This article was downloaded by: [Dae Jin Yi]On: 18 March 2013, At: 21:12Publisher: RoutledgeInforma Ltd Registered in England and Wales Registered Number: 1072954 Registeredoffice: Mortimer House, 37-41 Mortimer Street, London W1T 3JH, UK

Journal of Economic Policy ReformPublication details, including instructions for authors andsubscription information:http://www.tandfonline.com/loi/gpre20

No taxation, no democracy? Taxation,income inequality, and democracyDae Jin Yi aa Hankuk University of Foreign Studies, Seoul, Korea (ROK)Version of record first published: 03 May 2012.

To cite this article: Dae Jin Yi (2012): No taxation, no democracy? Taxation, income inequality, anddemocracy, Journal of Economic Policy Reform, 15:2, 71-92

To link to this article: http://dx.doi.org/10.1080/17487870.2012.672252

PLEASE SCROLL DOWN FOR ARTICLE

Full terms and conditions of use: http://www.tandfonline.com/page/terms-and-conditions

This article may be used for research, teaching, and private study purposes. Anysubstantial or systematic reproduction, redistribution, reselling, loan, sub-licensing,systematic supply, or distribution in any form to anyone is expressly forbidden.

The publisher does not give any warranty express or implied or make any representationthat the contents will be complete or accurate or up to date. The accuracy of anyinstructions, formulae, and drug doses should be independently verified with primarysources. The publisher shall not be liable for any loss, actions, claims, proceedings,demand, or costs or damages whatsoever or howsoever caused arising directly orindirectly in connection with or arising out of the use of this material.

No taxation, no democracy? Taxation, income inequality, anddemocracy

Dae Jin Yi*

Hankuk University of Foreign Studies, Seoul, Korea (ROK)

Does taxation promote democracy? Revisiting this question, I hypothesize that theeffects of taxation on democracy tend to be relatively stronger in unequal societiesbecause higher income inequality can amplify the extent to which citizens dissatisfiedwith higher levels of taxation want to soak elites. Using event history models to ana-lyze a pooled time-series dataset of regime transitions that cover all countries from1970 to 2000 if data are applicable, I find empirical evidence that taxation has a con-ditional impact on democratization, but not on democratic breakdown. According tothe theory, higher taxation levels and greater income inequality should tend to promotedemocracy.

Keywords: democratization; taxation; income inequality; event history analysis

1. Introduction

Does taxation promote democracy? Do levels of taxation and distribution of tax burdenslead to different types of political outcomes? It is widely accepted that taxation tends topromote the emergence of representative institutions. This theoretical argument has beendeveloped from two different fields of democratization; one is a historical interpretationof state-building in early modern Europe and colonial America (e.g., Tilly 1990, 2004)and another, the so-called rentier state theory, is a qualitative exploration of the MiddleEast and North Africa (hereafter MENA), where authoritarian regimes still remain strongeven in the era of the third wave of democratization (e.g., Ross 2001).

Although these ideas are well developed theoretically, few statistically systematicstudies have attempted to account for the relationship between taxation and democracy.The only quantitative models to test the nexus between taxation and democracy havebeen recently developed by Ross (2004) and Herb (2005). Yet, not only are their empiri-cal findings inconclusive, but they also share the same methodological problem, modelmisspecification, which will be discussed later.

With the insufficiency of the empirical literature on the effects of taxation in mind, Iwish to emphasize that the claim that a higher level of taxation is associated with morerepresentative systems should be revisited. My analysis is distinctive, I believe, on bothmethodological and theoretical grounds. First, unlike empirical studies mentioned earlier,to fit the data nature of democratic transition and my theoretical purposes, I employedevent history analysis rather than the ordinary least squares regression method (OLS). Ifthere are good substantive reasons to think that some variables of interest influence

*Email: dyi@hufs.ac.kr

Journal of Economic Policy ReformVol. 15, No. 2, June 2012, 71–92

ISSN 1748-7870 print/ISSN 1748-7889 online� 2012 Taylor & Francishttp://dx.doi.org/10.1080/17487870.2012.672252http://www.tandfonline.com

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

democratization and democratic breakdown in different ways, statistical models designedby Ross (2004) and Herb (2005) may not be appropriate because estimates of those mod-els attempt to account for those transitions simultaneously. Yet, the event history modelspresented here enabled me to distinguish between the effects of variables of interest ondemocratization (the survival of non-democracy) and their effects on democratic break-down (the survival of democracy).

More important, my theoretical contribution is to attempt to refine such arguments bypointing out the ways in which taxation may bolster democracy. The starting point forthe analysis is that taxation differences do affect a democracy. But I argue that these taxa-tion differences will be influential only when some economic conditions are in place.More specifically, I hypothesize that the effects of taxation on democracy tend to be rela-tively stronger in societies with more inegalitarian structures of income distributionbecause higher income inequality can amplify the extent and depth of dissatisfactioncitizens derive from higher levels of taxation. Previewing my empirical findings, the evi-dence presented here strongly helps to buttress a key theoretical hypothesis, indicatingthat taxation has a conditional impact on democratization (but not the persistence of dem-ocratic regimes). According to the theory, higher taxation levels and greater incomeinequality should tend to promote democracy.

The following section reviews the relevant literature on democratization theory, focus-ing on the effects of taxation. After reviewing previous research, I offer my key hypothe-sis, a summary of the data used in my analysis, and my statistical technique: eventhistory models. I then present my own empirical findings from pooled time-series data ofregime transitions that cover all countries from 1970 to 2000 if data are applicable. Thelast section concludes and proposes new agendas for future research.

2. Review and hypotheses

2.1. Taxation and democracy

The hypothesis that taxation tends to promote democracy is based on one strand of fiscalsociology that explores the dynamics of state-building in early modern Europe and colo-nial America (Ardant 1975, Braun 1975, Ames and Rapp 1977, Poggi 1978; and see,particularly, North and Weingast 1989 on England). The historical facts on which schol-ars have focused are that monarchies in England, France, Spain, and Austro-Hungary hadto give up some of their authority to representative systems in exchange for the ability toraise new taxes. The point on which theories associating taxation with the advent of dem-ocratic political systems focus is the bargaining between political elites who want to raisethe money for warfare and citizens who want to wield certain political rights. Tilly(1990, 2004), in his study of middle and modern states in Europe, points out that, in thebargaining with the states that were in pursuit of revenues, citizens demanded moreaccountability from them and some influence over how their taxes were spent inexchange for new taxes.1

The other strand is the rentier state theory, which is one of the tools to explain therobustness of authoritarian regimes in the MENA region. The positive relationshipbetween resource rents and the persistence of authoritarian regimes is the nuts and boltsof the literature on the rentier state theory (Beblawi 1987, Crystal 1990, Luciani 1994,Anderson 1995). The specific mechanisms through which rents might promote the dura-bility of authoritarianism can be summarized in two categories. If we think of citizens asconsumers and governments as suppliers, the first mechanism might be called an effect oflow demand.2 It suggests that when governments acquire the exceptional profits from nat-

72 D. Jin Yi

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

ural resource extraction, ‘they are likely to tax their populations less heavily or not at all,and the public in turn will be less likely to demand accountability from – and representa-tion in – their government’ (Ross 2001, 332). Lisa Anderson (1995) argues that soft bud-get constraints resulting from external oil rents have left conditions in the regionprofoundly inauspicious for democracy.

The second is through what might be dubbed ‘an effect of high supply’. ‘Windfall’(Beblawi 1987) or ‘manna from heaven’ (Dunning 2008) derived from the sale of mineralresources tends to enables rentier governments to provide a wide range of free services tocitizens, such as education, health care, and housing; therefore, these structures of patron-age and the rentier economy have weakened the economic basis of potential oppositiongroups (Gause 1993). In theory, the contrast between these two types of mechanisms issomewhat clear-cut, but, in practice, they influence each other. Each tends to be the otherside of the coin. The democratic effect of taxation will be relatively stronger if citizensare dissatisfied with fewer government benefits. In this context, whether citizens demanddemocracy is a function of two responses to the increase of taxes and the decrease ofgovernment expenditure (government service). This logic also matches the findings ofpublic opinion studies, which suggest that the complaints of citizens regarding their gov-ernment’s low provision of public goods lead to political protest (Finkel and Muller1998, Muller et al. 1991). Thus, when we explicate and test a corollary of the rentierstate thesis, ‘no taxation, no democracy,’ we should consider the size of the tax burdenrelative to government services.

It seems there is a grain of truth in this argument, but relatively little scholarly atten-tion pertaining to testing its validity and generality has been devoted to synthesizing sys-tematic improvements in the field. As I show, samples selected for advocating thedemocratic effect of taxation have been limited in terms of historical and geographicalaspects. Historically, only a few early modern Western countries have been selected asthe examples of democracy facilitated by high level of taxation. Geographically, thehypothesis about the relationship between taxation and democracy has been isolatedwithin the MENA specialists. Most of the researchers on democratization have ignoredthe riddle of the MENA democracy deficit and, broadly, ‘no taxation, no democracy’ hasexisted as ‘a theoretical enclave’ for years.

Although some quantitative models to test the nexus between taxation and democracyhave been recently developed by Ross (2004) and Herb (2005), not only are their empiri-cal findings inconclusive, but they also share the same methodological problem, that is,model misspecification. They both draw their conclusion from the OLS method, con-structing somewhat different dependent variables from the Polity 98 dataset of Gurr andJaggers (1999) or from Freedom House’s democracy score, respectively. Yet, this widelyaccepted practice involves some serious systematic problems. Linear regression analysismakes a big assumption: The distances between adjacent categories are equal. But thePolity and Freedom House’s democracy datasets were originally ordinal variables.3 If the

Table 1. Matrix of regime transitions, 1970-2000.

Number of transitions to

From ND PD D Total (transitions only)

Non-democracy A (46) B 69 C 13 82 DemocratizationPartial Democracy D 36 Backsliding E (8) F 31 Consolidation 67Democracy G 11 H 11 I (28) 22 Democratic Breakdown

Journal of Economic Policy Reform 73

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

implicit assumption that the intervals between adjacent categories are the same is false,then a linear regression model of the ordinal dependent variable produces the biased coef-ficients and misleading results.4 Moreover, their empirical findings do not strongly sup-port the democratic effect of taxation.

My empirical result from the event history analysis, Model 1 in Table 2, also showsthat taxation does not significantly influence democratic transition. In these specifications,the coefficients for taxation are positive (as expected, higher levels of taxation are associ-ated with democratization), but they do not even approach statistical significance. Theseresults therefore offer little support for the hypothesis ‘no taxation, no democracy.’ Thiscounter-theoretical evidence raises some possibility that will require further elaboration.One possibility is that the political consequences of taxation may not be constant acrossthe countries, regardless of their economic conditions, especially income inequality, whichmay condition the effect of taxation on democracy. It may be reasonable to infer that theworse the system of income distribution becomes, the easier citizens lose their temperwith government’s acts to increase the tax burden relative to government services. There-fore, the analysis needs to be extended to income inequality and to employ an improvedmethodology. More systematic tests of the relationships among taxation, inequality, anddemocracy should be performed. These are the objectives of the analysis that follows.

2.2. The argument

With the insufficiency of the literature on the effects of taxation in mind, theoretically,this paper systematically tests a hypothesis that derives the events called democratizationand democratic breakdowns from a set of assumptions, arguing that the level of taxationin a society, together with the degree of the distribution of income, shapes underlyingpreferences for democracy. Before I address my key hypothesis further, I briefly reviewthe literature on the effect of inequality on democracy.

Academic scholars committed to qualitative evidence seem to agree that relatively equalsocieties foster a democratic transition (Dahl 1971, de Tocqueville 1990, Huntington1991).5 Contrary to the intellectual history mentioned above, Acemoglu and Robinson(2006), synthesizing their two earlier pieces (2000, 2001), address an inverted U-shapedrelationship between inequality and the likelihood of democratic transition. In fairly equalsocieties, a revolution becomes more costly for citizens because they already benefit fromthe current structure of the economy, so they do not claim any further rewards that may beoffered by the democratic rules. In contrast, in fairly unequal societies, albeit that it ishighly reasonable to predict an intensification of the redistributive demand of citizens onthe authority of non-democracy, the cost of redistributive taxation to elites in democraticpolitics and, accordingly, their animosity against democracy, should be higher. Hence,democratization may be less likely to happen in unequal societies. The researchers suggestthat democratic transitions are more likely to occur as societies are located somewherebetween the highest and lowest level of inequality, i.e., in the middle levels of inequality.

This literature on the effect of inequality on democratization parallels to some extentthat on the effect of taxation on democratization; although qualitative and analytical stud-ies (except for Acemoglu and Robinson 2000, 2006) seem to agree that equal authoritar-ian societies are more likely to transition to democracy, the empirical results arecontroversial.6 Probably the most serious problem with the current theories that linkinequality to democratization is the extent to which causal mechanisms are chosen on thegrounds of theoretical convenience. As Houle (2009, p. 593) points out, inequality has‘two opposite … effects on democratization’ by influencing the preferences of elites and

74 D. Jin Yi

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

Table2.

Determinantsof

democratizationanddemocratic

breakdow

n,1970–2000.

Cox

analysisa

Dem

ocratization

Dem

ocratic

breakdow

n

Model

1Model

2Model

3Model

4Model

5Model

6

Tax/Exp

0.382

0.070

�0.454

0.647

(0.441)

(0.425)

(0.416)

(0.829)

Inequality

0.008

0.009

0.027⁄

⁄⁄0.038⁄

⁄(0.008)

(0.010)

(0.008)

(0.015)

Inequality2

�0.0008⁄

–0.001

⁄(0.0004)

(0.0005)

Tax/Exp

×Inequality

0.074⁄

–0.085

(0.033)

(0.068)

GDPpercapita

(log)

0.171

0.252⁄

0.192

–0.453

⁄⁄–0

.466

⁄⁄⁄

–0.617

⁄⁄(0.142)

(0.111)

(0.150)

(0.151)

(0.102)

(0.211)

Growth

Rate

0.012

�0.039

�0.040

�0.034

0.013

�0.037

(0.035)

(0.029)

(0.032)

(0.036)

(0.032)

(0.054)

Trade

(log)

0.008

0.138

0.178

0.213

�0.035

0.131

(0.140)

(0.128)

(0.159)

(0.275)

(0.158)

(0.320)

Catholic

�0.002

�0.0005

�0.002

�0.0004

�0.004

�0.011

⁄(0.002)

(0.002)

(0.002)

(0.005)

(0.003)

(0.005)

Muslim

�0.018

⁄⁄�0

.013

⁄⁄�0

.015

⁄0.003

0.002

0.001

(0.007)

(0.004)

(0.007)

(0.004)

(0.002)

(0.003)

ELF

0.053

0.363

0.082

0.345

�0.443

0.481

(0.291)

(0.278)

(0.295)

(0.433)

(0.351)

(0.710)

Resource

�0.017

�0.018

�0.013

0.009

0.015⁄

⁄⁄0.015

(0.017)

(0.012)

(0.015)

(0.009)

(0.004)

(0.020)

BritishColony

0.375⁄

0.192

0.281^

�0.705

⁄�0

.639

⁄⁄⁄

�1.266

⁄⁄⁄

(0.151)

(0.139)

(0.147)

(0.350)

(0.201)

(0.312)

RegionaltotalD

1.176⁄

⁄1.291⁄

⁄⁄0.907⁄

⁄(0.432)

(0.366)

(0.297)

Regionaldemocracy

�3.276

⁄⁄⁄

�2.368

⁄⁄⁄

�2.410

⁄⁄⁄

(0.569)

(0.432)

(0.627)

PreviousD

Transitions

�0.163

�0.066

�0.182

(Contin

ued)

Journal of Economic Policy Reform 75

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

Table2.

(Contin

ued).

Cox

analysisa

Dem

ocratization

Dem

ocratic

breakdow

n

Model

1Model

2Model

3Model

4Model

5Model

6

(0.125)

(0.093)

(0.127)

PreviousD

Breakdown

0.731⁄

0.404⁄

0.817⁄

⁄(0.311)

(0.175)

(0.295)

Observatio

ns1564

2283

1140

1564

2283

1140

Countries

84130

7584

130

75Log

Likelihood

�3586.063

�5768.626

�2711.382

�2010.814

�3919.671

�1046.259

Chi-Square

110.49

194.95

73.64

179.23

257.01

155.12

Note:

Table

entriesarecoefficients.Rob

uststandard

errors

with

clustering

onthecoun

tryarecontainedin

parentheses.Prob>chi2

=0.00

00forallmod

els.

^p60.10

;⁄ p

60.05

;⁄⁄p60.01

;⁄⁄

⁄ p60.00

1.a N

otethat

Cox

mod

elsdo

notcontainaconstant.Itisabsorbed

into

thebaselin

ehazard.

76 D. Jin Yi

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

citizens in a contradictory way; inequality makes democracy less attractive to elites byincreasing potential costs from redistribution, but more appealing to citizens by increasingfuture booty from soaking elites. Despite these conflicting effects of inequality on democ-ratization, scholars often pick up only one of the two opposite effects to bolster their the-oretical inferences.7

The theoretical inference that inequality has competing effects on democratizationdoes not mean that inequality, in practice, has no significant impact on it. It is entirelygratuitous to assert that two conflicting effects mainly find equilibrium by offsetting eachother. Various factors can influence the balance of power between inequality’s twoeffects, such that one dominates in a particular instance. In this analysis, I argue that athigher levels of taxation, as citizens’ dissatisfaction grows, the costs to elites of repress-ing revolts will surpass the costs of redistribution, in which democratization is morelikely. Therefore, if I put my key hypothesis in another – but much similar – way, taxa-tion conditions the effect of inequality on democratization.

Returning to my starting point, the analysis here privileges the economic condition ofpolitical regimes, not because these always play significant roles, but rather because thepolitical effects of taxation should certainly depend on such an economic factor. Themodels suggest that the democratic effect of taxation is contingent upon the level ofincome inequality in a given society. At low levels of inequality, as income tends to bedistributed in a more balanced way among individuals, the burden of taxation would bebearable to citizens and the effect of taxation on the likelihood of democratic transitionwould be minor.

In contrast, in highly unequal societies, if elites do not intend to endure the potentiallygreater tax revenue of future democracy, the political outcome assumed is the status quoand the fiscal outcome predicted is the lower tax system to abbreviate the redistributivedemand of citizens, like most of the authoritarian regimes in MENA. However, a highlevel of taxation is more likely to induce democratization in societies with a high level ofinequality because the combination of high levels of taxation and highly unequal eco-nomic structure may aggravate the extent to which citizens want to soak elites. In thiscontext, the cost to elites of repression becomes higher than the taxes that will beimposed on them. It is the credible threat of a revolution by citizens that eventuallyinduce elites to accept democracy. In sum, my testable hypothesis emerging from my the-oretical argument is that income inequality enlarges the democratic effect of taxation; inparticular, the democratic effect of taxation may be stronger in societies with highdegrees of income inequality, but it may not the case in relatively equal countries.

3. Variables and data

3.1. Regime types

Among the hot issues that show sharp disagreement in comparative research on demo-cratic regimes is the choice of measures of democracy: dichotomous and gradedapproaches (Collier and Adcock 1999, Munck and Verkuilen 2002). In this study, follow-ing Przeworski et al. (2000) and Boix (2003), I apply a categorical measure of democ-racy based on a distinction between authoritarian and democratic regimes. Decidingwhich logical treatment of democracy should be appropriate for what purpose (Collierand Adcock 1999). The theoretical rationale is obvious. Operationalization of categoricalregime types allows me to investigate whether taxation reveals significant differences interms of an effect on the regime transformation, hinging upon the degrees of income dis-

Journal of Economic Policy Reform 77

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

tribution among individuals. The methodological rationale is even clearer. It is fairlycomplicated to analyze and hard to interpret ordered ranked outcomes that have manycategories, such as the Polity dataset (a 21-point scale) or the Freedom House’s democ-racy scores (a 7-point scale), through the ordered choice models.

Yet, to sketch the world as either democratic or authoritarian may ignore the possibil-ity of an intermediate case. Both Huntington (1996) and Diamond (1996) point out that agrowing number of countries exist ‘somewhere in the middle’ of the ‘democratic–non-democratic continuum’ (Huntington 1996, p. 10). In addition, some scholars note that thecountries belonging to an intermediate category show different patterns of political andsocioeconomic dynamics from either full democracies or full authoritarian regimes(Mansfield and Snyder 1995, Goldstone 2000, Zakaria 2003).

To address these theoretical concerns, I employ a trichotomous measure of democrati-zation: Non-Democracy (ND), Partial Democracy (PD), and Democracy (D). Empirically,to define a democracy in the panel sample from 1970 to 2000, I divide regimes, thedependent variable, relying on the Polity IV dataset (Marshall and Jaggers 2007), intonon-democracy (Polity2 value –10 to 0), partial democracies (+1 to +7), or democracies(+8 to +10). The reasoning for my choice of 8 and above for Democracy is offered byPolity IV. In the dataset, 8 points is a threshold for a ‘mature and internally coherentdemocracy,’ which satisfies the following conditions: (a) ‘fully competitive’ political par-ticipation, (b) ‘elective’ executive recruitment, and (c) ‘substantial’ constraints on thechief executive (Marshall and Jaggers 2007, Dataset users’ manual).

Table 1 presents the dynamic of transition from one regime category to another. Itshows the number of regime transitions that occurred in all countries during 1970-2000.The entries in parentheses in the diagonal cells indicate very stable countries: 46 stablenon-democracies (cell A), eight stable partial democracies (cell E), and 28 stable democ-racies (cell I). The numbers in the off-diagonal cells denote 171 regime transitions of thesix possible kinds.

The table reveals that democratization (autocratic breakdown) accounts for the lion’sshare of regime transitions, which have occurred roughly four times more than demo-cratic breakdown. In my operationalization, democratization includes the transformationof non-democracy into partial democracy or democracy (cells B and C, total 82), anddemocratic breakdown implies the transformation of democracy into partial democracy ornon-democracy (cells G and H, total 22). Almost half of partial democracies move intodemocracy (democratic consolidation,8 cell F), and the other half move back into non-democracy (backsliding, cell D).

3.2. Taxation and income inequality

To fit the theoretical hypothesis above and to combine the effects of low demand andhigh supply, Tax/Exp is measured by total tax revenues as a percentage of total gov-ernment expenditure, which covers all current and capital expenditures, including inter-est payments on past debts. Its theoretical implication is that citizens weigh the costsof funding the government against the benefits they receive. The data are from WorldTax Database at the University of Michigan (http://www.bus.umich.edu/otpr/otpr/default.asp).9

This measure tests a ‘cost-benefit model’ (Ross 2004), of which theoretical implica-tion is that citizens weigh the costs of funding the government against the benefits theyreceive. Another possible way to catch levels of taxation is Tax/GDP, total tax revenuesas a share of GDP. This measure tests a ‘pure anti-tax model’ (Ross 2004), of which the-

78 D. Jin Yi

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

oretical speculation is that the increase in taxes in non-democracies tends to simplyincrease the demand for democracy. As addressed previously, following the logic of pub-lic opinion studies (Muller et al. 1991, Finkel and Muller 1998), I expect that Tax/Expmay be a better measure to fit nicely with the theoretical hypothesis rather than Tax/GDP.Strictly speaking, whether citizens demand democracy is a function of two responses tothe increase of taxes and the decrease of government expenditure (government service) ina cost-benefit model (Tax/Exp), whereas it is a function of only a single response to theincrease of taxes in a pure anti-tax model (Tax/GDP).

The variable of income inequality comes from the Gini variable in the World IncomeInequality Database (UNU-WIDER 2007). This is by far the most comprehensive Ginidataset, including 4981 observations from 152 countries.10 Because the dataset is anaccumulation of various surveys with different methods, implying that all the data are notof the same quality, and some countries have more than one observation for certain years,I developed the following selection criteria to trim down the data.

First, I choose data produced with either clear income concepts or quality surveys,which are coded as 1 or 2 for the variable Quality in the dataset. I also drop any data thatdo not cover all areas, all population groups, and all age groups in those populationgroups (see AreaCovr, PopCovr, and AgeCovr in the users’ guide). Then I choose datawhose income-sharing units, IncSharU, are family or household, not individual. This isbecause I believe inequality is better measured by family or household incomes than byindividual incomes. Another thing to be considered here is whether data have any equiva-lence scales, Equivsc, to take different family or household sizes into account. Finally, tomagnify the number of observations and to smooth out annual anomalies, I employ afive-year average of adjusted Gini coefficients.

3.3. Control variables

In addition to the hypothesized variables, I incorporate two economic variables that arecommonly theorized as the factors to affect regime transition. The first is the level of eco-nomic development (Lipset 1959, Przeworski et al. 2000, Boix and Stokes 2003), mea-sured as the natural log of GDP per capita (in constant dollars, Chain Index, expressedin international prices, base 2000), taken from the Penn World Tables. As the second eco-nomic variable, I include rates of economic growth, Growth Rate, which is oftenexpected to have a positive impact on the longevity of regimes (Gasiorowski 1995, Hag-gard and Kaufman 1995, Przeworski et al. 2000). I adopt an additional method, thesquare root of the two-year prior moving average of a country’s annual rate of growthrate, to weaken the weights of annual eccentricity and higher values.

Consistent with conventions in the field, I also control for some social and culturalfactors that have been found to influence the transition to and the survival of democra-cies. The first is the ethnolinguistic fractionalization index, ELF, which is commonly the-orized as a contributor to democratic breakdown (Muller and Seligson 1994). Data comefrom Roeder (2001). As the cultural variables, I include Catholic and Muslim, measuredas the percentages of a country’s population that are Catholic and Muslim. Previousresearchers have found evidence that democracy is less likely to flourish in countries withlarge Muslim populations in that Islam emphasizes authority and community over anindividual’s right and freedom (Lipset 1994, Barro 1999, Fish 2002). In a similar vein,some studies have noted that there is a negative correlation between Catholic populationand democracy (Huntington 1991, Lipset 1994, Cheibub 1998).

Journal of Economic Policy Reform 79

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

I also construct a variable, Resource, to account for the role of natural resources inregime transition, which is measured as the share of a country’s gross national incomederived from the depletion of energy resources (petroleum, natural gas, and coal) (seeBolt et al. 2002). This variable captures the theorem of the ‘resource curse’ (Ross 1999),in which the lion’s share of national income from extractable natural resources is consid-ered a boon for the prospects for authoritarian rule (Ross 2001, Smith 2004, Ulfelder2007). Furthermore, another dummy variable is added for countries that had experiencedBritish colonial rule, of which inherited legal institutions seem particularly conducive todemocracy. The data for Muslim, Catholic, and British Colony are from La Porta et al.(1999).

In addition, two international control variables are included. First, researchers seem toagree that trade openness matters in the choice of political regime, and this variablereceives some empirical support in the literature (Gasiorowski 1995, Epstein et al. 2006),even though they disagree on how trade openness influences regime transition on theoret-ical grounds (see Boix 2003, pp. 142–143). Trade is measured as the natural log of thesum of the total imports and exports as a share of a country’s GDP. The data of GrowthRate, Resource, and Trade come from the World Bank World Development Indicator.

Second, the democratic domino theory asserts that democracy is contagious (Huntington1991, Gasiorowski and Power 1998, Gleditsch and Ward 2006, Leeson and Dean 2009).Thus, I include two measures of the percentage of (partial) democracies in a country’sregion, Regional Total Democracy (partial democracy + democracy) and Regional Democ-racy. The former is employed in democratization specifications and the latter, in democraticbreakdown specifications. Whereas regions are often divided according to either the Corre-lates of War coding (Reiter 2001, Lai and Melkonian-Hoover 2005) or Gasiorowski (1995;Bernhard et al. 2001), I divided the world into the following regions: advanced industrialcountries (Western Europe, North America, Australia, and New Zealand), Asia (includingFiji, Papua New Guinea, Solomon Islands), Latin America, Middle East and North Africa,Post-Communist countries, and Sub-Saharan Africa. This taxonomy may better fit thespecific geopolitical history of regime transition in the world.

My final control variable is a measure of a country’s experience of regime change.Past democratic transition or breakdown brings political learning that affects the prospectsfor regime transition (Huntington 1991, Acemoglu and Robinson 2001). To operationalizea country’s history with regime transition, I construct two variables, Previous DemocraticTransition and Previous Democratic Breakdown, which are the total number of previousexperiences with democratization (ND → PD or D) and democratic breakdown (D → PDor ND), respectively, that a country has experienced since 1960. As the variables thatcount the number of regional (partial) democracies, the former is used in democratizationspecifications and the latter in democratic breakdown specifications.

4. Model specification

I employ event history analysis to address the question: what affects the duration of non-democracies, partial democracies, and democracies? Event history analysis has becomecommon in empirical work on democratization over the past few years (Gasiorowski1995, Feng and Zak 1999, Przeworski et al. 2000, Bernhard et al. 2001, Reiter 2001,Pevehouse 2002, Lai and Melkonian-Hoover 2005, Epstein et al. 2006, Ulfelder 2007).Event history analysis enables one to answer a more extensive set of questions than con-ventional analyses by using information about the number, timing, and sequence ofchanges in the dependent variable; it enables one to find not only causes but also patterns

80 D. Jin Yi

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

of change concerning the time to the occurrence of an event (Box-Steffensmeier andJones 2004, Box-Steffensmeier and Zorn 2001).

Based on the theoretical assumption and the structure of the data, I use two differentmethods of event history analysis. First, one distinctive characteristic of data in this studyis that a country can experience events – democratic transition, democratic breakdown,consolidation, or backsliding, in my models – more than once. This is problematic if thetraditional duration model is used to estimate multiple events data because the traditionalmodel assumes that the event times are independent. This assumption is highly likely tobe violated in multiple event data. If we do not address correlation among repeated transi-tions, the standard errors will be underestimated. Thus, I employ a repeated-failures variantof the standard duration model, which adjusts the variance of the parameter estimates byclustering on subjects to account for the repeated nature of the data (Box-Steffensmeierand Jones 1997, Kelly and Lim 2000, Box-Steffensmeier and Zorn 2002).

Whereas the first method of event history analysis concerns the multiple occurrencesof the same events, that is, democratization or democratic breakdown, the second methodconcerns the multiple events of different types. If partial democracies are patients, theytend to fail either to democracy or to non-democracy. Interested in the different ways inwhich a partial democracy may fall, I may assume that different covariates can affect onekind of event occurrence, but not another. For instance, in my study, I can trace whetherand how key covariates, that is, levels of taxation and income inequality, may influencethe regime transitions of partial democracies toward non-democracy or democracy in thedifferent ways. If I find that a variable of interest has a positive effect on transition todemocracy but a negative effect on transition to non-democracy, this may support therobustness of my conclusion.

Thus, for this analysis, I employ an independent competing risks model to catchmultistate processes among Cox competing risks specifications.11 It assumes stochasticindependence, which means that risks associated with each of the different events actindependently. This assumption is analogous to independence of irrelevant alternatives(IIA) in discrete choice modeling. Estimation of an independent competing risks model issomewhat straightforward. In this analysis, I estimate separately two different univariateCox models, first consolidation and then backsliding models, while treating one regimetransition of interest as complete and the other as censored. In addition to computationalconvenience, another reason for adopting this method is that the independent risks models‘provide estimates of the hazard shapes which, at least in mean values, approximate thetrue shapes’ (Sueyoshi 1992, p. 52) even in the case in which the assumption of indepen-dence is not correct.

5. Results

I proceed to analyze the results from three sets of analyses that focused on the potentialcauses of democratization (the survival of non-democracy, ND → PD or D), democraticbreakdown (the survival of democracy, D → PD or ND), and the backsliding (PD →ND) or democratic consolidation (PD → D) of partial democracy, respectively. In the testfor multicollinearity through employing Stata’s variance inflation factor (VIF) command,I find no hint of significant collinearity, given that none of the variables has a VIF scoreof more than 3. Tables 2 and 3 display the main results of my analysis, and the resultsclearly support my argument that the influence of taxation on democracy is contingentupon the level of income inequality. The failure events for the Models 1, 2, and 3 inTable 2 are the transitions of non-democracy into (partial) democracy and those for the

Journal of Economic Policy Reform 81

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

Models 4, 5, and 6 in Table 2 are the transitions of democracy into partial democracy ornon-democracy.

I utilize a Cox semi-parametric model12 because it is more flexible due to having noassumptions about a functional form of the baseline hazard rate, making it an appropriatestrategy for many applications.13 The coefficients assess the impact of the explanatoryvariables on the hazard rate. As such, a positive sign implies that an increase in the valueof the explanatory variable leads to an increase in the hazard rate, indicating shorter sur-vival time, and a negative sign implies that the hazard rate experiencing the failure eventis proportionally lower, indicating longer survival time.

Among the control variables, one interesting pattern from the democratization anddemocratic breakdown models concerns the estimated effects of GDP per capita. Democ-ratization models in Table 2 suggest that GDP per capita has positive coefficients but ismarginally significant in only Model 2, indicating that there is no statistically strong evi-dence that economic development leads non-democracies to become (partial) democraciesmore quickly. Democratic breakdown models in Table 2 indicate that GDP per capitahas negative and statistically significant effects on prospects for a democratic breakdown,implying that higher economic development increases the probability of the democraticsurvival. Taken together, the results suggest that economic development may not increasestrongly the likelihood that non-democracies will transition to democracy, but it causesdemocracies, once established, to be less likely to break down to partial democracies ornon-democracies. This is clearly consistent with the ‘exogenous’ theory of democratiza-tion coined by Przeworski and his colleagues (Przeworski and Limongi 1997, Przeworskiet al. 2000).

The significance of the control variables for Regional Total Democracy and RegionalDemocracy remains consistent across the six models in Table 2, strengthening the overallimpression of the democratic domino theory that the increase in neighboring (partial)democracies is exerting considerable impact on whether the regimes in question will sur-vive or collapse; countries surrounding democratic neighbors tend to undergo a transitionto democracy and, once they succeed, maintain their democratic systems. The coefficientsof Muslim go in the expected direction and reach the standard thresholds of statistical sig-nificance across all democratization models, but they do not retain significance in thedemocratic breakdown models. It means that as Muslims over the whole populationincrease, the chance of democratization is likely to decrease. However, the likelihood ofdemocratic survival is not significantly associated with rising Islamic population.

The first explanatory variable of theoretical interest, Tax/Exp, does not achieve statisti-cal significance throughout all models. Although it takes on the anticipated signs, e.g.,positive in democratization Model 1 and negative in democratic breakdown Model 4 inTable 2, implying that higher levels of taxation boost the rate of failure for democratiza-tion but reduce it for democratic breakdown, one may not infer that higher levels of taxa-tion appear to be systematically related to more democratic regimes, because itscoefficient is not significant.14

As the second variables of interest, Inequality and its square, Inequality2, retain theanticipated signs in Model 2 in Table 2, but the joint test of the null hypothesis thatInequality and Inequality2 have no explanatory power is not rejected even at the 10%level. It means that inequality may not affect the likelihood of democratization in aninverted-U shape as theorized by Acemoglu and Robinson (2006). By and large, thereseems to be little evidence to support their final hypothesis.

The effect of Inequality on democratic breakdown is somewhat different. Table 2illustrates that the effect of inequality on democratic breakdown is more straightforward

82 D. Jin Yi

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

than its effect on democratization. In the specifications of democratic breakdown, I didnot include a square term of inequality because in theory, no one hypothesizes a nonlin-ear relationship between inequality and democratic breakdown, and, in practice, the coef-ficient of Inequality, without its square term, is strongly statistically significant. Asconventional wisdom, inequality is not good for the stability of a democracy: it isstrongly significant and negative signs indicate that higher levels of inequality are likelyto increase the likelihood of democratic breakdown. All in all, this analysis finds someevidence in favor of the previous prediction. Inequality harms democracy as the indicatorto facilitate the instability of established democracy, which largely echoes the empiricalfinding of Houle (2009).

My leading hypothesis is to examine whether taxation triggers the failure of regimesin question in conjunction with the levels of income inequality. This is achieved by mul-tiplying the appropriate pairs of explanatory variables, here Tax/Exp and Inequality, andincluding the product term, Tax/Exp × Inequality. Model 3 suggests that taxation’sinteraction term with inequality has a significant positive effect on democratization, butModel 6 indicates that its statistical significance seems to disappear, although its substan-tial direction (negative) of a coefficient does match what I hypothesize when it comes tothe models of democratic breakdown. These estimates largely confirm the key part of myargument: taxation and inequality seem to exert a discernible, interactive impact on thefailure of non-democracies in the sense that the existence of one factor strengthens theinfluence of the other. More specifically, higher levels of taxation with rising inequalityhave a facilitating effect on democratization, but their democracy-friendly effects, oncedemocracies are established, appear to remain weak and insignificant.

To see how the effects of taxation on democratization vary across the different levelsof inequality, I start by finding the Gini values that correspond to the mean, one standarddeviation below the mean, and one standard deviation above the mean. Because the vari-able of Inequality is centered, these values are approximately 0, –10, and +10, respec-tively (its standard deviation in the estimation sample is 10.66). The effect of taxation ata given value of α inequality is given by β1 + aβ2, which comes from the form of Model3 in Table 2 (β1 is a parameter of Tax/Exp and β2 is one of Tax/Exp × Inequality). Therelevant null hypothesis, H0: β1 + aβ2 = 0, can be tested by employing the following teststatistic:

t ¼ b̂1 þ ab̂2

r̂b̂1þab̂2

where

r̂b̂1þab̂2¼

ffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffir̂2b̂1þ a2r̂2

b̂2þ 2ar̂b̂1b̂2

q

Table 3. t-test statistics of tax/exp across different levels of income inequality.

β1 + aβ2 Statistic p

Gini One standard deviation below the mean, -10 �0.67 �1.1020 >0.10The mean, 0 0.07 0.16 0.869One standard deviation above the mean, 10 0.81 1.7744 <0.05

Journal of Economic Policy Reform 83

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

Table4.

Regim

echange

ofpartialdemocracy,1970-2000.

Cox

analysis

Consolid

ation

Backslid

ing

Model

1Model

2Model

3Model

4Model

5Model

6

Tax/Exp

�0.878

�1.173

�0.167

1.171

(0.747)

(0.775)

(0.391)

(1.043)

Inequality

�0.014

�0.026

�0.001

0.001

(0.013)

(0.016)

(0.009)

(0.024)

Inequality2

�0.001

�0.001

(0.001)

(0.001)

Tax/Exp

×Inequality

0.087

�0.147

(0.101)

(0.094)

GDPpercapita

(log)

0.725⁄

⁄0.590⁄

⁄0.441^

�0.328

⁄�0

.276

⁄�0

.890

⁄⁄⁄

(0.277)

(0.189)

(0.254)

(0.153)

(0.130)

(0.223)

Growth

Rate

0.091

0.076^

0.054

0.014

0.055

0.138^

(0.061)

(0.043)

(0.060)

(0.039)

(0.038)

(0.079)

Trade

(log)

�0.051

�0.055

�0.184

0.245

�0.064

�0.051

(0.199)

(0.198)

(0.219)

(0.251)

(0.165)

(0.299)

Catholic

0.0004

0.002

0.002

0.006

0.002

0.006

(0.002)

(0.002)

(0.002)

(0.007)

(0.004)

(0.009)

Muslim

�0.014

�0.021

⁄�0

.018

0.004

0.006⁄

0.003

(0.104)

(0.010)

(0.015)

(0.006)

(0.003)

(0.011)

ELF

�0.155

0.268

�0.059

�0.273

�0.584

�1.098

⁄(0.390)

(0.338)

(0.417)

(0.352)

(0.356)

(0.540)

Resource

0.011

0.005

0.020

0.004

0.013⁄

⁄0.010

(0.024)

(0.019)

(0.021)

(0.009)

(0.005)

(0.018)

Britishcolony

0.182

0.185

0.133

�0.650

�0.131

�0.808

(0.174)

(0.170)

(0.209)

(0.477)

(0.251)

(0.899)

Regionaldemocracy

0.666

0.429

0.538

(0.421)

(0.343)

(0.422)

RegionaltotalD

�1.978

⁄⁄⁄

�1.613

⁄⁄⁄

�1.478

⁄(0.604)

(0.478)

(0.632)

PreviousD

transitio

ns�0

.293

�0.179

�0.277

(Contin

ued)

84 D. Jin Yi

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

Table4.

(Contin

ued).

Cox

analysis

Consolid

ation

Backslid

ing

Model

1Model

2Model

3Model

4Model

5Model

6

(0.233)

(0.175)

(0.268)

PreviousD

breakdow

n�0

.935

⁄�0

.974

⁄⁄�1

.375

⁄⁄(0.442)

(0.324)

(0.454)

Observatio

ns1130

1617

926

670

1144

396

Countries

70101

6452

9743

Log

Likelihood

�2798.218

�3888.651

�2187.844

�1078.847

�1984.310

�422.600

Chi-Square

111.12

253.46

113.72

81.00

109.19

93.33

Note:

Table

entriesarecoefficients.Rob

uststandard

errors

with

clustering

onthecoun

tryarecontainedin

parentheses.Prob>chi2

=0.00

00forallmod

els.

^p60.10

;⁄ p

60.05

;⁄⁄p60.01

;⁄⁄

⁄ p60.00

1.

Journal of Economic Policy Reform 85

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

Table 3 presents the test statistics obtained by substituting a with the values –10, 0,and +10, along with their statistical significance when referred to t-distribution with n –K – 1 degrees of freedom.15 The last statistic substituted by 10, one standard deviationabove the mean, is only statistically significant at the level of 5%. This pattern indicatesthat an increase of taxation appear to help promote democratization if countries suffer rel-atively higher levels of income inequality, a Gini score of approximately 47, but that itseffect on democratization does not remain on the edge of statistical significance if coun-tries enjoy the middle level or equal distributive structures.

Table 4 shows the results of estimation for the probability that partial democraciesconsolidate or collapse. At first glance, compared with backsliding, models of democraticconsolidation provide few hints as to the potential factors that aid democratic consolida-tion. Note first that among the explanatory variables, GDP per capita is the strongestfactor to influence the likelihood of both consolidation and backsliding: higher GDP percapita is a promoting factor for democratic consolidation and also an inhibiting factor forbacksliding. Democratic neighbors and previous democratic experience do not influencethe likelihood of consolidation, but (partially) democratic neighbors and previous demo-cratic breakdown strongly stabilize partial democracies. Turning to the variables of inter-est, levels of taxation and inequality, and their interaction term do not make partialdemocracies more likely to consolidate or collapse.

Admittedly, it is plausible to suspect that the empirical findings can be biased simplydue to failing to control for regional differences that are closely correlated with the varia-tions of inequality and taxation. As argued earlier, regime transitions have not only beenaffected by a region’s specific geopolitical contexts, but the extents of taxation andinequality also vary virtually across regions. As Table 5 shows, levels of taxation inMENA (49%), in which almost all countries are robust non-democracies, are much lowerthan those of other regions – almost half of advanced industrial democracies (84%) – andlevels of income inequality in Latin America, in which democracies are among the mostfragile, are the worst in the world, no matter how one measures them. The hypothesizednegative link between lower taxation and democracies and the apparent positive linkbetween higher inequality and democratic breakdown may be due to the effects of beingfrom MENA and Latin America, respectively, on both of the dependent and the explana-tory variables.

To check the robustness of my empirical findings against these potential spurious rela-tionships, I also re-estimated each specification, including regional dummies. If regionalcontexts serve as a source of spuriousness, the observable implication is that the statisti-cal significance should dwindle with their inclusion. Yet, this is not the case. As reportedin Table 6, the significance and estimated signs of the explanatory variables of interestbarely change. Accordingly, the conditional effect of taxation on democratization and thenegative effect of inequality on democratic breakdown do not seem simply a proxy forregional differences.16

6. Discussions and conclusion

This article has returned to the theory linking levels of taxation with democracy. In doingso, it has reappraised the central argument of taxation’s effect on regime transition, focus-ing on the role of income inequality as a factor mediating the influence of taxation. Con-trary to fiscal sociology and the rentier state theory discussed earlier, my event historyanalysis does not support their core claim. Higher levels of taxation are not linearly asso-

86 D. Jin Yi

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

ciated with the survival of democracy as well as the regime transition to democracy. Inline with Acemoglu and Robinson (2006), I also find an inverted U-shaped relationshipbetween income inequality and the likelihood of democratization, but its effect is not sta-tistically significant. In addition, higher levels of income inequality decrease the likeli-hood that democracies will remain democracies.

The main contribution of my analysis is to link taxation to income inequality in thestudy of the impact of factors affecting the survival of regime types, and to show howone reinforces the other. My results show that taxation and inequality act in synergy topromote democratization. More specifically, the interactive influence of taxation andinequality works in a different way. It is not necessarily the case that an increase in taxa-tion leads to improved chances of democratization when inequality is low. Yet, in highlyunequal countries, it is an increase in taxation that is associated with higher plausibilityof democratization. Confirming my hypotheses, this new discovery provides new insighton the study of democratization.

Its implication is very important in terms of theoretical and policy-wise bounds. If wefind valid evidence to believe that the nexus between taxation, income inequality, anddemocracy has existed in specific directions, we may, theoretically, understand a moresophisticated picture of democratization, and that, also practically, tax policies inauthoritarian regimes to lower or raise the levels of tax burdens may have pro- or anti-democratic effects, which produces policy-wise recommendation for supporters of democ-racy.

Employing trichotomous measures of democracy, this article may cast new light onthe significance of the intermediate category, that is, partial democracy. In line withEpstein et al. (2006), it is clear that a dichotomous classification of democratic regimesmay not be appropriate for capturing most of the real-world diversity in regime transition.Categorizing countries into non-democracies, partial democracies, and democracies, Ievaluate how the explanatory variables work in the democratic consolidation and back-sliding phases as well as the democratization and democratic breakdown phases. GDPper capita, which appears to follow an ‘exogenous’ route of democratization (Przeworskiand Limongi 1997, Przeworski et al. 2000), significantly increases the prospects of exist-ing partial democracies, both by enhancing democratic consolidation and by inhibitingbacksliding. However, some other factors, which are considered as strongly robust causesof democratization and democratic breakdown, do not affect the survival of partialdemocracies, especially in the consolidation phase.

A question then arises: what makes partial democracies move up or down the ladder?As shown in Table 1, being more volatile than other regimes types, partial democraciesseem to move at random (36 cases of backsliding versus 31 cases of consolidation). In

Table 5. Means of tax/exp and inequality by region.

Tax/Exp Inequality

Advanced Industrial Countries 0.84 30.0Africa 0.71 43.4Asia 0.71 38.8Latin America 0.78 52.8MENA 0.49 36.9Post-Communist Countries 0.77 30.3

Journal of Economic Policy Reform 87

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

Table6.

Dem

ocratizationanddemocratic

breakdow

nwith

regional

dummies,1970–2

000.

Cox

analysis

Dem

ocratization

Dem

ocratic

breakdow

n

Model

1Model

2Model

3Model

4Model

5Model

6

Tax/Exp

0.346

0.009

�0.478

0.661

(0.444)

(0.452)

(0.427)

(0.857)

Inequality

�0.006

�0.007

0.037⁄

⁄⁄0.061⁄

⁄⁄(0.009)

(0.010)

(0.009)

(0.018)

Inequality2

�0.001

�0.001

(0.0004)

(0.001)

Tax/Exp

×Inequality

0.057⁄

�0.073

(0.033)

(0.076)

GDPpercapita

(log)

0.524⁄

⁄⁄0.400⁄

⁄⁄0.488⁄

⁄⁄�0

.469

⁄⁄�0

.522

⁄⁄⁄

�0.609

⁄⁄(0.150)

(0.119)

(0.152)

(0.178)

(0.119)

(0.237)

Observatio

ns1564

2283

1140

1564

2283

1140

Countries

84130

7584

130

75Log

Likelihood

�3541.726

�5738.851

�2689.731

�2002.606

�3881.682

�1031.143

Chi-Square

143.09

197.06

99.58

119.95

210.43

122.15

Note:

Table

entriesarecoefficients.Rob

uststandard

errors

with

clustering

onthecoun

tryarecontainedin

parentheses.Prob>chi2

=.000

0forallmod

els.^p

60.10

;⁄ p

60.05

;⁄⁄p60.01

;⁄⁄

⁄ p60.00

1.

88 D. Jin Yi

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

an era of worldwide democratization, few questions have greater theoretical and practicalimportance than that concerning partial democracies. Yet, relatively little scholarly atten-tion has been devoted to their specific mechanisms for their movement. Theoretical expla-nations of democratization may be of little use if one does not acknowledge thisintermediate category. The need for this kind of investigation sets a very interestingagenda for future research.

Countries used in the empirical analysis

Numbers in parenthesis refer to one of the years for which the observation was availablein Model 1 in Table 2.

Albania (4), Argentina (14), Australia (30), Austria (30), Azerbaijan (6), Bahrain (19),Belarus (5), Belgium (29), Bolivia (14), Brazil (6), Bulgaria (9), Cameroon (18), Canada(25), Chile (28), Colombia (29), Congo (11), Costa Rica (28), Croatia (9), Cyprus (23),Czech Republic (7), Denmark (29), Dominican Republic (27), Egypt (22), Estonia (8),Finland (27), France (26), Gabon (13), Georgia (3), Germany (27), Greece (27), Guinea(6), Haiti (4), Honduras (8), Hungary (19), India (26), Indonesia (28), Iran (28), Ireland(28), Israel (28), Italy (27), Jamaica (10), Japan (24), Kazakhstan (3), Korea (28), Kuwait(20), Latvia (6), Lesotho (19), Liberia (13), Lithuania (7), Malaysia (26), Mauritius (19),Mexico (27), Morocco (26), Namibia (5), Nepal (28), Netherlands (25), Nicaragua (7),Norway (27), Pakistan (26), Panama (19), Paraguay (5), Peru (11), Poland (5), Portugal(24), Romania (8), Russia (4), Senegal (13), Slovak Republic (4), Slovenia (8), SouthAfrica (28), Spain (28), Sri Lanka (27), Sweden (30), Switzerland (23), Thailand (28),Trinidad (9), Tunisia (28), Turkey (2), UK (29), Uruguay (26), USA (26), Venezuela(30), Zambia (25), Zimbabwe (22).

AcknowledgmentsThe author wishes to thank Evelyne Huber, John Stephens, Layna Mosley, Andrew Reynolds, TomCarsey, and anonymous reviewers for their help in preparing this article. This work was supportedby Hankuk University of Foreign Studies Research Fund of 2012.

Notes1. For an analytical model of the nexus between taxation and democracy, see Bates and Lien

(1985).2. I borrow my terminology, the effects of low demand and high supply, from Ulfelder’s (2007,

p. 997) lexicon, ‘demand-side and supply-side explanations’.3. See Gleditsch and Ward (1997) for critiques of practices to handle the Polity as continuous

measures of democracy.4. See Long and Freese (2005, chap. 5) for a more thorough discussion of this issue.5. For a quantitative and analytical model, see Boix (2003).6. For a review of empirical studies on this issue, see Houle (2009, Table 1).7. Houle (2009, pp. 595–598) argues that, contrary to its effect on democratization, inequality’s

effect on democratic breakdown (in his term, ‘consolidation’) is straightforward; inequalityincreases the likelihood of democratic breakdown because it only influences the payoff ofelites (increase in costs from redistribution) but not that of citizens.

8. I use the term consolidation in a general mode to denote regime change in which partialdemocracies move up the ladder. There are no theoretical connotations for the definition of thischange.

9. The World Tax Database at the University of Michigan provides taxation data only for 1970–2000.

10. For more information, visit http://www.wider.unu.edu/wiid/wiid.htm and look at the user’sguide.

Journal of Economic Policy Reform 89

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

11. David and Moeschberger (1978) provide general discussions of competing risks models. Foran example of an independent competing risks model, see Zorn and Van Winkle (2000), andfor some methodological shortcomings of this approach, see Gordon (2002).

12. For the Cox models, I employ the Efron method for the approximation of the partial likelihoodfunction because it provides a more accurate approximation than the Breslow method when alarge number of tied cases exist.

13. On the issue of model selection, see Box-Steffensmeier and Jones (2004, chap. 6).14. I also run the model with 5-year or 10-year lags of Tax/Exp, following Ross (2004), but do

not find its statistical significance.15. It is straightforward to get a t-test on the effect of Tax/Exp when Inequality is at its mean

value. The parameter of Tax/Exp in Model 3 can be directly evaluated because it captures theeffect of Tax/Exp when Inequality is at 0, which is its centered mean.

16. It is interesting that GDP per capita becomes strongly significant in all models when regionaldummies are controlled.

ReferencesAcemoglu, D. and Robinson, J.A., 2000. Why did the West extend the franchise? Democracy,

inequality, and growth in historical perspective. The Quarterly Journal of Economics, 115(November), 1167–1199.

Acemoglu, D. and Robinson, J.A., 2001. A theory of political transitions. The American EconomicReview, 91 (September), 938–963.

Acemoglu, D. and Robinson, J.A., 2006. Economic origins of dictatorship and democracy. NewYork: Cambridge University Press.

Ames, E. and Rapp, R., 1977. The birth and death of taxes: a hypothesis. The Journal of Eco-nomic History, 37 (March), 161–178.

Anderson, L., 1995. Peace and democracy in the Middle East: the constraints of soft budgets.Journal of International Affairs, 49 (Summer), 25–44.

Ardant, G., 1975. Financial policy and economic infrastructure of modern states and nations. In:C. Tilly, ed. The formation of national states in Western Europe. Princeton: Princeton Univer-sity Press.

Barro, R.J., 1999. Determinants of democracy. Journal of Political Economy, 107 (6), 158–183.Bates, R. and Lien, D.D., 1985. A note on taxation, development, and representative government.

Politics and Society, 14, 53–70.Beblawi, H., 1987. The rentier state in the Arab World. In: H. Beblawi and G. Luciani, eds. The

rentier state: nation, state and the integration of the Arab World. New York: Croom Helm,85–98.

Bernhard, M., Nordstrom, T., and Reenock, C., 2001. Economic performance, institutional interme-diation, and democratic survival. The Journal of Politics, 63 (August), 775–803.

Boix, C., 2003. Democracy and redistribution. Cambridge: Cambridge University Press.Boix, C. and Stokes, S.C., 2003. Endogenous democratization. World Politics, 55 (July), 517–549.Bolt, K., Matete, M., and Clemens, M., 2002. Manual for calculating adjusted net savings. Wash-

ington, DC: World Bank, Environment Department.Box-Steffensmeier, J. and Jones, B., 1997. Time is of the essence. event history models in political

science. American Journal of Political Science, 41 (October), 1414–1461.Box-Steffensmeier, J. and Jones, B., 2004. Event history modeling: a guide for social scientists.

New York: Cambridge University Press.Box-Steffensmeier, J. and Zorn, C., 2001. Duration Models and proportional hazards in political

science. American Journal of Political Science, 45 (October), 972–988.Box-Steffensmeier, J. and Zorn, C., 2002. Duration models for repeated events. The Journal of

Politics, 64 (November), 1069–1094.Braun, R., 1975. Taxation, sociopolitical structure, and state-building: Great Britain and Branden-

burg-Prussia. In: C. Tilly, ed. The formation of national states in Western Europe. Princeton,NJ: Princeton University Press.

Cheibub, J.A., 1998. Political regimes and the extractive capacities of governments: taxation indemocracies and dictatorships. World Politics, 50 (April), 349–376.

Collier, D. and Adcock, R., 1999. Democracy and dichotomies: a pragmatic approach to choicesabout concepts. Annual Review of Political Science, 2, 537–565.

90 D. Jin Yi

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

Crystal, J., 1990. Oil and politics in the Gulf. New York: Cambridge University Press.Dahl, R.A., 1971. Polyarchy: participation and opposition. New Haven, CT: Yale University

Press.David, H.A. and Moeschberger, M.L., 1978. The theory of competing risks. New York: Macmillan.de Tocqueville, A., 1990. Democracy in America. New York: Vintage Books.Dunning, T., 2008. Crude democracy. Cambridge: Cambridge University Press.Diamond, L., 1996. Democracy in Latin America: degrees, illusions, and directions for consolida-

tion. In: T. Farer, ed. Beyond sovereignty: collectively defending democracy in the Americas.Baltimore, MD: Johns Hopkins University Press.

Epstein, D.L., Bates, R., Goldstone, J., Kristensen, I., and O’Halloran, S., 2006. Democratic transi-tions. American Journal of Political Science, 50 (July), 551–569.

Feng, Y. and Zak, P.J., 1999. The determinants of democratic transitions. Journal of Conflict Reso-lution, 43 (April), 162–177.

Finkel, S.E. and Muller, E.N., 1998. Rational choice and the dynamic of collective political action:evaluating alternative models with panel data. American Political Science Review, 92 (March),37–49.

Fish, S.M., 2002. Islam and authoritarianism. World Politics, 55 (October), 4–37.Gasiorowski, M.J., 1995. Economic crisis and political regime change: an event history analysis.

American Political Science Review, 89 (December), 882–897.Gasiorowski, M.J. and Power, T.J., 1998. The structural determinants of democratic consolidation:

evidence from the Third World. Comparative Political Studies, 31 (December), 740–771.Gause, G., 1993. Oil monarchies: domestic and security challenges in the Arab Gulf States.

Council on Foreign Relations Press.Gleditsch, K.S. and Ward, M.D., 1997. Double take: a reexamination of democracy and autocracy

in modern politics. Journal of Conflict Resolution, 41, 361–383.Gleditsch, K.S. and Ward, M.D., 2006. Diffusion and the international context of democratization.

International Organization, 60 (October), 911–933.Goldstone, J., 2000. State failure task force. phase III findings. McLean, VA: SAIC.Gordon, S.C., 2002. Stochastic dependence in competing risks. American Journal of Political Sci-

ence, 46 (January), 200–217.Gurr, T.R. and Jaggers, K., 1999. Polity 98: regime characteristics, 1800-1998. Database.Haggard, S. and Kaufman, R.R., 1995. The political economy of democratic transitions. Princeton,

NJ: Princeton University Press.Herb, M., 2005. No representation without taxation: rents, development, and democracy. Compara-

tive Politics, 37 (April), 297–316.Houle, C., 2009. Inequality and democracy: why inequality harms consolidation but does not affect

democratization. World Politics, 61 (October), 589–622.Huntington, S.P., 1991. The third wave: democratization in the late twentieth century. Norman:

University of Oklahoma.Huntington, S.P., 1996. Democracy for the long haul. Journal of Democracy, 7 (April), 3–13.Kelly, P.J. and Lim, L., 2000. Survival analysis for recurrent event data. Statistics in Medicine, 19,

13–33.La Porta, R., Lopez-de-Silanes, F., Shleifer, A., and Vishny, R., 1999. The quality of government.

The Journal of Law, Economics, and Organization, 15 (1), 222–279.Lai, B. and Melkonian-Hoover, R., 2005. Democratic progress and regress: the effect of parties on

the transitions of states to and away from democracy. Political Research Quarterly, 58 (Decem-ber), 551–564.

Leeson, P.T. and Dean, A.M., 2009. The democratic domino theory: an empirical investigation.American Journal of Political Science, 53 (July), 533–551.

Lipset, S.M., 1959. Some social requisites of democracy: economic development and politicallegitimacy. American Political Science Review, 53 (March), 69–105.

Lipset, S.M., 1994. The social requisites of democracy revisited: 1993 Presidential Address. Ameri-can Sociological Review, 59 (February), 1–22.

Long, J.S. and Freese, J., 2005. Regression models for categorical dependent variables using stata.College Station, Texas: Stata Press.

Luciani, G., 1994. The oil rent, the fiscal crisis of the state and democratization. In: G. Salame, ed.Democracy without democrats? The renewal of politics in the Muslim world. New York: I.B.Tauris.

Journal of Economic Policy Reform 91

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013

Mansfield, E. and Snyder, J., 1995. Democratization and war. Foreign Affairs, 74 (May/June), 79-97.

Marshall, M.G. and Jaggers, K., 2007. Polity IV dataset. College Park: University of Maryland,Center for International Development and Conflict Management.

Muller, E.N., Dietz, H.A., and Finkel, S.E., 1991. Discontent and the expected utility of rebellion:the case of Peru. American Political Science Review, 85 (December), 1261–1282.

Muller, E.N. and Seligson, M.A., 1994. Civic culture and democracy: the question of causal rela-tionships. American Political Science Review, 88 (September), 635–652.

Munck, G.L. and Verkuilen, J., 2002. Conceptualizing and measuring democracy: evaluating alter-native indices. Comparative Political Studies, 35 (February), 5–34.

North, D. and Weingast, B.R., 1989. Constitutions and commitment: the evolution of institutionalgoverning public choice in seventeenth-century England. The Journal of Economic History, 49(December), 803–832.

Pevehouse, J.C., 2002. With a little help from my friends? Regional organizations and the consoli-dation of democracy. American Journal of Political Science, 46 (July), 611–626.

Poggi, G., 1978. The development of the modern state: a sociological introduction. Stanford, CA:Stanford University Press.

Przeworski, A., Alvares, M.E., Cheibub, J.A., and Limongi, F., 2000. Democracy and develop-ment. Cambridge: Cambridge University Press.

Przeworski, A. and Limongi, F., 1997. Modernization: theories and facts. World Politics, 49 (Janu-ary), 155–183.

Reiter, D., 2001. Does peace nurture democracy? The Journal of Politics, 63 (August), 935–948.Roeder, P.G., 2001. Ethnolinguistic Fractionalization (ELF) Indices, 1961 and 1985. http://weber.

ucsd.edu/~proeder/elf.htm (October 13, 2009).Ross, M., 1999. The political economy of the resource curse. World Politics, 51 (January), 297–

322.Ross, M., 2001. Does oil hinder democracy? World Politics, 53 (April), 325–361.Ross, M., 2004. Does taxation lead to representation? British Journal of Political Science, 34,

229–249.Smith, B., 2004. Oil wealth and regime survival in the developing world, 1960–1999. American

Journal of Political Science, 48 (April), 232–246.Sueyoshi, G.T., 1992. Semiparametric proportional hazards estimation of competing risks models

with time-varying covariates. Journal of Econometrics, 51 (January), 25–58.Tilly, C., 1990. Coercion, capital, and European states, AD 990–1990. Cambridge, MA: Black-

well.Tilly, C., 2004. Contention and democracy in Europe, 1650–2000. Cambridge: Cambridge Univer-

sity Press.Ulfelder, J., 2007. Natural-resource wealth and the survival of autocracy. Comparative Political

Studies, 40 (August), 995–1018.UNU-WIDER, 2007. World Income Inequality Database.World Bank, 2008. World Development Indicators.Zakaria, F., 2003. The future of freedom. New York: W.W. Norton.Zorn, C.J.W. and Van Winkle, S.R., 2000. A competing risks model of supreme court vacancies,

1789–1992. Political Behavior, 22 (June), 145–166.

92 D. Jin Yi

Dow

nloa

ded

by [

Dae

Jin

Yi]

at 2

1:12

18

Mar

ch 2

013