
in au oc acies: A new app oach o quan i a i ely s udying
democ a iza ion. Poli ical Science Resea ch and Me hods
(h ps://doi.o g/10.1017/ps m.2022.11).
Wynia, Ga y W (1990) The Poli ics o La in Ame ican De el-
opmen . New Yo k: Camb idge Uni e si y P ess.
Zibla , Daniel (2006) How did Eu ope Democ a ize. Wo ld
Poli ics 58(2): 311–338.
SERAPHINE F MAERZ, b. 1982, PhD in Poli ical Science
(CEU Budapes , 2017); Resea ch Fellow a Goe he Uni e si y
F ank u and PI o DFG p ojec ‘Sma Au ho i a ianism?’
(2021– ); esea ch in e es s: au ho i a ianism, poli ical
communica ion, In e ne poli ics. Mos ecen a icles
published in Poli ical Science Resea ch and Me hods (PSRM),
B i ish Jou nal o Poli ical Science, and Democ a iza ion.
AMANDA B EDGELL, b. 1986, PhD in Poli ical Science
(Uni e si y o Flo ida, 2019); Assis an P o esso , Uni e si y
o Alabama (2020– ); Resea ch Associa e a he Va ie ies o
Democ acy (V-Dem) Ins i u e (2019– ); cu en main
in e es s: au ho i a ianism, democ a ic esilience, A ican
poli ics. Mos ecen book: Why Democ acies De elop and
Decline (Camb idge, 2022).
MATTHEW C WILSON, b. 1985, PhD in Poli ical
Science (Pennsyl ania S a e Uni e si y, 2015); Associa e
P o esso , Uni e si y o Sou h Ca olina (2019– ); Resea ch
Associa e a he Va ie ies o Democ acy (V-Dem) Ins i u e
(2018– ); cu en main in e es s: au ho i a ianism, egime
change, domes ic con lic . Mos ecen book: One oad o
iches? How s a e building and democ a iza ion a ec economic
de elopmen (Camb idge Elemen , 2021).
SEBASTIAN HELLMEIER, b. 1989, PhD in Poli ical
Science (Uni e si y o Kons anz, 2019); Pos doc o al
Resea ch Fellow, WZB Be lin Social Science Cen e
(2021– ); Resea ch Associa e a he Va ie ies o Democ acy
(V-Dem) Ins i u e (2021– ); main in e es s: au ho i a ian
egimes, con en ious poli ics, democ a iza ion.
STAFFAN I LINDBERG, b. 1969, PhD in Poli ical Science
(Lund Uni e si y, 2005); P o esso , Uni e si y o
Go henbu g (2013– ); Founding Di ec o , V-Dem Ins i u e
(2014– ); cu en main in e es s: democ acy ends,
au oc a iza ion, clima e u u e scena ios. Mos ecen book:
Why Democ acies De elop and Decline (Camb idge, 2022).
18 jou nal o PEACE RESEARCH XX(X)
