Modelling transitions to and from democracy

Democratization

Volumn 14, Issue 3, 2007, Pages 351-387

  Ulfelder, Jay ^a   Lustik, Michael ^a  

^a NONE

Author keywords

Autocracy; Democracy; Event history analysis; Forecasting; Transitions

Indexed keywords

AUTHORITARIANISM; DEMOCRACY; DEMOCRATIZATION; ECONOMIC DEVELOPMENT; LEADERSHIP; MODELING; MODERNIZATION; POLITICAL CHANGE;

EID: 34249727767     PISSN: 13510347     EISSN: 1743890X     Source Type: Journal    
DOI: 10.1080/13510340701303196     Document Type: Article

References (125)

1. Although the definitions, data sources and measures are generally the same, these models are not a subset of the Task Force's longstanding research on political instability. For more information on that research or the Task Force, see http://globalpolicy.gmu.edu/pitf/index.htm. The views expressed here are the authors' alone; they do not represent the official views of the US government, the US intelligence community, or the Central Intelligence Agency, which funds the Task Force's work.
2. See especially Michael T. Hannan and Glenn R. Carroll, 'Dynamics of Formal Political Structure: An Event-History Analysis', American Sociological Review, Vol. 46, No. 1 (1981), pp.19-35
3. Mark J. Gasiorowski, 'Economic Crisis and Political Regime Change: An Event History Analysis,' American Political Science Review, Vol. 89, No. 4 (1995): Pp.882-97
4. and, most recent and most prominent, Adam Przeworski, Michael E. Alvarez, José Antonio Cheibub, and Fernando Limongi, Democracy and Development: Political Institutions and Well-Being in the World, 1950-1990 (Cambridge: Cambridge University Press, 2000).
5. Robert A. Dahl, Polyarchy: Participation and Opposition (New Haven, CT: Yale University Press, 1971).
6. Adam Przeworski et al. (note 2). Of the 3,658 country-years for which data were available on our measure and the one used by Przeworski et al., 91.2 per cent are coded identically. Most of the divergent cases (222 of 323 country-years, or 6.0 per cent of the total sample) are coded as democracies under our scheme but dictatorships in theirs, and most of those differences are due to those authors' alternation rule, which requires an electorally induced change in government before a country may be considered democratic. Unsurprisingly, we found a similarly strong correlation between our measure and one based on splitting the Polity scale at 0. The theoretical and practical motivations for our approach are discussed in the 'Definitions and Measures' section of this article.
7. As King and Zeng argue, however, modelling for predictive accuracy and modelling to discern causal structure are not necessarily at cross-purposes. See Gary King and Langche Zeng, 'Improving Forecasts of State Failure', World Politics, Vol. 53, No. 4 (2001), pp.634-6.
8. See for example Dahl (note 3)
9. Kenneth A. Bollen and Robert W. Jackman, 'Democracy, Stability, and Dichotomies', American Sociological Review, Vol. 54, No. 4 (1989), pp.612-21
10. Larry Diamond, Developing Democracy: Toward Consolidation (Baltimore, MD: The Johns Hopkins University Press, 1999). Dahl (p. 8) labels the four corners of the two-dimensional space defined by his concepts of contestation and inclusiveness, but he also notes that, 'Perhaps the preponderant number of national regimes in the world today would fall into the [unlabeled] mid-area. Many significant changes in regimes, then, involves shifts within, into, or out of this important central area, as these regimes become more (or less) inclusive and increase (or reduce) opportunities for public contestation.'
11. Giovanni Sartori, The Theory of Democracy Revisited (Chatham, NJ: Chatham House, 1987), p.184,
12. and Przeworski et al. (note 2).
13. For a strong critique of continuous measures derived from Freedom House and Polity and an intriguing effort to create a continuous measure linked to a binary definition of regime types, see James Vreeland, 'A Continuous Schumpeterian Measure of Democracy', paper presented at the Annual Meeting of the American Political Science Association, Philadelphia, PA (2003).
14. David Collier and Robert Adcock, 'Democracy and Dichotomies: A Pragmatic Approach to Choices about Concepts', Annual Review of Political Science (1999), p.539.
15. It is possible, of course, to design a statistical analysis that avoids this assumption about inverse effects by treating regime transitions as discrete events but includes more than two regime types; see for example Gasiorowski (note 2) and David L. Epstein, Robert Bates, Jack Goldstone, Ida Kristensen, and Sharyn O'Halloran, 'Democratic Transitions', paper prepared for presentation at the Annual Meeting of the American Political Science Association, Washington DC (2005). Even the addition of one category can significantly reduce the size of the samples available for estimating statistical models, however, and this problem quickly worsens as the number of categories increases, making it difficult to obtain reliable estimates of variables' effects.
16. Joseph Schumpeter, Capitalism, Socialism, and Democracy (New York: Harper, 1942),
17. and Samuel Huntington, The Third Wave: Democratization in the Late Twentieth Century (Norman, OK and London: University of Oklahoma Press, 1991).
18. Scholars have rightly cautioned against equating democracy with elections - what Terry Lynn Karl aptly calls the 'fallacy of electoralism' ('Dilemmas of Democratization in Latin America', Comparative Politics, Vol. 23, No. 1 (1990), pp. 1-23) - because a myopic focus on the process of casting and counting ballots may lead to the neglect of other institutions essential to the realization of democratic principles. Still, virtually every major definition of democracy put forward in recent decades identifies elections as the critical procedural component of this concept, and our approach is consistent with that pattern.
19. A focus on accountability as the key feature of democracy is proposed by Philippe C. Schmitter and Terry Lynn Karl, 'What Democracy Is...and Is Not', Journal of Democracy, Vol. 2, No. 1 (1991), pp.75-88.
20. See Juan J. Linz and Alfred Stepan, Problems of Democratic Transition and Consolidation: Southern Europe, South America, and Post-Communist Europe (Baltimore, MD: The Johns Hopkins University Press, 1996)
21. Diamond (note 6)
22. Barbara Geddes, 'Authoritarian Breakdown: Empirical Test of a Game Theoretic Argument', paper presented at the Annual Meeting of the American Political Science Association, Atlanta, GA (1999)
23. and Juan J. Linz, Totalitarian and Authoritarian Regimes (Boulder, CO: Lynne Rienner, 2000) for efforts to develop sets of categories that approach typologies. Recent scholarship has often invoked new or additional categories on something of an ad hoc basis, coining terms like 'electoral autocracy' or 'competitive authoritarianism' to represent patterns that may have specific implications for the durability of authoritarian rule and potential paths to democracy, without attempting systematically to classify autocratic regimes. For a thoughtful critique of the proliferation of types and terms,
24. see Ariel C. Harmony and Hector E. Schamis, 'Babel in Democratization Studies', Journal of Democracy, Vol. 16, No. 4 (2005), pp.113-28.
25. See Keith Jaggers and Ted Robert Gurr, 'Tracking Democracy's Third Wave with the Polity III Data', Journal of Peace Research, Vol. 32, No. 4 (1995), pp.469-482.
26. Harry Eckstein and Ted Robert Gurr, Patterns of Authority: A Structural Basis for Political Inquiry (New York: Wiley and Sons, 1975).
27. The lack of comparable historical depth was one reason we chose not to use the data set compiled by Przeworski et al. (note 2) for their analysis. We also had concerns about possible sample bias resulting from those authors' alternation rule. Those concerns are discussed in the section of this paper describing the findings on horizontal accountability emerging from our model of transitions from democracy to autocracy.
28. Monty G. Marshall and Keith Jaggers, Polity IV Project: Political Regime Characteristics and Transitions, 1800-2002, Dataset Users' Manual (College Park: Center for International Development and Conflict Management, University of Maryland, 2002), p.58.
29. Ibid., p.25.
30. The distinction between suppressed (PARCOMP = 2) and factional (PARCOMP = 3) participation is not always a sharp one in the real world, but it is an important one in our classification scheme, so it deserves some additional discussion. According to Polity, 'Suppressed competition is distinguished from factional competition by the systematic, persisting nature of the restrictions: Large classes of people, groups, or types of peaceful political competition are continuously excluded from the political process', as might be evidenced by: Prohibitions on certain kinds of political organizations, either by the type of group or the people involved, that affect more than 20 percent of the adult population; prohibitions on certain kinds of political actions, most notably competing in elections; or systematic harassment of the political opposition, although this may also occur under repressed or factional competition, depending on the character of the regime and the opposition and the persistence of political groups (Marshall and Jaggers, note 17, p. 27).
31. For a description of the previous generation of our research on this topic, see pp.161-79 in Robert H. Bates, David L. Epstein, Jack A. Goldstone, Ted Robert Gurr, Barbara Harff, Colin H. Kahl, Kristen Knight, Marc A. Levy, Michael Lustik, Monty G. Marshall, Thomas M. Parris, Jay Ulfelder, and Mark R. Woodward, Political Instability Task Force Report: Phase IV Findings (McLean, VA: Science Applications International Corp., 18 November 2003), available upon request through http://globalpolicy.gmu.edu/pitf/index.htm.
32. Marshall and Jaggers (note 17), p.23.
33. Robert Dahl, for example, does not include executive constraints in his list of seven minimal conditions that must be present for polyarchy to exist; see Dilemmas of Pluralist Democracy (New Haven, CT: Yale University Press, 1982), p.11.
34. Schmitter and Karl (note 12, pp.83-4) take this point a step further, explicitly identifying checks and balances as a concept that is 'deliberately excluded from [their] generic definition of democracy' because, 'It is not necessary that the different branches of government be systematically pitted against one another' for democracy to obtain. Constraints on executive authority are more often identified as an issue related to the quality, character, or consolidation of democracy
35. see, for example, Diamond (note 6), pp.98-9,
36. and Guillermo O'Donnell, 'Horizontal Accountability in New Democracies', Journal of Democracy, Vol. 9, No. 3 (1998), pp.112-26.
37. We are aware that our use of the term 'transition' to refer to the instant in which a country crosses a qualitative threshold differs from common scholarly practice, in which the term is used to refer to a 'more or less lengthy period of exceptional politics whose outcome is more or less uncertain'; see Carsten Q. Schneider and Philippe C. Schmitter, 'Liberalization, Transition and Consolidation: Measuring the Components of Democratization', Democratization, Vol. 11, No. 5 (2004), p.68.
38. Readers familiar with Polity will know that it also includes three categorical codings - interruption (-66), interregnum (-77), and transition (-88) - that are used to identify periods when patterns of political authority are either irrelevant or indeterminate. We considered movement into and out of interruption (foreign occupation) and interregnum (collapse of central state authority) to be censoring events rather than transitions, so occurrences such as the installation of democracy in Cambodia in the early 1990s after a period of United Nations intervention are not included in our analysis. In cases coded by Polity as transitional (-88), we used supplemental research to determine whether that period contained an event that met our definition of a transition, and if so, in which year it occurred.
39. Continuous-time models, such as the Cox model, may be applied to data measured at discrete intervals, but those models are likely to produce biased parameter estimates if there are very many ties in the duration data - say, autocracies that endured for the same number of years. Discrete-time models are adequate for the approximation of continuous-time models when the conditional probability of the event's occurrence at any given discrete time point, given that the event has not yet occurred, is about 0.10 or less, and life tables generated from our data set indicate this is indeed the case. For more on discrete-time event history analysis, see Paul D. Allison, Event History Analysis: Regression for Longitudinal Event Data (Newbury Park: Sage Publications), pp.14-22
40. Kazuo Yamaguchi, Event History Analysis (Newbury Park, CA: Sage, 1991), pp.15-70
41. Nathaniel Beck, Jonathan N. Katz, and Richard Tucker, 'Taking Time Seriously: Time-Series-Cross-Section Analysis with a Binary Dependent Variable', American Journal of Political Science, Vol. 42, No. 4 (1998), pp.1260-88
42. and Janet M. Box-Steffensmeier and Bradford S. Jones, Event History Modeling: A Guide for Social Scientists (New York: Cambridge University Press, 2004), pp.69-83 and 87-8.
43. Allison (note 25), p.16.
44. All of these options are discussed in Box-Steffensmeier and Jones (note 25), pp.74-77.
45. See Philip Hougaard, 'Fundamentals of Survival Data', Biometrics, Vol. 55, No, 1 (1999), pp.13-22, and Box-Steffensmeier and Jones (note 25).
46. See Janet M. Box-Steffensmeier and Christopher J.W. Zorn, 'Duration Models and Proportional Hazards in Political Science', American Journal of Political Science, Vol. 45, No. 4 (2001), pp.951-67, for a discussion of the importance of testing for non-proportional hazards and various techniques to do so.
47. We describe this stepwise process as 'informal' because we did not establish strict statistical criteria for exclusion in advance or execute the procedure through a software routine that would produce the same effect.
48. Scott Menard, Applied Logistic Regression Analysis: Second Edition (Thousand Oaks, CA: Sage, 2002), pp.51-6.
49. Seymour Martin Lipset, Political Man: The Social Bases of Politics [expanded ed.] (Baltimore, MD: The Johns Hopkins University Press, 1981), p.51.
50. Among the most influential quantitative studies to document this relationship are Robert W. Jackman, 'On the Relations of Economic Development to Democratic Performance', American Journal of Political Science, Vol. 17, No. 3 (1973), pp.611-21
51. Kenneth A. Bollen, 'Political Democracy and the Timing of Development', American Sociological Review, Vol. 44, No. 4 (1979), pp. 72-87
52. and Ross E. Burkhart and Michael S. Lewis-Beck, 'Comparative Democracy: The Economic Development Thesis', American Political Science Review Vol. 88, No. 4 (1994), pp.903-10.
53. Michael T. Hannan and Glenn R. Carroll, 'Dynamics of Formal Political Structure: An Event-History Analysis', American Sociological Review Vol. 46 (1981), pp. 19-35
54. and Przeworski et al. (note 2), pp. 92-98.
55. Carles Boix and Susan Stokes, 'Endogenous Democratization', World Politics Vol. 55, No. 4 (July 2003), pp.517-49.
56. We measure a country's cumulative experience with democracy as the square root of years as a democracy since independence or 1800, based on the application of our regime categorization scheme to the Polity data. From a substantive perspective, this variable helps to deal with the possibility that first attempts at democracy differ from later attempts; see Alfred Stepan, 'Paths toward Redemocratization: Theoretical and Comparative Considerations', in Guillermo O'Donnell, Philippe C. Schmitter and Laurence Whitehead (eds), Transitions from Authoritarian Rule: Comparative Perspectives (Baltimore, MD: The Johns Hopkins University Press, 1986), pp.64-84
57. and Juan J. Linz, 'Transitions to Democracy', The Washington Quarterly, Vol. 2, No. 2 (Summer 1990), pp.12-34. From a methodological perspective, this variable accounts for the connection between repeated episodes of autocracy in a single country, thereby addressing estimation problems that can arise when the same individual appears more than once in the risk set. The value of this variable does not change during an episode of authoritarian rule, and it is equal to 0 for countries with no democratic experience; in other words, it indicates both whether a country has any democratic experience, and if so, how much.
58. We used infant mortality rates - deaths before age 1 per 1,000 live births, as estimated by the US Bureau of the Census - to measure economic development. To eliminate a longitudinal trend that could bias our estimates, we normalized this measure to the annual global median. To ensure that our results were robust to this choice of measures, we also estimated the same models using non-normalized but logged infant mortality and the more conventional GDP per capita (constant 1995 $US as reported in the World Bank's World Development Indicators, logged) and obtained similar results.
59. On rentier state theory, see especially Hussein Mahdavy, 'The Patterns and Problems of Economic Development in Rentier States: The Case of Iran', in M.A. Cook (ed.), Studies in Economic History of the Middle East from the Rise of Islam to the Present Day (London: Oxford University Press, 1970)
60. Hazem Beblawi and Giacomo Luciani (eds), The Rentier State (New York: Croom Helm, 1987)
61. Terry Lynn Karl, The Paradox of Plenty: Oil Booms and Petro-States (Berkeley, CA: University of California Press, 1997)
62. and Michael L. Ross, 'Does Oil Hinder Democracy?', World Politics, Vol.53, No. 3 (2001), pp.325-61.
63. We tested several measures of natural-resource wealth, including variables from the World Bank's World Development Indicators and the US Department of Energy's Energy Information Administration. The measure we used in the results reported here is derived from World Bank data on income from natural-resource depletion as a share of gross national income (GNI). Specifically, we took the square root of the sum of income from energy resources and mineral sources as a share of GNI.
64. Dahl (note 3), pp.33-47.
65. See Barrington Moore, Social Origins of Dictatorship and Democracy: Lord and Peasant in the Making of the Modern World (Boston, MA: Beacon Press, 1966)
66. and Dietrich Rueschemeyer, Evelyne Huber Stephens, and John D. Stephens, Capitalist Development and Democracy (Chicago, IL: University of Chicago Press, 1992).
67. Apart from the rentier states, many of the world's remaining longtime autocracies are found in Asia and enjoy a measure of direct or indirect support from China (for example, North Korea, Laos, Vietnam, and Burma). If our assessment is correct, then China and its client states rightfully belong in the set of oscillators; they certainly would not yet fit in the 'durable democracy' group, and most of them lack the resource wealth that would make them ideal candidates for the 'durable autocracy' set. Our model suggests that prospects for a democratic transition in China depend less on rising wealth than they do on other dynamics, such as leadership changes, decisions about political liberalization, the persistence of antigovernment mobilization, and the possibility of an economic slowdown, that are certainly linked to, but not determined by, the pace and character of economic development. If and when China does attempt democracy, we would expect the prospects for democratic transition to improve in its client states as well, possibly leaving the rentier states isolated as the last major bastion of durable authoritarian rule.
68. Jacques Delacroix, 'The Distributive State in the World System', Studies in Comparative International Development, Vol. 15, No. 3 (1980), pp.3-21
69. Huntington (note 7), pp.59-72.
70. Eva Bellin, 'The Robustness of Authoritarianism in the Middle East: Exceptionalism in Comparative Perspective', Comparative Politics, Vol. 36, No. 2 (2004), pp.139-57
71. and Bruce Bueno de Mesquita and George W. Downs, 'Development and Democracy', Foreign Affairs Vol. 84, No. 5 (September/ October 2005).
72. Rueschemeyer, Stephens, and Stephens (note 40), p.5.
73. Linz and Stepan (note 12), pp.8-10.
74. The newfound durability of democracy in many Latin American countries - still developing by European standards, but often much wealthier than most of their African counterparts - is provisional testament to this possibility.
75. There are exceptions to this pattern, of course, and practitioners interested in promoting democracy would do well to study them closely for clues as to how they have managed this feat. For a recent and excellent example of just such a study, see M. Stephen Fish, 'Mongolia: Democracy without Prerequisites', Journal of Democracy, Vol. 9, No. 3 (1998), pp.127-41.
76. Douglass C. North, Institutions, Institutional Change and Economic Performance (Cambridge: Cambridge University Press, 1990). A succinct statement on the concept of 'path dependency' can be found at p.112.
77. Guillermo O'Donnell and Philippe C. Schmitter, Transitions from Authoritarian Rule: Tentative Conclusions about Uncertain Democracies (Baltimore, MD: The Johns Hopkins University Press, 1986), p.10.
78. Our finding on collective action is based on variables in the Banks Cross-National Time-Series data set. We tested measures of anti-government demonstrations, general strikes, and riots alone and in combination, as single-year snapshots and three-year cumulative sums, and as any-vs-none dummies and continuous measures. Results were generally consistent across these variations in measurement. In the Banks data set, an anti-government demonstration is defined as 'any peaceful public gathering of at least 100 people for the primary purpose of displaying or voicing their opposition to government policies or authority, excluding demonstrations of a distinctly anti-foreign nature'; a general strike is defined as 'any strike of 1,000 or more industrial or service workers that involves more than one employer and is aimed at national government policies or authority'; and a riot is defined as 'any violent demonstration or clash of more than 100 citizens involving the use of physical force'. All of the relevant variables are based on reports from the daily files of the New York Times.
79. We should point out that our results probably underplay the effect of contentious collective action, due to some unavoidable aspects of our discrete-time statistical design. To ensure that we are not counting turmoil aimed at a newly installed democratic regime, we did not count events that occurred in the same year as the transition to democracy. An 'ocular' review of the data, however, suggests that several autocracies experienced a significant spike in demonstrations in the same year they transitioned. Examples include Albania in 1990 and again in 1997, Madagascar in 1991, and Nepal and Mongolia in 1990. While we suspect that nearly all of those demonstrations targeted the authoritarian regime, we cannot count those events in our analysis without risking errors of commission that would muddy any causal interpretations. Neither does our analysis include the demise of East Germany and the Soviet Union, both of which occurred in the wake of huge spikes in protest activity, because those events are considered cases of state collapse and thus appear in the dataset as right-censored cases (that is, cases where the state disappeared or observation ended without a transition to democracy having occurred).
80. Jay Ulfelder, 'Contentious Collective Action and the Breakdown of Authoritarian Regimes', International Political Science Review, Vol. 26, No. 3 (2005), pp.311-34.
81. See especially Stephan Haggard and Robert R. Kaufman, The Political Economy of Democratic Transitions (Princeton, NJ: Princeton University Press, 1995)
82. Geddes (note 13)
83. and Daron Acemoglu and James A. Robinson, 'A Theory of Political Transitions', American Economic Review, Vol. 91, No. 4 (2001), pp.938-63.
84. For studies emphasizing the effects of international forces on transitions to democracy, see Laurence Whitehead, 'International Aspects of Democratization', in Guillermo O'Donnell, Philippe C. Schmitter and Laurence Whitehead (eds), Transitions from Authoritarian Rule: Comparative Perspectives (Baltimore: The Johns Hopkins University Press, 1986), pp.3-46
85. Huntington (note 10), pp.85-106
86. Kristian Skrede Gleditsch and Michael D. Ward, 'Diffusion and the International Context of Democratization', paper presented at 'The International Diffusion of Democracy and Markets' conference, University of California, Los Angeles, CA (7-8 March 2003)
87. Antoaneta Dimitrova and Geoffrey Pridham, 'International Actors and Democracy Promotion in Central and Eastern Europe: The Integration Model and Its Limits', Democratization, Vol. 11, No. 5 (2004), pp.91-112
88. Barbara Wejnert, 'Diffusion, Development, and Democracy: 1800-1999', American Sociological Review, Vol. 70, No. 1 (2005), pp.53-81.
89. Steven Levitsky and Lucan Way, 'International Linkage and Democratization', Journal of Democracy, Vol. 16, No. 3 (2005), pp.20-34.
90. With a p-value of 0.09 when adjusting for other model variables, trade openness (measured as imports plus exports as a share of Gross Domestic Product) showed the strongest evidence of a link, but the effect was opposite the predicted one, suggesting that, if anything, countries less open to foreign trade were more likely to transition in the past 30 years, other things being equal.
91. Protests undertaken in 2005 in countries such as Azerbaijan, Egypt, Kyrgyzstan, and Lebanon by demonstrators citing Ukraine's 'Orange Revolution' as one source of inspiration offer anecdotal evidence to support the notion that demonstration and contagion effects might operate through this particular mechanism.
92. See note 53.
93. Daniel Brinks and Michael Coppedge, 'Patterns of Diffusion in the Third Wave of Democracy', paper prepared for presentation at the Annual Meeting of the American Political Science Association, San Francisco, CA (September 2001).
94. The seven cases of backsliding after 15 or more consecutive years of democracy that we observe are Brazil 1964, Chile 1973, Fiji 1987, The Gambia 1994, Peru 1968, Philippines 1972, and Uruguay 1973. Using a different set of cases and measures, Mark J. Gasiorowski and Timothy J. Power detect a similar drop off in the risk of democracy's demise, albeit after 12 instead of 14 years; see 'The Structural Determinants of Democratic Consolidation: Evidence from the Third World', Comparative Political Studies Vol. 31, No. 6 (1998), pp.740-71.
95. See especially Przeworski et al. (note 2).
96. Our infant mortality data were provided by the US Census Bureau. They are based on figures reported every five years by the United Nations, but they have been interpolated and adjusted to account for expected underreporting as well as calamities such as famine, drought, or war. To avoid conflating broader temporal trends with any effects from economic development, we considered using a normalized version of this variable, but we chose to rely on the raw estimates (logged) in order to test modernization theory, which emphasizes absolute levels rather than relative changes. When we added other measures of wealth or modernization to a base model that included infant mortality, we found that infant mortality remained highly significant while the other variable was generally not significant, and its addition failed to improve the fit of the model.
97. According to our data set, Angola had the highest infant mortality rate in 2003 with 193.8 deaths before age one per 1,000 live births, while Singapore had the lowest with just 2.3 deaths per 1,000 live births.
98. Przeworski et al. (note 2), pp.101-6.
99. In 'Federalist No. 10', Madison defines a faction as 'a number of citizens, whether amounting to a majority or a minority of the whole, who are united and actuated by some common impulse of passion, or of interest, adverse to the rights of other citizens, or to the permanent and aggregate interests of the community'; see 'Federalist Paper No. 10', in Alexander Hamilton, James Madison and John Jay, The Federalist Papers (New York: Mentor Books, 1961), p.78. In Polity, factionalism is identified by a coding of 3 on a variable measuring the competitiveness of political participation (PARCOMP). While PARCOMP is basically scalar - 1 indicates fully repressed participation and 5 indicates fully open participation - the factional coding is probably best considered indicative of movement along an additional dimension related to the extent of political polarization. As such, we believe it is best treated in quantitative analyses as a separate indicator variable, as we do here.
100. Ibid., p.80.
101. Madison identifies the combination of representative government and federalism as a check against factionalism in large societies, but the occurrence of the American Civil War suggests it is an imperfect solution at best. Recent scholarship has focused on the electoral system as a critical set of institutions that can encourage or discourage factionalism in ethnically or religiously diverse societies; for more on this subject, see Benjamin Reilly, 'Engineering Parties and Party Systems', paper presented at the Annual Meeting of the American Political Science Association, Philadelphia, PA (2003)
102. and Pippa Norris, Electoral Engineering: Voting Rules and Political Behavior (Cambridge: Cambridge University Press, 2004).
103. See for example Guillermo O'Donnell, 'Horizontal Accountability in New Democracies', Journal of Democracy, Vol. 9, No. 3 (1998), pp.112-26
104. and Wolfgang Merkel, 'Embedded and Defective Democracies', Democratization, Vol. 11, No. 5 (2004), pp.33-58.
105. Witold J. Henisz, 'The Institutional Environment for Infrastructure Investment', Industrial and Corporate Change, Vol. 11, No. 2 (2002), pp.355-389.
106. Diamond (note 6), pp.93-112.
107. Wolfgang Merkel, 'Embedded and Defective Democracies', Democratization Vol. 11, No. 5 (2004), pp.33-58.
108. Adam Przeworski, Michael Alvarez, José Antonio Cheibub, and Fernando Limongi, 'What Makes Democracy Endure?' Journal of Democracy, Vol. 7, No. 1 (1996), pp.39-55.
109. See also Juan J. Linz and Arturo Valenzuela (eds), The Failure of Presidential Democracy (Baltimore, MD: The Johns Hopkins University Press, 1995).
110. To distinguish presidential from parliamentary democracies, we relied on a Banks variable that identifies the type of executive, recoding it as a dummy variable equal to '1' where the effective chief executive was a president and equal to '0' in all other situations, most of which identify a premier as the person in charge.
111. Przeworski et al. (note 72), p.44-45.
112. Fourteen onsets occurred during ongoing episodes of democracy, and 10 others occurred in democracies established following periods of state collapse or foreign occupation.
113. N = 95 (46 factional): Infant mortality, logged = 0.99, p 0.01; new country dummy = 22.39, p 0.01; presidential system = -0.41, p 0.42. We obtain essentially identical results when we add a categorical variable to control for prior episodes of democracy.
114. John M. Carey, 'Institutional Design and Party Systems', in Larry Diamond, Marc F. Plattner, Yun-han Chu, and Hung-mao Tein (eds), Consolidating the Third Wave of Democracies: Themes and Perspectives (Baltimore, MD: The Johns Hopkins University Press, 1997), p.80.
115. A similar point is made by George Tsebelis, Veto Players: How Political Institutions Work (Princeton, NJ: Princeton University Press, 2002).
116. This scenario might arise, for example, if autocrats considering their chances of becoming competitively elected presidents accounted for their prospects for electoral victory in their calculations about whether to support a transition to democracy, and if so, what kind of democracy to install. Potential presidents who can expect to win and sustain an electoral majority well into the future would seem more likely to endorse elections that are procedurally competitive than those whose electoral support is less certain. If those expectations are reasonably accurate, then we might see a selection effect whereby presidential systems are more likely to be installed in situations where alternation is likely to be delayed. In some of those cases, the president might subvert democracy when his electoral mandate finally does wear out, or democracy might die for other reasons before alternation occurs; under either scenario, however, the episode of plausible democracy would be treated as a period of dictatorship by Przeworski et al. (note 2), and the average age of presidential democracies at their death would be reduced.
117. Przeworski et al. (note 72), p.48. Our negative finding is consistent with the conclusions from Scott Mainwaring and Matthew Soberg Shugart (eds), Presidentialism and Democracy in Latin America (New York: Cambridge University Press, 1997).
118. Larry Diamond and Juan J. Linz, 'Introduction: Politics, Society, and Democracy in Latin America', in Larry Diamond, Juan J. Linz, and Seymour Martin Lipset (eds), Democracy in Developing Countries: Latin America (Boulder, CO: Lynne Rienner, 1989), p.17.
119. See also Juan J. Linz, The Breakdown of Democratic Regimes: Crisis, Breakdown, Reequilibration (Baltimore, MD: The Johns Hopkins University Press, 1978) and Larry Diamond (note 6) pp.78-88.
120. See for example Przeworski et al. (note 2), especially pp.106-12 and Table 2.17 on p.124
121. Gasiorowski and Power (note 60).
122. Margaret E. Keck and Kathryn Sikkink, Activists beyond Borders:
123. Amartya Sen, 'Democracy as a Universal Value', Journal of Democracy Vol. 10, No. 3 (1999), pp.3-17.
124. Michael McFaul, 'Democracy Promotion as a World Value', The Washington Quarterly, Vol. 28, No. 1 (2004), p.156.
125. For purposes of this exercise, we followed the same rule we used in our postdictive accuracy assessments, basing the predicted status of each case on a cut point that most closely balanced sensitivity (the proportion of true positives) with specificity (the proportion of true negatives) in the historical data from which each iterative model was estimated. This is not the only way to set a cut point, however, and a different decision rule would produce different results. For example, given the rarity of regime transitions, we could achieve better than 90 per cent overall accuracy in most years by simply predicting no events. Of course, we think that approach is less informative than one that makes some attempt to identify likely cases, so we adopted a rule that we think is more intuitive.