economic security and democratic capital: why do some democracies survive and others fail?

Es

Ta

b

c

a

ARRA

KDEED

JDPP

1

tmsnfudiscaf

s

1T

2h

Journal of Behavioral and Experimental Economics 50 (2014) 13–28

Contents lists available at ScienceDirect

Journal of Behavioral and Experimental Economics

j ourna l h o mepage: www.elsev ier .com/ locate / jbee

conomic security and democratic capital: Why do some democraciesurvive and others fail?�

homas D. Jeitschkoa, Susan J. Linza, Jose Noguerab, Anastasia Semykinac,∗

Michigan State University, United StatesUniversity of Santiago de Chile, ChileFlorida State University, United States

r t i c l e i n f o

rticle history:eceived 16 June 2013eceived in revised form 19 January 2014ccepted 20 January 2014

eywords:emocracy breakdownxpectations

a b s t r a c t

We develop a theoretical framework where the chance of any given democratic society maintaining itsdemocratic status is determined by two key factors: economic security and democratic capital. Our modelpredicts democracy breakdown is more likely (1) the lower the level of democratic capital, (2) the lowerthe anticipated growth in democracy, (3) the greater the anticipated growth after democracy break-down, and finally, (4) the smaller the difference between anticipated growth in continued democracyand after democracy breakdown. We test the model using a newly constructed data set and the Polity IVdata. We find that if expected economic growth under democracy is greater than anticipated economic

conomic securityemocratic capital

EL classification:721648

growth under the alternative regime then the probability of breakdown is lower. Moreover, an increasein democratic capital decreases the probability of democracy breakdown. The country’s most recent owndemocratic experience appears to have a more important impact on democracy survival, while the effectof foreign democratic capital cannot be distinguished from the time-specific shocks that are common toall countries in the world.

© 2014 Elsevier Inc. All rights reserved.

ltime

ptbtlc

. Introduction

Even before the extensive economic and political changes ofhe 1990s and recent rebellion movements in the Middle East,

any questioned the factors which make a democratic regimeustainable. Ongoing debates surround the issue of how eco-omic performance is related to democratization. Some papers

ocus on the ‘modernization’ hypothesis, where greater levels ofrbanization and wealth are associated with a greater degree ofemocratization, and where higher incomes are expected, if not to

nduce democracy, then at least to make democratic regimes moretable (Barro, 1999; Lipset, 1959; Przeworski et al., 1996, 2000). In
ontrast, Acemoglu et al. (2008) argue that the observed positivessociation between income and democracy stems from commonactors that determine both variables. Persson and Tabellini (2009)
� We thank seminar participants at Warwick University, Royal Holloway, Univer-ity of London, and University of New Hampshire for helpful comments.∗ Corresponding author at: Department of Economics, Florida State University,13 Collegiate Loop, Tallahassee, FL 32306-2180, Unites States.el.: +1 850 644 4557; fax: +1 850 644 4535.

E-mail address: [email protected] (A. Semykina).

dwr

igslatn

214-8043/$ – see front matter © 2014 Elsevier Inc. All rights reserved.ttp://dx.doi.org/10.1016/j.socec.2014.01.005

ink economic performance and democratization by suggestinghat democracy may entail higher returns to investment, whichn turn will determine expectations of future economic perfor-

ance under democracy and thus, beneficially affect both currentconomic performance and democracy consolidation.

Despite the wealth of literature linking democracy to economicerformance, no studies examine the impact of relative (as opposedo absolute) economic performance on the likelihood of democracyreakdown. We address this literature gap. Specifically, we arguehat even though some societies may generally be less wealthy andess inclined to support democracy due to historical reasons, theirhances for democracy consolidation are nevertheless better if aemocratic regime is expected to produce higher levels of wealthhen compared to the possible alternative of a non-democratic

egime.We begin by developing a theoretical framework that formal-

zes a trade-off between economic security and ability to affectoverning institutions; a trade-off that is present in all democraticocieties. We then formulate testable hypotheses about how the
ikelihood of democracy breakdown is related to the degree ofttachment to democratic values and principles, and about expec-ations of future economic performance under democratic andon-democratic regimes. Our theoretical framework implies that
dx.doi.org/10.1016/j.socec.2014.01.005

http://www.sciencedirect.com/science/journal/00905720

http://www.elsevier.com/locate/jbee

http://crossmark.crossref.org/dialog/?doi=10.1016/j.socec.2014.01.005&domain=pdf

mailto:[email protected]

dx.doi.org/10.1016/j.socec.2014.01.005

1 and E

tdtg

cc2cPsctdt(iiiWtb

drfcioui

tbeigtWr

2

di22tdrPdtrem

rda2t

wlddtroed

lioomaTtaiCpbus

dridMlan

3

wtdo

deetw

attm

4 T.D. Jeitschko et al. / Journal of Behavioral

he risk of democracy breakdown declines not only when theegree of attachment to democratic values increases, but also whenhe difference between actual growth in democracy and expectedrowth after democracy breakdown becomes larger.

To test the implications of our theoretical framework, weonstruct a new data set that covers all currently independentountries in all years during the period 1800–2006 (see Noguera,007a for details). As a robustness check, we also test our theoreti-al predictions using the well-known but more limited in coverageolity IV data.1 Employing traditional duration (survival) analy-is, we find that general patterns observed in both data sets areonsistent with our predictions. Anticipated growth difference hashe expected negative effect – if expected economic growth underemocracy (proxied by the country’s past GDP growth) is greaterhan anticipated economic growth under the alternative regimeproxied by past GDP growth in ‘peer’ countries), then the probabil-ty of democracy breakdown, conditional on no breakdown so far,s lower. Moreover, an increase in accumulated democratic capitals negatively related to the probability of democracy breakdown.

e find that democracy breakdown is more likely to occur duringhe first few years of a democracy episode, with the likelihood ofreakdown declining over time.

While it may be tempting to also consider the switch from non-emocracy to democracy, we refrain from doing so. Modeling theeversion from non-democracy would necessitate developing a dif-erent model, especially if taking into account the wide variety ofultural, institutional, economic and historical factors characteriz-ng non-democratic regimes. Moreover, from the empirical pointf view, information on the start of non-democracy episodes isnavailable in many instances (especially early ones), which makes

t difficult to properly account for duration dependence.The paper is organized as follows. Section 2 reviews the previous

heoretical and empirical literature on democracy consolidation-reakdown. Section 3 develops a theoretical framework thatmphasizes the role of preferences for and trade-offs between anndividual’s economic security and ability to exercise control overoverning institutions. Section 4 describes the data and estima-ion strategy used to test the hypotheses generated by the model.

e discuss the empirical results in Section 5, and offer concludingemarks in Section 6.

. Previous literature

A substantial empirical literature investigates the link betweenemocratic consolidation and economic well-being, typically find-

ng this link to be positive (Bernhard, Nordstrom, and Reenock,001; Bernhard, Reenock, and Nordstrom, 2003; Epstein et al.,006; Przeworski and Limongi, 1993, 1997). While proponents ofhe modernization hypothesis suggest a causal effect of income onemocratization, others cast doubt on the causality in these cor-elations (Acemoglu et al., 2008; DeHaan and Siermann, 1995).ersson and Tabellini (2009) propose an alternative explanation:emocracy may have an indirect effect on current performance
hrough the expected returns to investment under a democraticegime in the future. In societies where returns to investment arexpected to be higher under democracy, greater levels of invest-ent in both economic development and democratization occur,
1 Numerous papers focusing on democracy questions express concerns abouteadily-available data and corresponding democracy measures, and consequentlyevelop new data sets and democracy measures (see, for example, Cheibub, Gandhi,nd Vreeland, 2010; DeHaan and Siermann, 1995; Gasiorowski, 1993; Potrafke,012; Treisman, 2010). Few consider multiple measures and use multiple data setso check for robustness.

ettuttdrCrt

xperimental Economics 50 (2014) 13–28

hich lead to better current economic performance and greaterikelihood of democracy survival. They argue that because of theescribed incentives, countries sort themselves into autocratic andemocratic regimes. Based on the comparative statics derived fromheir model, Persson and Tabellini suggest that exit from democ-acy should be negatively related to income, while the probabilityf exit from autocracy should not be responsive to income. Theirmpirical results are consistent with the comparative statics pre-ictions.

In other strands of the democracy consolidation-breakdowniterature, the survival or failure of democratic regimes is empir-cally linked to country history, country affiliations, and regionalr global trends. Country history typically is measured in termsf the country’s previous experience with democracy and com-itment to democratic values (see, for example, Diskin, Diskin,

nd Hazan, 2005; Peffley and Rohrschneider, 2003; Sullivan andransue, 1999). Country affiliations are measured in terms of par-icipation in networks and international organizations (Mansfieldnd Pevehouse, 2006; Pevehouse, 2002). Regional and global polit-cal trends measure the relative share of democracies (Brinks andoppedge, 2006; Bunce, 2001). Persson and Tabellini (2009) incor-orate a country’s previous democratic experiences in their modely introducing the concept of ‘democratic capital,’ which they alsose in their empirical analysis, finding that the chance of democracyurvival is positively linked to democratic capital.

In our theoretical framework ‘democratic capital’ impacts theecision to maintain a democratic regime not only because democ-atization may be associated with higher income, but also becausendividuals learn to appreciate democratic values and thus mayerive direct benefits from living in a democratic environment.oreover, we extend the existing literature by empirically ana-

yzing the link between the likelihood of democracy breakdownnd the difference in economic performance under democratic andon-democratic regimes.

. Theoretical framework

In developing the framework to formalize our investigatione first consider a static setting in which the main intuition for

he trade-offs that people face is established. We then present aynamic extension that generates the propositions investigated inur empirical analysis.

Since our focus is the incidence of democracy survival andemocracy breakdown, we begin by considering a democratic soci-ty. Members in a democratic society care not only about theirconomic security, which is influenced by macroeconomic condi-ions, but also about their ability to implement governance choices,hich is influenced by the strength of democratic institutions.

In a democratic society, the fact that both economic security andbility to influence the governing institutions are important implieshe existence of a willingness to engage in trade-offs between thewo. Agents are willing to forsake their right to control decision-

akers, provided that they achieve a sufficiently large increase inconomic security. Since economic security is the more fundamen-al need, the trade-off between economic security and the abilityo influence the governing institutions is not constant, but dependspon the existing level of wealth and the degree of commitmento democratic values in the society. Thus, the greater one’s wealth,he greater the marginal rate of substitutions between wealth andemocratic values, and the more willing one is to sacrifice material
esources for the attainment of higher levels of democratization.onversely, poorer individuals are less able to sacrifice materialesources, and therefore have a decreased marginal rate of substitu-ion, making them more willing to give up their ability to influence

and E

ie

3a

ormvwtetipp

igawa‘

u

fon

dgutdcaae

dftiddtS

F

mocGocatut

wipd

3

ddroartbddeoww

u

ddobacdme

ti

rrdnaetdf

T.D. Jeitschko et al. / Journal of Behavioral

nstitutions in exchange for a given incremental increase in theirconomic security.

.1. Economic security and freedom in democracy: the pivotalgent

To analyze democracy survival or breakdown we use the notionf a pivotal agent because choices (trade-offs) resulting in democ-acy breakdown may not reflect conditions associated with aedian voter framework. This is especially true in the case of a

iolent takeover – the pivotal agent does not necessarily coincideith a somehow-determined median voter.2 Indeed, if agents con-

emplate post-breakdown redistributive policies, it is clear that ifnough agents stand to gain sufficiently as a result of the switch,hen a revolution, putsch, coup d’etat or similar event may bemplemented with less than a majority of the population in sup-ort. Even traditional military coups need some support from theopulation in order to consolidate their power (Noguera, 2007b).

In a democracy, the pivotal agent cares about two character-stics: the overall wealth in society and the ability to influenceoverning institutions. Specifically, preferences are represented by

felicity (per period utility) function, in which the arguments areealth (Wt) and the degree of attachment to democratic values

nd principles that, following Persson and Tabellini (2009), we calldemocratic capital’ (Ft):

(Wt, Ft) = Wt(1 + Ft)˛, ̨ > 1. (1)

Wealth is defined in the conventional manner and requires nourther explanation, other than to note that it can also be thoughtf in terms of a Cobb–Douglass composite good: W = Cˇ

1 C1−ˇ2 . The

otion of democratic capital, however, requires some discussion.In a democracy, the ability to influence governing institutions

epends not only upon the current state of political rights formallyranted, but also upon attitudes in society toward democratic val-es. Benefits of democracy may be fully enjoyed only in societieshat have consistently placed a high value on the principles ofemocracy. Because this societal trait – an appreciation of demo-ratic values – requires investment over time to achieve, the pivotalgent’s utility depends on the overall level of democratic capitalccumulated in the society, which captures the country’s previousxperiences with democracy.

In order to formalize these ideas, let the current degree ofemocratization in a society be captured by ft ≥ 0. Specifically,

t measures the ability of individuals at time t to participate inhe decision-making process by exercising control over governingnstitutions. We let higher levels of f indicate a greater degree ofemocratization. Over time, in societies that place a high value onemocratic principles and commit to higher levels of democratiza-ion, there is a cumulative effect; what we call democratic capital.pecifically, we assume that

t =∞∑

�=0

��ft−� = ft + �Ft−1, 0 < � < 1 (2)

2 Should the pivotal agent coincide with the median voter, democracy breakdownay come about in several ways. First, and most obviously, it may occur in the form

f a self-coup; that is, the majority may elect a government that abandons a demo-ratic system – an instance for which there is ample historical precedence (Hitler inermany in 1933, Fujimori in Peru in 1992). Second, a more or less forceful takeoverf government may occur by a group that eo ipso is not part of the democratic pro-ess, but still has popular support. This may be the case with a military coup, butlso with a guerilla force that, due to wide popular support, experiences no morehan token resistance to a government takeover. Benhabib and Przeworski (2006)se the notion of a pivotal agent to capture the sentiment within large groups – inheir case, groups are separated by income.

dbonep

ctbIadd

xperimental Economics 50 (2014) 13–28 15

here Ft is the democratic capital in the society at time t, �Ft−1s the value of democratic capital in society carried over from theast, and ft reflects the current commitment to and investment inemocratization.

.2. Democracy breakdown and expectations

Having characterized a willingness to engage in trade-offs inemocracy, we now consider democracy breakdown. By definition,emocracy breakdown entails a reduction in the degree of democ-atization. Indeed, democracy breakdown implies a deteriorationf the institutions that are necessary to support individual rightsnd abilities to exercise control over the decision-making process,egardless of the amount of democratic capital accumulated inhe society. While such deterioration need not be characterizedy an instantaneous break, we assume that if democracy break-own occurs, the pivotal agent’s utility becomes independent of theemocratic capital, as the benefits associated with it are no longerxperienced. For expositional ease, when democracy breakdownccurs, utility is derived solely from economic security. Therefore,e use an indicator function 1D to capture democracy survival andrite the instantaneous utility function as

(Wt, Ft) = Wt(1 + 1DFt)˛, with 1D =

{1 in democracy,

0 in breakdown.(3)

From this specification it is clear that, for any given value ofemocratic capital, an agent is willing to surrender benefits ofemocratization, provided that a sufficiently high level of wealth isbtained in compensation. Thus, for a society to support democracyreakdown and the corresponding loss of institutions that protectnd promote democratic principles, it must be the case that a suffi-iently large number of people are willing to accept the reduction inemocratization in exchange for greater economic security. Such aove is captured by the pivotal agent, in anticipation of perceived

conomic outcomes under democracy and democracy breakdown.For the static setting – which captures the essence of the

rade-offs in our argument – multiple scenarios for breakdown arellustrated in Fig. 1.

Fig. 1 depicts possible constellations of expectations foremaining in democracy. For example, while remaining in democ-acy, moving from an initial point associated with the level ofemocratic capital of Ft to point Dg represents expectations of eco-omic growth; a move to Dn denotes expectations of no change;nd a move to Dr reflects expectations of a (possibly continued)conomic recession; finally a move to point Dd denotes expecta-ions of an economic depression. In all of these cases, while wealthepends on economic growth, the survival of democracy implies aurther investment in democratic capital and, thus, increases in F.

All of these possible outcomes associated with remaining inemocracy imply different thresholds for triggering democracyreakdown – some of which may be rather low. Indeed, for the casef an expected (continued) economic depression in democracy, aon-democracy would be preferred at lower than the current lev-ls of wealth (depicted as the indifference of point Bd compared tooint Dd).

Taking a longer time horizon necessitates specifications con-erning expectations about the future path of a country under thewo possible current cases: continued democracy or democracyreakdown. To this end, we consider a three-period framework.

n the current period, t, a decision is formed about the immedi-te future, i.e., whether to support democracy and thus maintainemocracy in period t + 1, or whether to support democracy break-own. Should democracy survive in the immediate future, the

16 T.D. Jeitschko et al. / Journal of Behavioral and Experimental Economics 50 (2014) 13–28

nd de

qm

imaieoii

aeotnrPcamtrewb

tb

tpeiei1mtteo

dovrvuw

Gbbb

b

td

U

wac

w

1

Fig. 1. Expectations a

uestion of possible democracy breakdown is raised again in theore distant future, period t + 2.In contrast, once democracy breakdown occurs, the ability to

nfluence decision-makers, and thus the ability to agitate for andake changes are restricted so that the notion of the pivotal agent

s constituted above is no longer relevant, even if a transition backnto democracy were a preferred outcome. We capture this inher-nt asymmetry between democracy breakdown and the emergencef democracy by assuming that a reversal of democracy breakdowns not expected to take place should democracy breakdown occurn period t + 1.

In order to determine the optimal path, individuals living in democracy must form expectations about how they would fareconomically under non-democracy, accepting the immediate lossf welfare associated with the loss of ability to influence insti-utions. Individuals may expect that the governing forces underon-democracy will improve economic growth by redirecting, orestructuring, economic activity. Without needing to deal witharliaments or other bargaining scenarios common in democra-ies, individuals may perceive leaders in non-democracy as betterble to focus on attaining a higher growth rate.3 This perceptionay be especially true in the presence of severe economic down-

urns in established democracies, as well as in new democraticegimes when economic performance falls short of individuals’xpectations, provided that beliefs about economic performancehen leaving the democratic system remain the same as they were

efore.We suppose that agents’ beliefs are represented by distribu-

ions where the means under continued democracy and democracyreakdown are given by �d

t+1 and �nt+1.

3 In non-democratic regimes, especially in countries where manufacturing con-ributes less than one-quarter of total economic activity, it is likely that planners’references rather than consumer sovereignty will prevail, allowing for greatermphasis on producer goods, and enabling the country to experience a correspond-ngly higher growth rate than would occur in a democratic regime and market-basedconomy at the same level of economic development. In practice, historically speak-ng, this outcome occurred in the Soviet Union in the 1930s, and in China in the950s.The former example is particularly telling, as during the Great Depression,any countries were quite envious of the purported growth rates reported from

he Soviet Union, leading many to consider free-market democracies as a failed sys-em that lagged behind in terms of the performance of the Soviet style plannedconomy. More recently, some doubt has been cast on how greatly the Soviet Unionut-performed European and North American countries (Rosefielde, 2003).

w

P

((((

tp

mocracy breakdown.

In order to determine whether the optimal path involvesemocracy breakdown in period t + 1, beliefs about the sec-nd period must be assessed. As the felicity takes on differentalues in t + 2 with continued democracy survival vs. democ-acy breakdown, what matters to assessments is the expectedalue of the felicity rather than growth itself. Specifically, letting∗t+2:= max{ud

t+2, unt+2}, to evaluate the expectations of the future

hen democracy survives in t + 1, one must determine Etu∗t+2.

To determine the appropriate expectation of u∗t+2, let Gd

t+2 andnt+2 denote the distributions of ud

t+2 and unt+2 that are implied by

eliefs about future growth if democracy survives in t + 1. Then,y the Bapat–Beg Theorem (see Bapat and Beg, 1989), the distri-ution of u∗

t+2 if breakdown does not take place in t + 1 is given

y Gmax:=Gdt+2Gn

t+2.4 Hence, normalizing the level of democratiza-

ion experienced in democracy to ft = f̄ for all t, the implications ofemocracy survival and breakdown are given by,

t ={

(1 + �dt+1)Wt(1 + f̄ + �Ft)

˛ + ˇ

∫u∗

t+2dGmax

(1 + �nt+1)Wt + ˇEt(1 + �n

t+2)(1 + �nt+1)Wt

(6)

here once again the top branch obtains under democracy survivalnd the bottom branch under democracy breakdown in t + 1, and ˇaptures the rate at which the distant future is discounted at.

Consequently, democracy breakdown occurs in period t + 1henever

+ �dt+1 <

(1 + �nt+1)(1 + ̌ + ˇEt�n

t+2)

(1 + f̄ + �Ft)˛ + ˇ

∫v∗

t+2dGmax, (7)

here v∗t+2 = u∗

t+2/[(1 + �dt+1)Wt].

From this, then, follows readily,

roposition. Ceteris paribus, democracy survival is more likely

1) the greater the amount of democratic capital in society, Ft;2) the greater the growth in democracy, �d

t+1;
3) the smaller the growth absent democracy, �n
t+1; and4) the greater the difference between the growth in continued democ-

racy and the growth after democracy breakdown.

4 The Bapat-Beg Theorem does not always yield closed form solutions for dis-ributions, but Rarakat and Abdelkader (2004) give some examples where this isossible.

and E

sis

otiieulpcrterar

4

apaltt(oceb

ona(tdidra(

h2ppPaT

o

dh

np

mteoob

4

irteBstaAFnEwcip

s

D

w7rddptstHrttcodtlp


The proof follows immediately from Eq. (7), since increases in Ftrictly increase Gd

t+2 in the sense of first order stochastic dom-nance, which weakly increases Gmax in the sense of first ordertochastic dominance.

The first part of the proposition is familiar from the literaturen democracy consolidation. The second is similar to the litera-ure that suggests that democracy survival is a function of societalncome or wealth. However, it should be noted that we specif-cally focus on economic performance as measured by growthxpectations, rather than absolute levels of wealth. This allowss to capture instances in which poor democratic countries are

ess susceptible to democracy breakdown. The third part of theroposition is novel in that it emphasizes expectations about aounter-factual economic performance associated with democ-acy breakdown. The fourth part of the proposition, a corollaryo the previous two parts, is therefore also distinctive, as itmphasizes that expectations about performance under democ-acy must be considered relative to aspiration levels, i.e., beliefsbout possible alternative economic performance under democ-acy breakdown.

. Data

To facilitate empirical testing of our hypotheses, we constructed data set that covers 1800–2006 and includes all currently inde-endent countries; a total of 187 countries. Specifically, we identify

political regime prevailing in a given country in each particu-ar year based on the level of voter participation and extent ofhe competition in the political election process. Our classifica-ion approach is similar to Bernhard, Nordstrom, and Reenock2001), Przeworski et al. (2000), and others. The major advantagef our data is the broader coverage compared to data sets whichonsider countries only since their independence. Extended cov-rage permits better measures of democratic capital, as discussedelow.

Based on our classification, political regimes may fall into onef the following six categories, each associated with a particularumerical value: autocracy = 1, bureaucracy & monarchy = 2, semi-utocracy = 4, oligarchy = 5, semi-democracy = 6 and democracy = 7where “3” is missing to emphasize a significant difference betweenhe most autocratic and other regimes).5 These data have beeneveloped by Noguera (2007a) and have not been previously used

n any other research.6 We refer to this data set as ‘constructedata.’ In the constructed data, 140 countries had a democraticegime at least once during the considered period. In total, therere 215 episodes of democracy and 101 instances of breakdownsee Table 1a).

In addition, we also perform estimation using Polity IV data thatave previously been used by many authors (e.g. Acemoglu et al.,008; Persson and Tabellini, 2009) and cover all countries with theopulation of more than 500,000 people from the year of inde-

7
endence (or from 1800, whichever comes last) until 2004. In theolity IV data, we follow Persson and Tabellini (2009) and identify
country as a democracy if polity2 variable is greater than zero (seeable 1b). The Polity IV measure of democracy is broader than the

5 We also re-ran the regressions where the regime variable was recoded to taken values 1, 2, 3, 4, 5, 6. The results were practically unchanged.6 A brief summary of the regime classification can be found in Appendix A. The

etailed discussion of the regime identification strategy and complete list of regimeistories for each country are provided in Noguera (2007a).7 We also considered using the Freedom House Measure of Democracy. Unfortu-ately, it was not suitable for our purposes because it covers only a relatively recenteriod (1972–2007), which is too short to make our analysis feasible.

oesp

oo

c


easure obtained from our constructed data, which is in part dueo polity2 score being partially based on the executive powers andxecutive selections, while the constructed data focus exclusivelyn rights to vote and fair elections. Overall, there are 226 episodesf democracy in the Polity data, and 124 instances of democracyreakdown.

.1. Measures of democratic capital

As noted earlier, democratic capital grows slowly over time,nfluenced not only by conditions within each society, but also byegional and global conditions. Therefore, we decompose Ft intohree components: a country’s own political experiences, politicalxperiences in peer countries, and experiences of other countries.y peers we mean countries that are culturally and economicallyimilar and are characterized by relatively strong connections dueo historical reasons and/or geographic proximity. Specifically, wessign each country to one of twelve peer groups: Caribbean, Eastsia, Former socialist countries in Central and Eastern Europe,ormer Soviet Union, Islamic, Latin America, North America, Ocea-ia, South Asia, Southeast Asia, Sub-Sahara Africa and Westernurope (see Appendix B). In addition to estimating specificationshere foreign democratic capital is broken into the democratic

apital of peers and that other countries, we also consider spec-fications where all foreign countries are treated equally, i.e. theeer group is the whole world.

A country’s own democratic capital is defined as a discountedum of the country’s regime indices for the previous 50 years:

emocraticCapitalownit =

∑50�=1ı�f1,t−�

D, (8)

here fi,t−� is the regime index (ranging from 1 = autocracy to = democracy) when using the constructed data and the democ-acy status (1 if democracy, 0 otherwise) when using Polity IVata,8 in country i in period t − �; ı is a discount factor, and theenominator, D = ˙50

�=1ı� , is used for normalization purposes. Weerform estimation for ı = 0.94 and ı = 0.99, where each reflectshe fact that more recent experiences are most memorable inhaping beliefs and actions.9 Unfortunately, due to lack of iden-ification, it is not possible to let ı be different in every period.owever, we relax the fixed ı assumption by allowing the cur-

ent episode of democracy to have a weight that is different fromhat of the more distant past. Specifically, in the estimating equa-ion we include DemocraticCapitalown measured at the start of theurrent democracy episode. Information about the current episodef democracy is incorporated separately in a form of democracyuration. Other things being equal, we expect to find negative dura-ion dependence: the longer the country remains a democracy, theess likely it is to make a transition to non-democracy in the nexteriod.

Similar to Persson and Tabellini (2009), the other componentsf the democratic capital are each defined as a weighted sum ofither regime indices (when using constructed data) or polity2cores (when using Polity IV data) of the considered countries in the
revious year, where each county’s weight is proportional to
8 We use the binary democracy variable rather than polity2 score when creatingwn democratic capital to avoid a substantial data loss. Also, this definition makesur variable similar to that used by Persson and Tabellini (2009).9 The choice of ıis based on Persson and Tabellini (2009) who find that the dis-

ount factor typically ranges from 0.94 to 0.99


Table 1aDemocracy episodes, constructed data.

Albania 1920–1928, 2005–2006 Guatemala 1946–1954, 1996–2006 Palestine 1994–2006Antigua and Barbuda 1981–1990, 2004–2006 Guinea Bissau 1993–1999, 2000–2003, 2005–2006 Panama 1990–2006Argentina 1916–1930, 1946–1951, 1973–1976,

1983–2006Guyana 1993–2006 Papua New Guinea 1975–2006

Australia 1923–2006 Haiti 1990–1991, 1994–1999 Paraguay 1993–2006Austria 1919–1933, 1945–2006 Honduras 1998–2006 Peru 1945–1948, 1956–1962, 1964–1968,

1980–1992, 2001–2006Azerbaijan 1992–1993 Hungary 1990–2006 Philippines 1946–1965, 1986–2006Bahamas 1973–2006 Iceland 1915–1940, 1944–2006 Poland 1919–1926, 1989–2006Bangladesh 1947–1956, 1972–1975, 1991–1995,

1996–2003India 1951–1975, 1977–2006 Portugal 1976–2006

Barbados 1966–2006 Indonesia 1955–1959, 1999–2006 Romania 1929–1938, 1989–2006Belgium 1921–1940, 1945–2006 Ireland 1922–2006 Russia 1991–1993, 1995–1999Belize 1981–2006 Israel 1947–2006 Saint Kitts and Nevis, 1983–2006Benin 1960–1963, 1991–2006 Italy 1946–2006 Saint Lucia 1979–2006Bermuda 1964–2006 Jamaica 1962–2006 Saint Vincent and the Grenadines 1979–2006Bolivia 1956–1965, 1980–1981, 1982–2006 Japan 1952–2006 Samoa 1962–2006Botswana 1966–2006 Kenya 1963–1967, 2002–2006 Sao Tome and Principe 1991–2006Brazil 1946–1964, 1985–2006 Kiribati & Tuvalu Gilbert Islands 1972–2006 Senegal 2000–2006Bulgaria 1919–1921, 1990–2006 Laos 1954–1959 Serbia 2000–2006Burma Myanmar 1948–1962 Latvia 1921–1935, 1991–2006 Seychelles 1976–1977, 1993–2006Burundi 2003–2006 Lebanon 1946–1975 Sierra Leone 1961–1967, 1996–1997, 2002–2006Canada 1960–2006 Lesotho 2003–2006 Slovak Republic 1990–2006Cape Verde Islands 1991–2006 Liberia 2005–2006 Slovenia 1990–2006Central African Republic 1991–2006 Lithuania 1921–1927, 1991–2006 Solomon Islands 1979–2006Chile 1933–1973, 1989–2006 Luxembourg 1919–1940, 1945–2006 Somalia 1960–1969Colombia 1974– 2006 Macedonia 1991–2006 South Africa 1994–2006Comoros Islands 1996–2006 Madagascar 1961–1972, 1991–2006 South Korea 1960–1961, 1988–2006Costa Rica 1948–2006 Malawi 1994–2003 Spain 1931–1936, 1978–2006Croatia 2000–2006 Malaysia 1958–1973 Sri Lanka 1949–2006Cuba 1944–1952 Maldive Islands 1969–1975 Sudan 1957–1958, 1966–1971, 1987–1989Cyprus 1977–2006 Mali 1992–2006 Suriname 1975–1980, 1987–1990, 1991–2006Czech Republic 1918–1939, 1990–2006 Malta and Gozo 1964–2006 Sweden 1919–2006Denmark 1915–1941, 1945–2006 Marshall Islands 1966–2006 Switzerland 1971–2006Djibouti 1977–1981 Mauritius 1968–2006 Syria 1955–1959, 1962–1963Dominica 1978–2006 Mexico 2000–2006 Taiwan 1991–2006Dominican Republic 1962–1963, 1970–2006 Micronesia, Federated States of 1984–2006 Tanzania 1960–1962Ecuador 1948–1961, 1980–1996, 2003–2004 Moldova 1994–2006 Thailand 1975–1976, 1992–2006El Salvador 1984–2006 Mongolia 1991–2006 Trinidad and Tobago 1962–2006Estonia 1919–1935, 1991–2006 Mozambique 1994–2006 Turkey 1961–1972, 1973–1980, 1983–2006Fiji 1969–1987, 1999–2001 Namibia 1991–2006 Uganda 1962–1967Finland 1917–2006 Nepal 1959–1960, 1991–2002 Ukraine 2004–2006France 1945–2006 Netherlands 1919–1940, 1945–2006 United Kingdom 1928–2006Gabon 1960–1963, 1964–1968 New Zealand 1891–2006 United States of America 1920–2006Gambia 1965–1994 Nicaragua 1990–2006 Uruguay 1932–1934, 1939–1973, 1985–2006Georgia 2003–2006 Niger 1993–1996, 1999–2006 Vanuatu New Hebrides 1980–2006Germany 1919–1933, 1949–2006 Nigeria 1960–1966, 1978–1983, 1999–2006 Venezuela 1946–1948, 1958–1999Ghana 1957–1960, 1969–1972, 1979–1981,

1996–2006Norway 1913–1940, 1945–2006 Zaire (Congo, Democratic Republic) 1960–1963,

1992–1997Greece 1950–1967, 1974–2006 Pakistan 1947–1956, 1972–1977, 1988–1999 Zambia 1991–1997Grenada 1975–1979, 1985–2006 Palau 1984–2006

Source: Noguera (2007a).T he an

tG

D

D

daait

waot

o

otal: 140 countries, 215 episodes of democracy. Episodes in italics are not used in t

he size of the economy as measured by the country’s realDP10:

emocraticCapitalpeersit

=∑

j ∈ Gpi

,j/∈iGDPj,t−1f ownj,t−1∑

j ∈ Gpi

,j/∈iGDPj,t−1, (9)

∑own

emocraticCapitalotherit = j ∈ Go

iGDPj,t−1f

j,t−1∑j ∈ Go

iGDPj,t−1

, (10)

10 By construction, the impact of a particular country’s political experience on theemocratic capital measure is greater for the countries with large economies, suchs the United States, and it is very small for economies like the Marshall Islands. Wenticipate that as voting rights and democratic values become more prevalent innfluential economies and spread around the world, this will improve the chanceshat the remaining countries will choose a democratic path.

rdctaDmfcv

alysis because GDP data were missing for the whole episode.

here Gip is a set of countries that belong to the same peer group

s country i, as identified in Appendix B, and Gio is a set of all

ther countries. In some specifications the peer group is definedo include all countries in the world.

By construction, all democratic capital measures that are basedn constructed data can each take on values from 1 (autoc-acy) to 7 (democracy). In practice, the amount of variationepends on the measure. As seen in Table 2, among the demo-ratic capital variables that were created using constructed datahere appears to be much more variation in DemocraticCapitalown

nd DemocraticCapitalpeers, than in DemocraticCapitalother andemocraticCapitalworld, which is as expected because the latter two
easures include more countries. The average value is the lowest
or DemocraticCapitalown. Similarly, when looking at the demo-ratic capital measures created from the Polity IV data, there is lessariation in DemocraticCapitalother and DemocraticCapitalworld than

T.D. Jeitschko et al. / Journal of Behavioral and Experimental Economics 50 (2014) 13–28 19

Table 1bDemocracy episodes, polity IV data.

Albania 1990–1996, 1997–2004 Gambia 1965–1994 Nigeria 1960–1966, 1979–1984, 1999–2004Algeria 2004–2004 Georgia 1991–2004 Norway 1898–2004Argentina 1880–1930, 1937–1943, 1973–1976,

1983–2004Germany 1890–1933, 1949–2004 Pakistan 1948–1958, 1962–1970, 1972–1977,

1988–1999Armenia 1991–1996, 1998–2004 Ghana 1970–1972, 1979–1981, 1996–2004 Panama 1955–1968, 1989–2004Australia 1901–2004 Greece 1863–1922, 1926–1936, 1944–1967,

1974–2004Papua New Guinea, 1975–2004

Austria 1919–1933, 1945–2004 Guatemala 1879–1896, 1898–1900, 1921–1931,1944–1954, 1966–1974, 1986–2004

Paraguay 1937–1940, 1989–2004

Azerbaijan 1992–1993 Guinea Bissau 1994–1998, 1999–2003 Peru 1886–1919, 1933–1948, 1956–1962,1963–1968, 1979–1992, 1993–2004

Bangladesh 1972–1974, 1991–2004 Guyana 1966–1978, 1992–2004 Philippines 1935–1972, 1986–2004Belarus 1991–1995 Haiti 1918–1935, 1990–1991, 1994–2000 Poland 1918–1926, 1989–2004Belgium 1853–2004 Honduras 1894–1904, 1908–1936, 1980–2004 Portugal 1908–1928, 1975–2004Benin 1960–1963, 1991–2004 Hungary 1989–2004 Romania 1990–2004Bolivia 1880–1936, 1982–2004 India 1950–2004 Russia 1992–2004Bosnia Herzegovina 1939–1941 Indonesia 1946–1950, 1999–2004 Senegal 2000–2004Botswana 1966–2004 Iran 1997–2004 Serbia 1843–1858, 1903–1941, 2000–2004Brazil 1946–1964, 1985–2004 Ireland 1921–2004 Sierra Leone 1961–1967, 1968–1971, 1996–1997,

2001–2004Bulgaria 1918–1919, 1990–2004 Israel 1948–2004 Singapore 1959–1962Burkina Fasso 1977–1980 Italy 1945–2004 Slovak Republic 1945–1947, 1990–2004Burma Myanmar 1948–1962 Jamaica 1959–2004 Slovenia 1939–1941, 1991–2004Cambodia Kampuchea Khmer 1990–1997,

1998–2004Japan 1868–2004 Solomon Islands 1978–2000, 2004

Canada 1867–2004 Kenya 1963–1966, 2002–2004 Somalia 1960–1969Central African Republic 1993–2003 Laos 1955–1960 South Africa 1910–2004Chile 1874–1924, 1935–1973, 1989–2004 Latvia 1920–1934, 1991–2004 South Korea 1960–1961, 1963–1972, 1987–2004China 1912–1913 Lebanon 1943–1975 Spain 1871–1873, 1879–1923, 1930–1939,

1976–2004Colombia 1843–1860, 1867–1886, 1930–1948,

1957–2004Lesotho 1966–1970, 1993–1998, 1999–2004 Sri Lanka 1948–2004

Comoros Islands 1975–1976, 1990–1995,1996–1999, 2002–2004

Liberia 1847–1884 Sudan 1954–1958, 1965–1970, 1986–1989

Costa Rica 1843–2004 Lithuania 1918–1926, 1991–2004 Sweden 1910–2004Croatia 1939–1941, 1999–2004 Macedonia 1939–1941, 1991–2004 Switzerland 1848–2004Cuba 1902–1952 Madagascar 1991–2004 Syria 1944–1949, 1950–1951, 1954–1957Cyprus 1960–1963, 1968–2004 Malawi 1994–2004 Taiwan 1992–2004Czech Republic 1918–1939, 1945–1947,

1990–2004Malaysia 1957–2004 Tanzania 2000–2004

Denmark 1849–1866, 1904–2004 Mali 1992–2004 Thailand 1969–1971, 1974–1976, 1978–1991,1992–2004

Djibouti 1999–2004 Mauritius 1968–2004 Trinidad and Tobago 1962–2004Dominican Republic 1962–1963, 1978–2004 Mexico 1994–2004 Turkey 1946–1971, 1973–1980, 1983–2004Ecuador 1948–1961, 1968–1970, 1979–2004 Moldova 1991–2004 Uganda 1962–1966, 1980–1985Egypt 1922–1929, 1935–1952 Mongolia 1990–2004 Ukraine 1991–2004El Salvador 1982–2004 Mozembique 1994–2004 United Kingdom 1843–2004Equatorial Guinea 1968–1969 Namibia 1990–2004 United States of America 1843–2004Estonia 1917–1935, 1991–2004 Nepal 1959–1960, 1990–2002 Uruguay 1910–1934, 1952–1972, 1985–2004Ethiopia 1855–1930, 1993–2004 Netherlands 1917–2004 Venezuela 1958–2004Fiji 1970–1987, 1990–2004 New Zealand 1857–2004 Zambia 1964–1968, 1991–2004Finland 1917–2004 Nicaragua 1990–2004 Zimbabwe 1970–1987France 1848–1851, 1872–1940, 1945–2004 Niger 1991–1996, 1999–2004

Source: Polity IV website, http://www.systemicpeace.org/polity/polity4.htm.T he an

iccf−s

4

evtt

nperiod helps to smooth the time series, so that there is no need toaccount for business cycles.12

11 Data for GDP per capita measures were taken fromUSDA ERS International Macroeconomic Data Set: 1969–2007http://www.ers.usda.gov/Data/Macroeconomics/; Penn World Tables:1950–1968, Center for International Comparisons, University of Pennsyl-

otal: 125 countries, 226 episodes of democracy. Episodes in italics are not used in t

n DemocraticCapitalpeers. The comparisons between own demo-ratic capital and other measures are more complicated in thease of Polity IV data because the polity2 variable and hence,oreign democratic capital measures, can take on values from10 to 10, while own democratic capital is measured on a [0,1]

cale.

.2. Expectations of future economic growth

When measuring expectations we follow a common approach
mployed in the forecasting literature, and use the most recentalues of a variable to predict its future values. Specifically, we usehe average GDP growth over the last five years as a measure ofhe expected future GDP growth under the status quo (democracy,
vAh

t

alysis because GDP data were missing for the whole episode.

o regime change).11 Averaging the GDP growth over a five-year

ania, http://pwt.econ.upenn.edu/; and the Maddison Historical Statistics,ngus Maddison Web Site, Economics Department, University of Wisconsinttp://www.ggdc.net/maddison/.12 As a robustness check, we re-ran our regressions using single-year values forhe measures of economic performance, and results were qualitatively unchanged.

http://www.systemicpeace.org/polity/polity4.htm

http://www.ers.usda.gov/Data/Macroeconomics/

http://pwt.econ.upenn.edu/

http://www.ggdc.net/maddison/


Table 2Descriptive statistics.

Variable name Constructed DATA Polity IV data

Democratic CapitalOwn at the start of democracy episode, ı = 0.94 3.901(1.486)

0.084(0.173)

Democratic CapitalOwn at the start of democracy episode, ı = 0.99 4.016(1.461)

0.104(0.209)

Democratic CapitalPeers 5.726(1.493)

4.754(5.084)

Democratic CapitalOther 6.121(0.473)

7.054(1.291)

Democratic CapitalWorld 6.159(0.480)

7.106(1.283)

Log (average per capita GDP in t − 1) 8.082(1.394)

7.934(1.354)

Average per capita GDP growth in t − 1 (percentage points) 2.070(3.149)

1.932(3.256)

Difference in average GDP growth in t − 1 (percentage points) −0.378(3.009)

−0.381(3.043)

Number of observations 4128 4105Number of democracy episodes 208 193Number of countries 140 121

Averages for the per capita GDP, per capita GDP growth and difference in the GDP growth are computed over the past five years. GDP growth difference is defined as adifference between the anticipated GDP growth after democracy breakdown and the expected GDP growth under democracy.S

pefirtusaK

poecgpgccacvvBehcr

tdhttue

w

ao

4

tcactAtpcbehFfgSa

5

(ttmrhayt

tandard deviations in parentheses under the sample means.

Regarding the measurement of the expected growth after theolitical transition (from democracy to non-democracy), the gen-ral approach is the same: use the immediate past to predict theuture. Because we don’t observe the outcome under dictatorshipn the country’s immediate past, we use the most recent expe-ience of peer countries instead. This approach is motivated byheoretical and empirical literature that suggests that individualsse experiences of others (peers or individuals belonging to theame social group) to infer about outcomes that are possible tochieve (Clark, Kristensen, and Westergard-Nielsen, 2009; Foley,idder, and Powell, 2002, Major, 1994, for example).

A preferred measure would incorporate past experiences ofeer countries that are non-democracies. Unfortunately, the setf countries with non-democratic regimes is very small (or evenmpty) for several peer groups. Conversely, expanding the set of theountries to include non-democratic states from outside of the peerroup would generate inappropriate reference points. For exam-le, countries in Latin America are more likely to form economicrowth expectations based on the economic performance of otherountries in the region than based on economic performance ofountries in Eastern Europe. Therefore, we define expected growthfter the regime switch as a growth rate in real GDP of all otherountries in the peer group. This measure, although imperfect, isalid because recent economic performance of peer countries pro-ides information about the outcomes that are possible to achieve.etter economic performance in peer countries (as compared to theconomic performance of the country in question) would signal of aigher chance that replacing an economically unsuccessful demo-ratic regime with a different (non-democratic) regime wouldesult in improved economic performance.

It is important to note that actual realizations of growth afterhe transition (or no transition) from democracy to non-democracyo not matter. All that matters is expectations. If actual growthappens to be lower or higher than the forecasted value dueo unforeseen circumstances, it does not invalidate our expec-ations measures because those circumstances would have been
nforeseen, and therefore, would be irrelevant to forming priorxpectations.
The difference between the two growth rates (expected growthhen maintaining a democratic regime minus the expected growth

tt

d

fter the regime change) is the variable that we use in our analysisf democracy breakdown.

.3. Other variables

Although the goal of our empirical analysis is to study howhe likelihood of democracy breakdown is related to democraticapital and relative economic performance, in our regressions welso include the natural logarithm of the country’s real GDP perapita and the growth rate in GDP per capita, each averaged overhe past five years (summary statistics are reported in Table 2).s mentioned earlier, establishing a causal relationship between

hese variables and democratization is complicated because of theresence of unobserved heterogeneity and possibility of reverseausality. In this paper, we do not study a causal relationshipetween measures of economic performance and democracy; how-ver, we include these measures to account for cross-countryeterogeneity due to variation in historical and cultural factors.or the same reason, in all regressions we also include dummiesor international and civil war times, indicators for colonial ori-in (Ottoman, Austrian, Dutch, Belgian, German, British, French,panish, and Portuguese), socialist legal origin, African location,nd Middle East location.

. Estimation strategy

We study democratic regime stability by employing durationsurvival) analysis. Specifically, we consider the determinants ofhe hazard rate – the probability of a democracy breakdown, givenhat a country has remained a democracy up until the current

oment. Because we use yearly data, and the episodes of democ-acy may be as short as 2–3 years, we employ the discrete-timeazard. This approach also allows us to account for the fact that

country’s political regime is recorded at the end of the calendarear, although the actual democracy breakdown may occur at anyime during the year (Jenkins, 2005; Wooldridge, 2002). Moreover,
he incorporation of time-varying covariates is straightforward inhis context.
Formally, let ht(X) denote the probability of a switch to non-emocracy some time during year t, conditional on observing a

and E

dihi

l

y

l

dCwinec

mapT

h

doydreiew

aidar

oAptrIs

dltdcebfe

e

diit3Pit

6

trhfimobire

alfim(dpttcmottdiPstasdcTi

iridlddemocratic capital is included in the regression. It is also consistentwith results reported by Goldstone et al. (2010), who find that anadverse regime change is less likely to occur if preceded by a period


emocratic regime until the beginning of year t. The function ht(X)s a discrete time hazard function (Jenkins, 2005). Then, the likeli-ood for country i to experience democracy breakdown in year T

s

i = hT (Xit)T−1∏t=1

(1 − ht(Xit)). (11)

On the other hand, for a country that remains a democracy inear T, the likelihood is

j =T∏

t=1

(1 − ht(Xjt)). (12)

For each country in our sample, it either retains or gives upemocracy until the end of 2006 (our last year of observation).ountries remaining as democracies are subject to right censoring,hich is reflected in the likelihood function (Eq. (12)). The start-

ng years of democracies are known for all countries, so there iso need to account for left censoring. Naturally, T may be differ-nt for each country, as the duration of democracies differs acrossountries, but we suppress subscript i to simplify the notation.

Next we obtain a standard cloglog (complementary log–log)odel by limiting our attention to proportional hazard models and

ssuming a typical exponential form of the country-specific com-onent of the hazard function (Jenkins, 2005; Wooldridge, 2002).hus, the resulting model is

t(Xit) = 1 − exp[− exp(Xit ̌ + �t + log vi)]. (13)

In the equation above, Xit is a vector of covariates (measures ofemocratic capital, difference in the expected future growth, andther variables); parameter �t summarizes the baseline hazard forear t and characterizes duration dependence. In our regressions,uration dependence is modeled using time dummies, which cor-espond to different values of t. In Eq. (13), vi is an unobservedffect that captures unobserved heterogeneity across countries ands assumed to be independent of covariates. We estimate the mod-ls with and without vi. In models with unobserved heterogeneity,e assume that vi has a log normal distribution.

Because we use a discrete-time hazard, the dependent vari-ble is a binary response variable that takes on a value of zerof there is no regime change in period t (the country remains aemocracy) and it equals one if a country experiences a switch to

non-democratic regime. Once a switch has occurred, the democ-acy episode is considered to be completed.

A number of countries experienced more than one episodef democracy during the considered period (examples includergentina, Denmark, Spain, and others). We follow commonractice and treat such episodes as separate events, but allowhe country’s previous democratic experience to enter the cur-ent democracy episode via the own democratic capital measure.n models with unobserved heterogeneity, such episodes have theame unobserved country effect, vi.

As is standard in duration models, in our estimation eachemocracy episode is interpreted as a single observation and the

ikelihood function stated in Eqs. (11) and (12) is assumed to behe true probability of breakdown and survival, respectively, con-itional on no breakdown prior to the current period. Thus, theorresponding estimation method is a full maximum likelihoodstimator, which implies that all previous information has alreadyeen incorporated, and no serial correlation may be present. There-
ore, there is no need to account for serial dependence in ourstimation.
Moreover, it is important to note that not all democracypisodes were used in the estimation. A number of episodes were

csm


ropped from the regression analysis because GDP data were miss-ng for the entire duration of the democracy episode. Also, due todentification issues, it is not possible to include year dummies inhe regressions. When year dummies were included, more than0% of observations had to be dropped from the analysis (similar toersson and Tabellini, 2009). Therefore, in some specifications wenclude five-year dummies instead of one-year dummies to capturehe time-specific shocks that are common to all countries.

. Empirical results

The results from complementary log–log regressions that usehe constructed and Polity IV data are presented in Tables 3a and 3b,espectively. These results are for the models without unobservedeterogeneity. The last row provides information on goodness oft – the percentage of outcomes that are correctly predicted. Toake the predictions, we chose thresholds such that the fraction

f predicted transitions would be the same as the fraction of actualreakdowns in the data (Wooldridge, 2002, Chapter 15). The results

ndicate that our models predict the outcome in each given periodather accurately, with the percentage correctly predicted alwaysxceeding 95 percent.

In the constructed data, the risk of democracy breakdown is neg-tively related to country’s own democratic capital (Table 3a). Thisink is stronger and highly significant in the regressions that includeve-year dummies: when the value of the own democratic capitaleasured at the start of the democracy episode changes from 1

autocracy) to 7 (democracy), the predicted probability of break-own (conditional on no breakdown so far) is reduced by about 84ercent.13 Similarly, coefficients on the foreign democratic capi-al are typically negative; however, they become insignificant oncehe five-year dummies are included. Because five-year dummiesapture time-specific economic and political shocks that are com-on to all countries (such as international economic crises, periods

f global political instability, or general changes in individual atti-udes toward democratic values), the latter result indicates thathe effect of the foreign democratic capital on democracy break-own cannot be distinguished from the effect of the overall changes

n the economic and political environment in the world. In theolity IV data, the coefficient on the own democratic capital mea-ure is insignificant (Table 3b). Similar to the results obtained onhe constructed data, the Polity IV regressions produce negativend significant coefficients on the foreign democratic capital mea-ures, but those coefficients become insignificant after the five-yearummies are added to the regressions. In all regressions, coeffi-ients on time dummies indicate negative duration dependence.14

he likelihood of a transition away from democracy is the highestn the first 12 years and is lower afterwards.

Our finding that the coefficient on DemocraticCapitalown is oftennsignificant, while coefficients on time (duration) dummies areather large and statistically significant implies that after account-ng for the most recent own democratic experience, past ownemocratic experience is not so important. This finding is simi-

ar to Persson and Tabellini (2009), who also find the effect of pastomestic democratic capital is insignificant when current domestic

13 The effects are computed as exp(b) − 1, where b is an estimate of ˇ.14 Dummy variables were specified for the intervals of democracy duration. Thehoice of the intervals was in part determined based on observation (similar tran-ition rates over the range of t) and was also partly dictated by data properties (forany values of t breakdowns were extremely rare – 0 or 1 breakdowns).

22

T.D.

Jeitschko et

al. /

Journal of

Behavioral and

Experimental

Economics

50 (2014)

13–28

Table 3aComplementary log–log estimates of democracy breakdown using constructed data.

Variables ı = 0.94 ı = 0.99

(1) (2) (3) (4) (5) (6) (7) (8)

Log (average per capita GDP in t − 1) −0.593***

(0.115)−0.566***

(0.121)−0.649***

(0.110)−0.563***

(0.115)−0.586***

(0.115)−0.565***

(0.121)−0.645***

(0.110)−0.561***

(0.114)Average per capita GDP growth in t − 1 0.055

(0.043)0.031(0.049)

0.061(0.043)

0.032(0.049)

0.058(0.043)

0.033(0.049)

0.063(0.043)

0.034(0.048)

Difference in average GDP growth in t − 1 −0.171***

(0.046)−0.151***

(0.050)−0.171***

(0.046)−0.153***

(0.050)−0.174***

(0.046)−0.154***

(0.050)−0.174***

(0.046)−0.156***

(0.049)DemocraticCapitalOwn at the start of democracy

episode−0.196*

(0.106)−0.308***

(0.116)−0.182*

(0.105)−0.313***

(0.116)−0.215*

(0.111)−0.304**

(0.119)−0.193*

(0.109)−0.312***

(0.119)DemocraticCapitalPeers −0.160*

(0.084)0.032(0.101)

−0.160*

(0.084)0.029(0.101)

DemocraticCapitalOther −0.859***

(0.215)0.375(0.637)

−0.850***

(0.213)0.288(0.632)

DemocraticCapitalWorld −0.990***

(0.200)−0.018(0.703)

−0.977***

(0.197)−0.037(0.705)

Dummy for 2 ≤ t ≤ 5 1.489***

(0.382)1.414***

(0.382)1.491***

(0.382)1.422***

(0.382)1.489***

(0.382)1.423***

(0.383)1.491***

(0.382)1.428***

(0.383)Dummy for 6 ≤ t ≤ 12 1.244***

(0.347)1.324***

(0.350)1.252***

(0.347)1.330***

(0.350)1.242***

(0.347)1.325***

(0.351)1.251***

(0.348)1.329***

(0.351)Dummy for 13 ≤ t ≤ 20 0.779**

(0.341)0.744**

(0.351)0.791**

(0.340)0.750**

(0.351)0.791**

(0.341)0.778**

(0.351)0.802**

(0.340)0.783**

(0.351)Dummy for 21 ≤ t ≤ 30 1.010**

(0.498)1.041**

(0.511)0.950*

(0.500)1.063**

(0.510)1.005**

(0.498)1.023**

(0.512)0.943*

(0.500)1.041**

(0.510)Constant 6.053***

(1.695)−17.722(689.738)

6.355***

(1.644)−15.009(689.828)

6.004***

(1.670)−17.126(681.968)

6.273***

(1.618)−14.867(681.767)

5-year dummies included? No Yes No Yes No Yes No YesObservations 4039 3862 4039 3862 4039 3862 4039 3862Number of democracy episodes 204 204 204 204 204 204 204 204Number of countries 136 136 136 136 136 136 136 136Percentage correctly predicted 96.29 96.2 95.83 96.17 96.37 96.12 95.81 96.2

Dependent variable is a binary variable that equals one if democracy breaks down and zero if a country remains a democracy. Averages for the per capita GDP, per capita GDP growth and difference in the GDP growth arecomputed over the past five years. GDP growth difference is defined as a difference between the GDP growth in continued democracy and anticipated GDP growth after democracy breakdown. All regressions include dummiesfor international and civil war times, indicators for colonial origin (Ottoman, Austrian, Dutch, Belgian, German, British, French, Spanish, and Portuguese), socialist legal origin, African location, and Middle East location.Reference category: t > 30.

*** Significant at 1%.** Significant at 5%.* Significant at 10%.

T.D.

Jeitschko et

al. /

Journal of

Behavioral and

Experimental

Economics

50 (2014)

13–28

23

Table 3bComplementary log–log estimates of democracy breakdown using polity IV data.

Variables ı = 0.94 ı = 0.99

(1) (2) (3) (4) (5) (6) (7) (8)


(0.133)−0.924***

(0.149)−0.839***

(0.129)−0.956***

(0.144)−0.753***

(0.133)−0.923***

(0.150)−0.833***

(0.129)−0.955***

(0.145)Average per capita GDP growth in t − 1 −0.004

(0.038)−0.029(0.039)

−0.010(0.037)

−0.041(0.037)

−0.004(0.038)

−0.029(0.039)

−0.010(0.037)

−0.041(0.037)


(0.040)−0.116***

(0.041)−0.111***

(0.037)−0.108***

(0.038)−0.127***

(0.040)−0.115***

(0.041)−0.111***

(0.037)−0.108***

(0.038)DemocraticCapitalOwn at the start of democracy episode 0.398

(0.682)0.254(0.719)

0.406(0.682)

0.288(0.718)

0.092(0.615)

0.082(0.660)

0.104(0.614)

0.098(0.658)

DemocraticCapitalPeers −0.080***

(0.030)−0.018(0.037)

−0.080***

(0.030)−0.018(0.037)

DemocraticCapitalOther −0.131(0.100)

0.412(0.293)

−0.132(0.101)

0.414(0.294)


(0.089)0.192(0.381)

−0.231**

(0.090)0.188(0.381)

Dummy for 2 ≤ t ≤ 5 0.691**

(0.339)0.677*

(0.346)0.641*

(0.340)0.640*

(0.344)0.717**

(0.339)0.690**

(0.346)0.665*

(0.341)0.655*

(0.345)Dummy for 6 ≤ t ≤ 12 0.958***

(0.307)1.082***

(0.317)0.882***

(0.306)1.062***

(0.316)0.978***

(0.310)1.088***

(0.319)0.900***

(0.309)1.070***

(0.319)Dummy for 13 ≤ t ≤ 20 0.443

(0.332)0.482(0.354)

0.451(0.333)

0.494(0.353)

0.447(0.333)

0.479(0.354)

0.453(0.333)

0.490(0.353)

Dummy for 21 ≤ t ≤ 30 0.767*

(0.421)0.674(0.440)

0.775*

(0.421)0.678(0.440)

0.770*

(0.421)0.671(0.441)

0.777*

(0.421)0.676(0.441)

Constant 2.408*

(1.297)−1.954(2.802)

3.374***

(1.199)−0.051(3.444)

2.395*

(1.294)−1.934(2.802)

3.360***

(1.197)0.017(3.440)


Dependent variable is a binary variable that equals one if democracy breaks down and zero if a country remains a democracy. Averages for the per capita GDP, per capita GDP growth and difference in the GDP growth arecomputed over the past five years. GDP growth difference is defined as a difference between the GDP growth in continued democracy and anticipated GDP growth after democracy breakdown. All regressions include dummiesfor international and civil war times, indicators for colonial origin (Ottoman, Austrian, Dutch, Belgian, German, British, French, Spanish, and Portuguese), socialist legal origin, African location, and Middle East location.Standard errors in parentheses.Reference category: t > 30.


24

T.D.

Jeitschko et

al. /

Journal of

Behavioral and

Experimental

Economics

50 (2014)

13–28

Table 4Complementary log–log estimates of democracy breakdown; episodes ending in 1939–1941 excluded (ı = 0.94).

Variables Constructed data Polity IV data

(1) (2) (3) (4) (5) (6) (7) (8)


(0.116)−0.561***

(0.124)−0.617***

(0.112)−0.549***

(0.117)−0.770***

(0.134)−0.958***

(0.153)−0.849***

(0.131)−0.995***


(0.045)0.019(0.050)

0.047(0.045)

0.019(0.050)

−0.018(0.039)

−0.047(0.040)

−0.022(0.038)

−0.057(0.039)


(0.048)−0.129**

(0.051)−0.147***

(0.048)−0.130**

(0.051)−0.112***

(0.042)−0.109**

(0.042)−0.099**

(0.039)−0.100**

(0.040)DemocraticCapitalOwn at the start of democracy episode −0.194*

(0.115)−0.330***

(0.125)−0.168(0.114)

−0.333***

(0.125)0.325(0.702)

0.314(0.738)

0.356(0.703)

0.340(0.737)

DemocraticCapitalPeers −0.134(0.089)

0.036(0.103)

−0.080***

(0.030)−0.023(0.038)

DemocraticCapitalOther −0.651**

(0.271)0.090(0.761)

−0.122(0.106)

0.306(0.335)


(0.257)0.019(0.857)

−0.205**

(0.096)0.263(0.419)

Dummy for 2 ≤ t ≤ 5 1.515***

(0.388)1.381***

(0.390)1.524***

(0.388)1.385***

(0.390)0.726**

(0.342)0.681*

(0.348)0.665*

(0.343)0.652*

(0.346)Dummy for 6 ≤ t ≤ 12 1.235***

(0.353)1.322***

(0.360)1.243***

(0.353)1.322***

(0.359)0.964***

(0.311)1.096***

(0.321)0.882***

(0.310)1.079***

(0.321)Dummy for 13 ≤ t ≤ 20 0.755**

(0.345)0.740**

(0.358)0.760**

(0.344)0.737**

(0.358)0.549*

(0.334)0.615*

(0.354)0.534(0.335)

0.633*

(0.354)Dummy for 21 ≤ t ≤ 30 0.139

(0.677)0.317(0.688)

0.105(0.676)

0.324(0.688)

0.706(0.441)

0.572(0.461)

0.706(0.441)

0.578(0.461)

Constant 4.423**

(1.972)−16.558(865.350)

4.573**

(1.928)−15.947(866.666)

2.420*

(1.350)−0.791(3.137)

3.281***

(1.253)−0.391(3.740)


Dependent variable is a binary variable that equals one if democracy breaks down and zero if a country remains a democracy. Averages for the per capita GDP, per capita GDP growth and difference in the GDP growth arecomputed over the past five years. GDP growth difference is defined as a difference between the GDP growth in continued democracy and anticipated GDP growth after democracy breakdown. All regressions include dummiesfor international and civil war times, indicators for colonial origin (Ottoman, Austrian, Dutch, Belgian, German, British, French, Spanish, and Portuguese), socialist legal origin, African location, and Middle East location.Standard errors in parentheses.Reference category: t > 30.


T.D.

Jeitschko et

al. /

Journal of

Behavioral and

Experimental

Economics

50 (2014)

13–28

25

Table 5Complementary log–log estimates of democracy breakdown, unobserved heterogeneity/frailty permitted (ı = 0.94).

Variables Constructed data Polity IV data

World War II included World War II excluded World War II included World War II excluded

(1) (2) (3) (4) (5) (6) (7) (8)


(0.128)−0.571***

(0.125)−0.509***

(0.130)−0.543***

(0.127)−0.706***

(0.182)−0.777***

(0.167)−0.707***

(0.182)−0.773***


(0.046)0.058(0.046)

0.045(0.048)

0.049(0.049)

−0.010(0.042)

−0.019(0.041)

−0.010(0.042)

−0.018(0.041)


(0.053)−0.190***

(0.053)−0.168***

(0.054)−0.173***

(0.054)−0.137***

(0.044)−0.115***

(0.040)−0.129***

(0.045)−0.109***

(0.040)DemocraticCapitalOwn at the start of democracy episode −0.167

(0.109)−0.157(0.108)

−0.177(0.118)

−0.155(0.117)

0.479(0.786)

0.490(0.741)

0.369(0.818)

0.448(0.763)

DemocraticCapitalPeers −0.140(0.090)

−0.119(0.095)

−0.092***

(0.034)−0.092***

(0.035)DemocraticCapitalOther −0.836***

(0.231)−0.628**

(0.293)−0.147(0.126)

−0.132(0.130)


(0.215)−0.733***

(0.278)−0.270**

(0.109)−0.255**

(0.113)Dummy for 2 ≤ t ≤ 5 1.569***

(0.403)1.571***

(0.403)1.605***

(0.412)1.609***

(0.412)1.013**

(0.431)0.960**

(0.414)1.013**

(0.432)0.975**

(0.416)Dummy for 6 ≤ t ≤ 12 1.250***

(0.372)1.268***

(0.373)1.253***

(0.381)1.267***

(0.381)1.194***

(0.377)1.101***

(0.378)1.163***

(0.380)1.080***

(0.381)Dummy for 13 ≤ t ≤ 20 0.906**

(0.363)0.910**

(0.364)0.885**

(0.372)0.886**

(0.372)0.860**

(0.382)0.848**

(0.380)0.869**

(0.385)0.846**

(0.382)Dummy for 21 ≤ t ≤ 30 0.947*

(0.556)0.890(0.560)

−0.004(0.804)

−0.026(0.804)

1.335**

(0.549)1.330**

(0.531)1.348**

(0.553)1.325**

(0.532)Constant 5.213***

(1.885)5.552***

(1.832)3.626*

(2.169)3.817*

(2.134)2.005(1.733)

3.012**

(1.526)1.851(1.756)

2.850*

(1.550)

5-year dummies included? No No No No No No No NoLikelihood ratio test for unobserved heterogeneity �̄2

1 = 0.0001(0.497)

�̄21 = 0.0000

(0.498)�̄2

1 = 0.0000(0.497)

�̄21 = 0.0001

(0.496)�̄2

1 = 0.002(0.482)

�̄21 = 0.0000

(0.498)�̄2

1 = 0.03(0.436)

�̄21 = 0.0001

(0.497)Observations 3577 3577 3454 3454 2620 2624 2610 2614Number of democracy episodes 188 188 183 183 161 161 160 160Number of countries 126 126 126 126 102 102 102 102

Dependent variable is a binary variable that equals one if democracy breaks down and zero if a country remains a democracy. Averages for the per capita GDP, per capita GDP growth and difference in the GDP growth arecomputed over the past five years. GDP growth difference is defined as a difference between the GDP growth in continued democracy and anticipated GDP growth after democracy breakdown. All regressions include dummiesfor international and civil war times, indicators for colonial origin (Ottoman, Austrian, Dutch, Belgian, German, British, French, Spanish, and Portuguese), socialist legal origin, African location, and Middle East location.Standard errors in parentheses; p-values in parentheses under the test statisticsReference category: t > 30.


2 and E

oir

scngrcfiitt(

sdmacctGtTtT

emttiTw

efi(fiisuuhatc

dmtcw

iaada

i

tt

7

actdtcstdrptnne

PwouoTeacsosc

tvwppdecwtcota


f democracy. Our finding that foreign democratic capital becomesnsignificant once five-year dummies are included is also similar toesults reported by Persson and Tabellini (2009).15

Likewise, the estimated effect of the growth difference is con-istent with our theoretical predictions. In all regressions, theoefficient on the growth difference is negative and statistically sig-ificant, suggesting that other things being equal, the anticipatedrowth difference is negatively related to the probability of democ-acy breakdown given that there was no breakdown up to theurrent moment. Specifically, based on the regressions that includeve-year dummies (even-numbered columns in Tables 3a and 3b),

f the country’s own GDP growth is by one percentage point greaterhan the GDP growth in the peer group, the hazard rate is predictedo decrease by about 14 (10) percent when using the constructedPolity IV) data.

Similar to many existing studies, our estimates indicate aignificant positive relationship between per capita income andemocracy. This result does not necessarily imply causality, butay be due to certain historical factors that affect both income

nd a country’s own democratic experiences. The coefficient on theountry’s average per capita GDP growth is not statistically signifi-ant. That is, after accounting for the country’s per capita GDP andhe anticipated difference in the GDP growth, the own per capitaDP growth conveys no additional information that would help

o predict the probability of democracy breakdown. The results inables 3a and 3b also demonstrate that changing the discount fac-or from 0.94 to 0.99 has little impact on the coefficient estimates.herefore, in what follows, we report results only for ı = 0.94.

Several incidents of democracy breakdown were forced byvents associated with World War II. Because the processes thatotivated these transitions are rather unusual, we re-estimated

he models on restricted samples, excluding episodes of democracyhat ended between 1939 and 1941, inclusive. The results reportedn Table 4 are quite similar to those displayed in Tables 3a and 3b.hus, our main conclusions hold for these restricted samples asell.

Results from the regressions that account for unobserved het-rogeneity are reported in Table 5. Unfortunately, models withve-year dummies could not be estimated due to technical reasonsthe estimation did not converge). Hence, Table 5 displays the coef-cient estimates and standard errors only for models that do not

nclude five-year dummies. Moreover, it was not possible to mea-ure the goodness of fit in these models because a country-specificnobserved effect could not be “integrated out,” and predicted val-es could not be obtained. As seen in Table 5, when unobservedeterogeneity is incorporated in the analysis, the estimates remainlmost unchanged.16 In fact, in all specifications the likelihood ratioest indicates that the hypothesis of no unobserved heterogeneityannot be rejected at the 10% significance level.

As a robustness check, we also considered specifications withifferent measures of democratic capital and economic perfor-ance. We ran regressions where own democratic capital rather

han regime indices were used for constructing the peer’s demo-
ratic capital measure. Additionally, we considered specificationshere GDP measures were based on the previous year data rather
15 Foreign democratic capital variable constructed by Persson and Tabellini (2009)s most comparable with our world democratic capital measure. Similar to Perssonnd Tabellini (2009) we find that the effect of the world democratic capital is neg-tive and significant when five-year dummies are excluded (for both constructedata and Polity IV data), but it becomes insignificant once five-year dummies aredded in the regression.16 As dictated by econometric theory, in the models with unobserved heterogene-ty we use only the episodes that have complete GDP per capita and growth data.

A

fmto

dam


han on averages over the past five years. In all those regressionshe results were qualitatively unchanged.

. Conclusions

Why do some democracies survive and others fail? We develop theoretical framework where the chance of any given demo-ratic society maintaining its democratic status is determined bywo key factors: economic security and degree of attachment toemocratic values in the society. We emphasize the role of rela-ive economic performance as opposed to focusing solely on theountry’s level of economic wealth. We argue that even thoughome societies may generally be less wealthy and less inclinedo support democracy due to historical reasons, their chances foremocracy consolidation are nevertheless better if a democraticegime is expected to produce higher levels of wealth when com-ared to a non-democratic regime. This allows us to formulateestable hypothesis about the direct link between the relative eco-omic performance and the chance of democracy survival that haveot been considered in the previous theoretical and empirical lit-ratures on democracy consolidation-breakdown.

In addition to employing a well-known and commonly usedolity IV data set, we construct a new, more complete, data set thate use to test the implications of the model. The estimates that we

btain are qualitatively the same for both data sets, which givess more confidence in our results. We find that general patternsbserved in the data are consistent with our theoretical predictions.hat is, anticipated growth difference has the expected negativeffect and an increase in the democratic capital decreases the prob-bility of democracy breakdown. The country’s own democraticapital appears to have a more important impact on democracyurvival than does foreign democratic capital. Indeed, the effectf democratic experiences in peer and other countries cannot beeparated from the time-specific shocks that are common to allountries in the world.

Our findings are important as they are useful for improvinghe general understanding of the factors contributing to the sur-ival or breakdown of democracy. This in turn will help to suggestays to make existing strategies and programs more effective inolitically fragile environments, as well as to identify strategies orolicies to avoid. For example, our results indicate that stability of aemocratic regime is influenced not only by political and economicnvironment within the country and political climate in otherountries, but also by economic performance in foreign countries,here the latter provides a reference point for identifying whether

he current democratic regime is economically successful. Currentonditions in the Middle East remind us about the relative fragilityf many young democratic regimes, which underscores the impor-ance of this topic. Future research could also focus on transitionsway from non-democracy to democracy.

ppendix A. Regime classification

We use the classification of political regimes that is based onree and fair elections (see Noguera, 2007a for details). By free we

ean that elections are universal, i.e. all individuals have the righto vote and run for office. By fair elections, we mean that electionsccur without falsifying results or voter intimidation.

The employed classification includes three fully or partiallyemocratic regimes (democracy, semi-democracy, and oligarchy)nd three non-democratic regimes (autocracy, bureaucracy &onarchy, and semi-autocracy).

and E

D

totsc

S

ohrt

O

artt

A

cbtEA

M

&ttlC(e

S

tCffu

cd

A

BNa

a

Hd

Gs

IQa

CGP

sPH

N

pT

BCbGMNSlR

FbS

R

A

B

B

B

B

B

B

B

C

C

D

D


emocracy

We classify a political regime as democracy if there is no restric-ion on competition (no political group can impede or createbstacles for other groups to compete in elections) and no restric-ion on participation/inclusion (i.e. political participation of allectors of the society is allowed, so that all members of the societyan politically organize or join political parties).

emi-democracy

In semi-democracies, competition is fair but an important sharef population is excluded from participation; that is, they do notave the right to vote. Typical examples are countries that enjoy aather free political environment, but women do not have the righto vote.

ligarchy

In oligarchy (or competitive oligarchy), participation reduces to small group (for instance, wealthy men). An example of such aegime would be the white predominance in South Africa duringhe Apartheid. A distinct characteristic of this political regime is thathe competition among those who are allowed to participate is fair.

utocracy

Autocracy is a political regime that severely restricts politi-al competition, so that it is not possible to alternate the poweretween different political actors. This political regime is charac-erized by absence of elections and despotic rule of one individual.xamples include Nazi Germany, as well as the rule of Generalugusto Pinochet in Chile and General Francisco Franco in Spain.

onarchy & Bureaucracy

Similar to autocracy, competition is inexistent in monarchy bureaucracy (e.g. contemporary Saudi Arabia or France before

he French revolution). The main difference between this regimeype and autocracy is that in autocracy, the dictator gains abso-ute power either through a coup (e.g. General Augusto Pinochet inhile, General Francisco Franco in Spain) or by means of electionsNazi Germany), while in monarchy & bureaucracy the power isither inherited or delegated to an appointed official.

emi-autocracy

In semi-autocracies elections are held, but either the competi-ion is severely restricted (e.g. the former Soviet Union, where theommunist party selected political candidates), or elections are not

air (several parties can participate, but the ruling party is able toalsify election results, so that opposition parties cannot competender acceptably fair conditions).

We code these regimes as follows: autocracy = 1, bureau-racy & monarchy = 2, semi-autocracy = 4, oligarchy = 5, semi-emocracy = 6, and democracy = 7.

ppendix B. Reference groups/‘peer’ countries

Caribbean: Antigua and Barbuda; Bahamas; Barbados; Belize;ermuda; Dominica; Grenada; Guyana; Jamaica; Saint Kitts and
evis; Saint Lucia; Saint Vincent and the Grenadines; Suriname;nd Trinidad and Tobago.
East Asia: China; Japan; Mongolia; North Korea; South Korea;nd Taiwan.

E


Former Socialist Central & Eastern Europe: Albania; Bosnia-erzegovina; Bulgaria; Croatia; Czech Republic; Hungary; Mace-onia; Poland; Romania; Serbia; Slovakia; and Slovenia.

Former Soviet Union Armenia; Azerbaijan; Belarus; Estonia;eorgia; Kazakhstan; Kyrgyzstan; Latvia; Lithuania; Moldova; Rus-ia; Tajikistan; Turkmenistan; Ukraine; and Uzbekistan.

Islamic Countries Afghanistan; Algeria; Bahrain; Egypt; Iran;raq; Jordan; Kuwait; Lebanon; Libya; Morocco; Oman; Palestine;atar; Saudi Arabia; Syria; Tunisia; Turkey; United Arab Emirates;nd Yemen United.

Latin America: Argentina; Bolivia; Brazil; Chile; Colombia;osta Rica; Cuba; Dominican Republic; Ecuador; El Salvador;uatemala; Haiti; Honduras; Mexico; Nicaragua; Panama;araguay; Peru; Uruguay; and Venezuela.

North America Canada and the United States of America.Oceania Australia; Fiji; Kiribati & Tubalu Gilbert Islands; Mar-

hall Islands; Federated States of Micronesia; New Zealand; Palau;apua New Guinea; Samoa; Solomon Islands; and Vanuatu Newebrides.

South Asia: Bangladesh; Bhutan; India; Maldives Islands;epal; Pakistan; and Sri Lanka.

Southeast Asia: Brunei; Burma Myanmar; Cambodia Kam-uchea Khmer; Indonesia; Laos; Malaysia; Philippines; Singapore;hailand; and Vietnam.

Sub-Saharan Africa: Angola; Benin; Botswana; Burkina Faso;urundi; Cameroon; Cape Verde Islands; Central African Republic;had; Comoros Islands; Republic of Congo; Cote D’Ivoire; Dji-outi; Equatorial Guinea; Eritrea; Ethiopia; Gabon; Gambia; Ghana;uinea; Guinea Bissau; Kenya; Lesotho; Liberia; Madagascar;alawi; Mali; Mauritania; Mauritius; Mozembique; Namibia;iger; Nigeria; Rwanda; Sao Tome and Principe; Senegal;eychelles; Sierra Leone; Somalia; South Africa; Sudan; Swazi-and; Tanzania; Togo; Tonga; Uganda; Zaire (Congo, Democraticepublic); Zambia; and Zimbabwe.

Western Europe: Austria; Belgium; Cyprus; Denmark; Finland;rance; Germany; Greece; Iceland; Ireland; Israel; Italy; Luxem-ourg; Malta and Gozo; Netherlands; Norway; Portugal; Spain;weden; Switzerland; and United Kingdom.

eferences

cemoglu, Daron, Johnson Simon, Robinson James and Yared Pierre, 2008. “Incomeand democracy.”. American Economic Review 98(3), 808–842.

apat, R.B. and M.I. Beg, 1989. “Order statistics from nonidentically distributedvariables and permanents.”. Sankhya Ser. A 51, 79–93.

arro, Robert J., 1999. “Determinants of democracy.”. Journal of Political Economy107(S6), S158–S183.

enhabib, Jess and Adam Przeworski, 2006. “The political economy of redistributionunder democracy.”. Economic Theory 29, 271–290.

ernhard, Michael, Timothy Nordstrom and Christopher Reenock, 2001. “Economicperformance, institutional intermediation and democratic survival.”. Journal ofPolitics 63(3 (August)), 775–803.

ernhard, Michael, Christopher Reenock and Timothy Nordstrom, 2003. “Economicperformance and survival in new democracies: is there a honeymoon effect?”.Comparative Political Economy 36(4 (May)), 404–431.

rinks, Daniel and Michael Coppedge, 2006. “Diffusion is no illusion – neighbor emu-lation in the third wave of democracy.”. Comparative Political Studies 39(May(4)), 463–489.

unce, Valerie, 2001. “Democratization and economic reform.”. Annual Review ofPolitical Science 4, 43–65.

heibub, Jose Anotonio, Jennifer Gandhi and James Raymond Vreeland, 2010.“Democracy and dictatorship revisited.”. Public Choice 143, 67–101.

lark, A., N. Kristensen and N. Westergard-Nielsen, 2009. “Job satisfaction and co-worker wages: status or signal?”. Economic Journal 119, 430–447.

eHaan, Jakob and Clemens Siermann, 1995. “New evidence on the relationshipbetween democracy and economic growth.”. Public Choice 86, 175–198.

iskin, Abraham, Hanna Diskin and Reuven Y. Hazan, 2005. “Why democracies col-
lapse: the reasons for democratic failure and success.”. International PoliticalScience Review 26(July (3)), 291–309.
pstein, David L., Robert Bates, Jack Goldstone, Ida Kristensen and Sharyn O’Halloran,2006. “Democratic transitions.”. American Journal of Political Science 50(July(3)), 551–569.

http://refhub.elsevier.com/S2214-8043(14)00023-8/sbref0005
































































































































































































2 and E

F

G

G

J

L

M

M

N

N

P

P

P

PP

P

P

P

R

R

S


oley, S., D.L. Kidder and G.N. Powell, 2002. “The perceived glass ceiling and justiceperceptions: an investigation of Hispanic law associates.”. Journal of Manage-ment 28(4), 471–496.

asiorowski, Mark, 1993. The Political Regime Change Data Set. Baton Rouge:Louisiana State University.

oldstone, Jack A., Robert H. Bates, David L. Epstein, Ted Robert Gurr, Michael B.Lustik, Monty G. Marshall, Jay Ulfelder and Mark Woodward, 2010. “A globalmodel for forecasting political instability.”. American Journal of Political Science54(1), 190–208.

enkins, Stephen, 2005. Survival Analysis. Colchester, UK: Institute forSocial and Economic Research, University of Essex, Available from:http://www.iser.essex.ac.uk/survival-analysis (unpublished manuscript).

ipset, Seymour M., 1959. “Some social requisites of democracy: economic devel-opment and political legitimacy.”. American Political Science Review 53,69–105.

ajor, B., 1994. “From social inequality to personal entitlement – the role of socialcomparisons, legitimacy appraisals, and group membership.”. In Zanna, M.P.(Ed.), Advances in Experimental Social Psychology, vol. 26, 293–355. San Diego,CA: Academic Press Inc.

ansfield, Edward D. and Jon C. Pevehouse, 2006. “Democratization and interna-tional organizations.”. International Organization 60(Winter), 137–167.

oguera, Jose, 2007a. Participation, Competition and Political Regimes. Departmentof Economics, University of New Hampshire, Working Paper (November), Avail-able from: http://papers.ssrn.com/sol3/papers.cfm?abstract id=1139423

oguera, Jose, 2007b. Democratic Stability and Breakdown. Department of Eco-
nomics, University of New Hampshire, Working Paper (February), Availablefrom: http://papers.ssrn.com/sol3/papers.cfm?abstract id=1139934
effley, Mark and Robert Rohrschneider, 2003. “Democratization and political tol-erance in seventeen countries: a multi-level model of democratic learning.”.Political Research Quarterly 56(September (3)), 243–257.

T

W


ersson, Torsten and Guido Tabellini, 2009. “Democratic capital: the nexus of politi-cal and economic change.”. American Economic Journal: Macroeconomics 1(July(2)), 88–126.

evehouse, Jon C., 2002. “With a little help from my friends? Regional organiza-tions and the consolidation of democracy.”. American Journal of Political Science46(July (3)), 611–626.

otrafke, Niklas, 2012. “Islam and democracy.”. Public Choice 151(1–2), 185–192.rzeworski, Adam and Fernando Limongi, 1993. “Political regimes and economic

growth.”. Journal of Economic Perspectives 7, 51–69.rzeworski, Adam and Fernando Limongi, 1997. “Modernization: theories and

facts.”. World Politics 49(2), 155–183.rzeworski, Adam, Michael Alvarez, Jose Antonio Cheibub and Fernando Limongi,

1996. “What makes democracies endure?”. Journal of Democracy 7(January (1)),39–55.

rzeworski, Adam, Michael Alvarez, Jose Antonio Cheibub and Fernando Limongi,2000. Democracy and Development: Political Institutions and Well-Being in theWorld, 1950–1990. New York: Cambridge University Press.

arakat, H.M. and Y.H. Abdelkader, 2004. “Computing the moments of order statis-tics from nonidentical random variables.”. Statistical Methods and Applications13, 13–24.

osefielde, Steven, 2003. “The riddle of Postwar Russian economic growth: statisticslied and were misconstrued.”. Europe-Asia Studies 55(3), 469–481.

ullivan, John L. and John E. Transue, 1999. “The psychological underpinnings ofdemocracy: a selective review of research on political tolerance, interpersonaltrust, and social capital.”. Annual Review of Psychology 50, 625–650.

reisman, Daniel, 2010. “Twenty years of political transition.”. In Gérard, Roland(Ed.), Economies in Transition: The Long-Run View, 109–133. Palgrave Macmil-lan.

ooldridge, Jeffrey, 2002. Econometric Analysis of Cross Section and Panel Data.Cambridge: MIT Press.

















































http://www.iser.essex.ac.uk/survival-analysis




































































http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1139423

http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1139934


































































































































































































































economic security and democratic capital: why do some democracies survive and others fail?

Documents