Open Collections

UBC Theses and Dissertations

UBC Theses Logo

UBC Theses and Dissertations

Undertow in the third wave : understanding the reversion from democracy Ferguson, Peter Allyn 2009

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
24-ubc_2009_spring_ferguson_peter.pdf [ 1.38MB ]
Metadata
JSON: 24-1.0067105.json
JSON-LD: 24-1.0067105-ld.json
RDF/XML (Pretty): 24-1.0067105-rdf.xml
RDF/JSON: 24-1.0067105-rdf.json
Turtle: 24-1.0067105-turtle.txt
N-Triples: 24-1.0067105-rdf-ntriples.txt
Original Record: 24-1.0067105-source.json
Full Text
24-1.0067105-fulltext.txt
Citation
24-1.0067105.ris

Full Text

UNDERTOW IN THE THIRD WAVE: UNDERSTANDING THE REVERSION FROM DEMOCRACY  by  PETER ALLYN FERGUSON   A THESIS SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF  DOCTOR OF PHILOSOPHY  in  THE FACULTY OF GRADUATE STUDIES  (Political Science)   THE UNIVERSITY OF BRITISH COLUMBIA  (Vancouver)      February 2009  © Peter Allyn Ferguson ii  Abstract  In 2007, the world suffered a net decline in freedom for the second successive year for the first time in fifteen years.  There are indications of global democratic stagnation. Coups and democratic reversions continue to occur.  Why do regimes sometimes experience reversions away from democracy?  An analysis of data from 1972-2003 indicates that for every $1 increase in Gross Domestic Product (GDP) per capita, the odds of a democratic reversion decrease 0.2%; for each 1% increase in GDP growth, the odds of a democratic reversal decrease 9.2%; and, for each 1 unit increase in Consumer Price Index (CPI), there is a 4.1% increase in the likelihood of democratic reversion.  When the analysis is limited strictly to a comparison of democratic reversion cases and ongoing democratic regimes, variables addressing political institutional configurations, vulnerabilities to international pressures and civilian control over the military are either insignificant or provide very little purchase for explaining variance on the dependent variable.  The dissertation includes thirty case studies of reversions from democracy, representing one universe of such cases from 1975-2003.  Based on an analysis of these cases, several conclusions may be drawn.  On economic issues, the case studies indicate we should be cautious in overstating the importance of economic performance and they draw attention to the problematic nature of analyses based on one year lags.  The importance of legislative gridlock, particularly during an economic crisis is highlighted. High levels of legislative fractionalization are found to increase reversion risks.  Younger democracies are also found to be more vulnerable, as each additional year a democratic regime is in existence decreases reversion risks by 3.8%.  The consideration of international influences on costs associated with reversion decisions is found to be relevant.  The case analysis indicates attempts to assert civilian control over the military are likely to increase reversion risks.  Based on a rational choice analysis and a case study of the Philippines, higher levels of democratic uncertainty are found to reduce reversion risks by allowing actors to tolerate lower levels of goods in light of the potential for future democratic change.  iii  Table of Contents  Abstract ............................................................................................................................... ii Table of Contents ............................................................................................................... iii List of Tables ..................................................................................................................... vi List of Figures .................................................................................................................. viii Acknowledgements ............................................................................................................ ix Dedication .......................................................................................................................... xi Chapter 1: Introduction ....................................................................................................... 1 Introduction ..................................................................................................................... 2 The Relevance of Democratic Reversion Research ........................................................ 5 The Limits of Consolidation Research ......................................................................... 11 State of the Democratic Reversion Literature ............................................................... 15 The Competition between Structure and Choice .......................................................... 20 The Plan ........................................................................................................................ 23 Chapter 2: Theoretical Orientation and Methodology ...................................................... 29 Introduction ................................................................................................................... 30 Economic Influences ..................................................................................................... 31 Political Institutional Influences ................................................................................... 35 International Influences ................................................................................................ 37 Military Influences ........................................................................................................ 39 Democratic Uncertainty ................................................................................................ 42 Data and Methods ......................................................................................................... 44 Conclusion .................................................................................................................... 59 Chapter 3: Economic Issues and Democratic Reversion .................................................. 72 Introduction ................................................................................................................... 73 Cross-National, Time-Series Analysis .......................................................................... 76 Case Study Data Analysis ............................................................................................. 85 Case Level Analysis ...................................................................................................... 96 iv  Conclusions ................................................................................................................. 117 Chapter 4: Political Institutions and Democratic Reversion ........................................... 124 Introduction ................................................................................................................. 125 Cross-National, Time-Series Analysis ........................................................................ 127 Case Study Data Analysis ........................................................................................... 138 Case Level Analysis .................................................................................................... 140 Conclusions ................................................................................................................. 145 Chapter 5: International Influences and Democratic Reversion ..................................... 155 Introduction ................................................................................................................. 156 Cross-National, Time-Series Analysis ........................................................................ 159 Case Study Data Analysis ........................................................................................... 161 Case Level Analysis .................................................................................................... 165 Conclusion .................................................................................................................. 181 Chapter 6: The Military and Democratic Reversion ....................................................... 188 Introduction ................................................................................................................. 189 Civilian Control .......................................................................................................... 193 Cross-National, Time-Series Analysis ........................................................................ 195 Case Study Data Analysis ........................................................................................... 198 Case Level Analysis .................................................................................................... 203 Conclusion .................................................................................................................. 223 Chapter 7: Democratic Uncertainty and Democratic Reversion ..................................... 231 Introduction ................................................................................................................. 232 Uncertainty in the Democratization Literature ........................................................... 234 Recent Interpretations of Uncertainty ......................................................................... 237 Flaws in Existing Interpretations ................................................................................ 241 Rehabilitating Democratic Uncertainty ...................................................................... 244 A Rational Choice Account ........................................................................................ 248 Discussion ................................................................................................................... 254 Conclusion .................................................................................................................. 257 Chapter 8: Democratic Uncertainty and Regime Stability in the Philippines ................ 260 v  Introduction ................................................................................................................. 261 Historical Background ................................................................................................ 262 Economic Shocks ........................................................................................................ 263 Military Threats .......................................................................................................... 266 Political Institutional Threats ...................................................................................... 270 International Threats ................................................................................................... 272 Additional Threats ...................................................................................................... 274 The Role of Democratic Uncertainty in the Philippines ............................................. 275 Conclusion .................................................................................................................. 288 Chapter 9: Conclusion..................................................................................................... 296 Introduction ................................................................................................................. 297 Economic Conditions .................................................................................................. 299 Political Situation ........................................................................................................ 301 International Influences .............................................................................................. 302 Civilian Control of the Military .................................................................................. 303 Democratic Uncertainty .............................................................................................. 305 The Philippines ........................................................................................................... 307 Conclusion .................................................................................................................. 308 Bibliography ................................................................................................................... 310   vi  List of Tables Table 2.0.1: Cases Losing “Free” Rating.......................................................................... 62 Table 2.0.2: Democratic Reversion Cases ........................................................................ 63 Table 2.0.3: Democratic Reversion Event Descriptions ................................................... 64 Table 3.0.1: Economic Development - Regime Type Rated 2-14 .................................. 119 Table 3.0.2: Economic Development - Regime Type Rated Free, Partly Free, Not Free ......................................................................................................................................... 119 Table 3.0.3: Economic Development - Democratic Reversions versus All Remaining Country Years ................................................................................................................. 120 Table 3.0.4: Economic Development - Democratic Reversions versus Democratic Country Years ................................................................................................................. 120 Table 3.0.5: GDP per capita (30 Reversion Cases) ........................................................ 121 Table 3.0.6: GDP per capita Growth (30 Reversion Cases) ........................................... 122 Table 3.0.7: Inflation (30 Reversion Cases) ................................................................... 123 Table 4.0.1: Political Institutions – Four Dependent Variables ...................................... 148 Table 4.0.2: Presidential versus Parliamentary System (30 Reversion Cases) ............... 149 Table 4.0.3: Electoral Rules – Plurality? ........................................................................ 150 Table 4.0.4: Electoral Rules – Proportional Representation? ......................................... 151 Table 4.0.5: The Legislature – Herfindahl Index Total .................................................. 152 Table 4.0.6: The Legislature – Fractionalization Index Total ........................................ 153 Table 4.0.7: Tenure of System or Chief Executive ......................................................... 154 Table 5.0.1: International Influences - Regime Type Rated 2-14 ................................... 183 Table 5.0.2: International Influences - Regime Type Rated Free, Partly Free, Not Free 183 Table 5.0.3: International Influences - Democratic Reversions versus All Remaining Country Years ................................................................................................................. 184 Table 5.0.4: International Influences - Democratic Reversions versus Democratic Country Years ................................................................................................................. 184 Table 5.0.5: Trade, % GDP (30 Reversion Cases) ......................................................... 185 Table 5.0.6: Trade in Services, % GDP (30 Reversion Cases) ....................................... 186 Table 5.0.7: Aid, % GNI (30 Reversion Cases) .............................................................. 187 Table 6.0.1: Civil Military Relations - Regime Type Rated 2-14 .................................. 226 Table 6.0.2: Civil Military Relations - Regime Type Rated Free, Partly Free, Not Free 226 Table 6.0.3: Civil Military Relations - Democratic Reversions versus All Remaining Country Years ................................................................................................................. 227 Table 6.0.4: Civil Military Relations - Democratic Reversions versus Democratic Country Years ................................................................................................................. 227 Table 6.0.5: Military Expenditures, % GDP (30 Reversion Cases) ................................ 228 Table 6.0.6: Military Personnel (30 Reversion Cases) ................................................... 229 Table 6.0.7: Civilian Defense Minister (30 Cases) ......................................................... 230 vii  Table 7.0.1: Relationship between level of uncertainty and policy outcomes ................ 241 Table 8.0.1: The Philippines -  Economic Data .............................................................. 292 Table 8.0.2: The Philippines – Military Data ................................................................. 293 Table 8.0.3: The Philippines - Military Officers in the Government.............................. 294 Table 8.0.4: The Philippines - Political Institutional Indicators ..................................... 295   viii  List of Figures  Figure 8.1: Polity Scores for the Philippines ……………………………………290 Figure 8.2: Real Income Growth, Philippines 1961-1994 ………………………291 ix  Acknowledgements  I owe numerous people thanks and my heartfelt appreciation for the support I received in completing this dissertation.  I have to start with Maxwell Cameron.  Without the ongoing support of Max, I would have been unable to complete the dissertation.  I owe many thanks to the other members of my dissertation committee: Lisa Sundstrom and Ken Carty.  I also appreciate the efforts of the members of the dissertation examining committee: Lenora Angeles (UBC), Ben Nyblade (UBC) and Kirk Bowman (Georgia Institute of Technology).  There were many other members of the UBC Political Science community whose support I will not soon forget.  In particular, I really appreciate the support I received from Lesley Burns, Scott Matthews, Kathryn Harrison and the Koerner’s Study Group. Outside of UBC, I owe a number of thanks for academic support.  The priests and teachers at Jesuit Prep in Dallas prepared me to think and write.  Loyola trained me to research, construct and present arguments.  In particular, I appreciate the support of my coaches, Jay and Barbara Busse, and my debate partners, John Doran, PJ Graham and Ryan Mills.  At Texas A&M, I learned to be a student.  I owe a debt to Jon Bond, who took a chance on me, and to Gretchen Casper, who taught me about democratization.  As well, I thank my Bolton Hall compatriots who provided me with support, especially Alex Tan, Rob Bohrer, Mike Rosenstiel, Chris Hanson, Angie Hinton-Anderson, Rachel Gibson and the BOFFLE gang.  The people at the University of Saskatchewan were essential to getting my academic career back on track.  In particular, I owe a debt to David Smith and John Courtney.  A lot of friends in Saskatoon saw me through some tough dissertation times, especially my Thursday night study group, my very good friend John Hyshka and the Saskatoon Kinsmen.  The Department of Political Science at The University of Western Ontario is now my home.  I owe a particular debt of gratitude to two UWO research assistants, Mark Mitchell and Diala al Dabbas.  My departmental colleagues have been more supportive than I could describe.  Rather than name names, I will just express my utmost appreciation and pledge to be an excellent colleague for years to come. Completing this project would not have been possible without the encouragement of my long-time friends and my family.  I would not be who I am if I did not count John Kessler and Larry and Kathy James as my friends.  Jamie Glean is the best friend a person could ever have.  I could always count on him to remind me to press on and breeze through things.  My in-laws, the Wolfe/Diener family, are an amazing group of people who I knew were in my corner.  My brothers, Jack and Mike, and their families have been there for me when I was in need.  My dad didn’t live to see this but he would have gotten a kick out of it all.  But in the end, there are three people that are most responsible for making me a good person and for whatever successes I experience in life. x  My mom, Terry, has been a rock.  She sacrificed no end to see me succeed.  She always believed in me and made it possible for me to believe in myself.  My wife, Cristine de Clercy, is the love of my life.  I would never have finished this dissertation without her. She made the sacrifice worthwhile.  And finally, the person who never fails to bring a smile to my face is my son Colin.  When you live with the happiest person on the face of the earth, it is easy to see the path forward.  Thanks to you all.  xi  Dedication        For Cris, Colin and Mama.  1  Chapter 1: Introduction 2  Introduction  Regimes occasionally experience reversions away from democracy.  Why?  This is the motivating question that drives the research in this dissertation.  A great deal of attention has been devoted to understanding why countries experience transitions to democracy and why those countries may or may not eventually become consolidated democracies.  Relatively less attention has been devoted to why countries experience reversions away from democracy.  As the number of these anomalous cases of reversion mount and as the world increasingly appears to be entering a phase of democratic stagnation, research examining democratic reversions becomes increasingly relevant. To address this question, the dissertation introduces a framework for explaining the reversion from democracy which relies on the notion that actors base their decision- making around support for democratic regimes on traditional pocketbook issues.  The dissertation draws on a number of the most widely employed theories in the study of democratization in order to examine them in the specific context of democratic reversions.  The concept of democratic uncertainty is introduced as an additional theoretical element for consideration.  The baseline argument uniting these various theoretical elements is that actors are concerned with the ability of democratic regimes to deliver an acceptable level of goods.  When regimes fail to deliver goods at an acceptable level, actors withdraw their support for the regime and the potential for a democratic reversion increase.  These actors are mainly interested in economic goods.  As such the dissertation begins with an explanation of the importance of both the level of economic development as well as economic performance.  Countries at higher levels of economic 3  development are less vulnerable to reversion.  As well, countries with higher levels of economic performance (in terms of economic growth and inflation rates) are less vulnerable to reversion.  A country’s political institutional configuration is also relevant to the issue of democratic reversion.  Here it is argued that when governments fall into legislative gridlock, the country is more vulnerable to democratic reversion because the government’s ability to address pocketbooks issues is perceived as compromised.  In this regard, executive structures, the political party systems and electoral rules are taken into consideration and presidential systems, systems with higher party fractionalization, and systems employing proportional representation voting are seen as more vulnerable to democratic reversion. Further, a country’s vulnerability to international pressure is considered.  As international pressure can directly affect pocketbook issues, countries that are less vulnerable to such pressures are more likely to experience democratic reversions.  As well, as the military is the most common veto player in a regime change situation, the pocketbook issues of the military are considered.  While democratic regimes are often counseled to assert civilian control over the military, this is often accomplished in a manner that affects the military’s pockets (its budget or personnel levels).  As such, the countries that confront the military in an effort to assert increasing civilian control are more vulnerable to democratic reversion. Finally, the idea of democratic uncertainty is introduced as a means to further understand what levels of goods are acceptable as well as what time horizon actors consider when evaluating their support for democratic regimes.  When actors perceive higher levels of democratic uncertainty, they are more willing to tolerate a lower level of 4  goods and take a longer time horizon into account when evaluating their regime support. So countries with lower levels of democratic uncertainty are more likely to experience democratic reversion. This approach makes five contributions to the democratization field.  First, whereas democratization research over the last two decades has substantially concentrated on examining the transition to or consolidation of democracy, the dissertation is explicitly oriented to the study of the reversion from democracy.  This orientation is important as such reversions continue to occur and seem to be of increasing global relevance.  Second, a number of the most important theories in the study of democratization are examined for their relevance to democratic reversions.  It is widely recognized that the elements involved in achieving democracy and sustaining it may not be the same.  These theories are thus tested across both time and space with an explicit eye to the relevance they hold for explaining democratic reversions.  In addition, the theories are tested against one universe of democratic reversion cases.  To accomplish this, various chapters draw on a set of thirty case studies of democratic reversion between 1975 and 2004 undertaken as part of the research for this dissertation.  Third, rather than exclusively employing a structural or process-based approach, the dissertation draws on both in an attempt to provide a more complete picture of democratic reversion.  Fourth, the attempt to include the idea of democratic uncertainty into an explanation of reversion represents a promising step toward addressing the unique benefits that democratic regimes may offer actors considering whether to withdraw their support for democracy. In this regard, the work on democratic uncertainty also explicitly draws into question the reliance on one year lags to analyze the impact of structural factors on the endurance of 5  democratic regimes.  Finally, the chapter length case study of the Philippines allows us to examine a framework that draws on both the structural and process approaches found in the literature in light of a case that did not revert from democracy despite numerous indications that it probably should not have remained democratic. The Relevance of Democratic Reversion Research  Given the explosive development of democracy over the last four decades, it seems reasonable to wonder if research examining the reversion away from democracy remains relevant.  It is important to remember that even though military coups occur with less frequency than they did, for example, in the 1960s and 1970s, militaries continue to oust democratically elected civilian governments from power.  And while the Fujimori autogolpe in Peru seemed in the 1990s like a novel way for a democracy to fall, the experience, for example, of post-Soviet states in Central Asia indicates executive seizures of power are not as novel as once thought.  Further, the stagnation of democratic development is an issue of increasing importance across a variety of regions in the world. In fact, as pointed out by Larry Diamond in the headline March/April 2008 Foreign Affairs article, Alarmingly, a January 2008 Freedom House survey found that for the first time since 1994, freedom around the world has suffered a net decline in two successive years.  The ratio of the number of countries whose scores had improved to the number whose scores had declines – a key indicator – was the worst since the fall of the Berlin Wall. 6  So it appears that there is some kind of a substantial, and potentially growing, undertow to the democratic wave we have experienced over the last four decades.  As such, it seems highly relevant to further explore democratic reversion. Since the beginnings of democratization research in the late 1950s, both research questions and analytical frameworks have shifted over time.  The issue driving research has moved from democratic preconditions to breakdown to transition to consolidation. The analytical framework has drifted between structural and process orientations.  One means of identifying these changes is to orient them with Huntington’s (1990, 16-26) democratic waves. Samuel Huntington (1990) argues there is an observable historical pattern of global political change that occurs in waves.  These waves of democratization are groups of transitions from non-democratic to democratic regimes that occur within a specified period of time and that significantly outnumber transitions in the opposite direction. Further, Huntington (1990, 15-6) points out that “history is also not unidirectional.  Each of the first two waves of democratization was followed by a reverse wave in which some but not all of the countries that had previously made the transition to democracy reverted to non-democratic rule.”  He (1990, 16-26) points to three waves of democratization. The first wave lasted from 1828-1926 and was reversed from 1922-1942.  The second wave lasted from 1943-1962 and was reversed from 1958-1975.  The third wave began in 1974 (with the end of dictatorship in Portugal) and continues today. Shifts in the research questions orienting the democratization literature have loosely tracked Huntington’s democratic waves.  In response to post-WWII transitions 7  (second wave), research in the 1960s was concerned with a search for preconditions to democracy.1  Confronted with the second reverse wave, attention in the 1970s turned to explaining the causes of democratic breakdowns.2  Just as a consensus had been achieved on the undemocratic and statist direction of change in the third world (Remmer 1997, 42- 3), the rising number of anomalies created by the growing third wave confronted scholars and research attention shifted to the causes of democratic transitions.3  In the early 1990s,  1 See, for example, Seymour Martin Lipset, “Some Social Requisites of Democracy: Economic Development and Political Legitimacy,” American Political Science Review 53 (1959); Samuel P. Huntington, Political Order in Changing Societies (New Haven: Yale University Press, 1968); Gabriel Almond and Sidney Verba, The Civic Culture (Princeton: Princeton University Press, 1963); Barrington Moore, Social Origins of Dictatorship and Democracy (Boston: Beacon Press, 1966); Robert A. Dahl, Polyarchy: Participation and Opposition, (New Haven: Yale University Press, 1971); Phillips Cutright, “National Political Development: Measures and Analysis,” American Sociological Review 28 (1963); Leonard Binder, Crises and Sequences in Political Development (Princeton: Princeton University Press, 1971). 2 See, for example, Juan J. Linz, “Crisis, Breakdown, and Reequilibration,” in The Breakdown of Democratic Regimes, ed. Juan J Linz and Alfred Stepan (Baltimore: Johns Hopkins University Press, 1978); Juan J. Linz and Alfred Stepan, eds., The Breakdown of Democratic Regimes: Europe (Baltimore: Johns Hopkins Press, 1978); Juan J. Linz and Alfred Stepan, eds., The Breakdown of Democratic Regimes: Latin America (Baltimore: Johns Hopkins Press, 1978);  Arturo Valenzuela, The Breakdown of Democratic Regimes: Chile (Baltimore: Johns Hopkins Press, 1978);  Jyotirindra Das Gupta, “A Season of Caesars,” Asian Survey 28 (April 1978); Martin C. Needler, “Military Motivations in the Seizure of Power,” Latin American Research Review 10 (Fall 1975); Youssef Cohen, “Democracy from Above: The Political Origins of Military Dictatorship in Brazil,” World Politics 60 (October 1987); Guillermo A. O’Donnell, Modernization and Bureaucratic-Authoritarianism: Studies in South American Politics (Berkeley: University of California Press 1973); Karen L. Remmer and Gilbert W. Merkx, “Bureaucratic-Authoritarianism Revisited,” Latin American Research Review 17 (No. 2 1982). 3 See, for example, Guillermo O’Donnell, Philipe C. Schmitter and Laurence Whitehead, eds., Transitions From Authoritarian Rule: Comparative Perspectives (Baltimore: The Johns Hopkins University Press, 1986); Guillermo O'Donnell and Philippe Schmitter, Transitions From Authoritarian Rule: Tentative Conclusions About Uncertain Democracies (Baltimore: Johns Hopkins University Press, 1986); Terry Lynn Karl and Philipe C. Schmitter, “Modes of Transition in Latin America, Southern and Eastern Europe,” International Social Science Journal 143 (May 1991); James Malloy and Mitchell A. Seligson, eds., Authoritarians and Democrats: Regime Transition in Latin America (Pittsburgh: University of Pittsburgh Press, 1987); Larry Diamond, Juan J. Linz, and Seymour Martin Lipset, Democracy in Developing Countries (Boulder: Lynne Rienner, 1988); Enrique Baloyra, Comparing New Democracies: Transition and Consolidation in Mediterranean Europe and the Southern Cone (Boulder: Westview Press, 1987). 8  many researchers recognized the need to distinguish completion of the transition phase from the ongoing struggle for democratic permanence and shifted their attention to the issue of democratic consolidation.4  More recently, a number of academics, mainly interested in Latin American politics, have explored issues surrounding autogolpes (presidential self-coups) in Peru and Guatemala.5 Recent trends in the third wave also point to the need for research oriented to the study of the reversion from democracy.  It is widely recognized that democracies, even consolidated one, can experience reversions.6  Diamond (1997, xvii), utilizing Freedom House indicators, points out that freedom levels in many third wave democracies has actually declined over the last decade.  Huntington (1997, 5) argues that each of the first two waves has been followed by a reverse wave and that there are now “indications that a new reverse wave may be gathering.”  Power and Gasiorowski (1997, 135) argue, “it is  4 See, for example, John Higley and Richard Gunther, eds., Elites and Democratic Consolidation in Latin American and Southern Europe (Cambridge: Cambridge University Press, 1992); Scott Mainwaring, Guillermo O'Donnell and J. Samuel Valenzuela, eds., Issues in Democratic Consolidation: The New South American Democracies in Comparative Perspective (Notre Dame: University of Notre Dame Press, 1992); Juan J. Linz, “Transitions to Democracy,” Washington Quarterly 13:3 (1990); Larry Diamond, Marc F. Plattner, Yun-Han Chu, and Hung-Mao Tien, eds., Consolidating The Third wave Democracies (Baltimore: The Johns Hopkins University Press, 1997). 5 See, for example, Maxwell A. Cameron, “Self-Coups: Peru, Guatemala, and Russia,” Journal of Democracy 9.1 (1998), p. 125-139; Maxwell A. Cameron, “Latin American Autogolpes: Dangerous Undertows in the Third Wave of Democratisation,” Third World Quarterly 19.2 (1988), p. 219-239; Philip Mauceri, “State Reform, Coalitions, and the Neoliberal Autogolpe in Peru,” Latin American Research Review 30.1 (1995), p. 7-37. 6 See, for example, Doh Chull Shin, “On the Third wave of Democratization: A Synthesis and Evaluation of Recent Theory and Research,” World Politics 47 (October 1994), p. 144; Timothy J. Power and Mark Gasiorowski, “Institutional Design and Democratic Consolidation in the Third World,” Comparative Political Studies 30:2 (April 1997), p. 133, Juan J. Linz and Alfred Stepan, “Toward Consolidated Democracies,” in Larry Diamond, Marc F. Plattner, Yun-Han Chu, and Hung-Mao Tien, eds., Consolidating The Third wave Democracies (Baltimore: The Johns Hopkins University Press, 1997), p. 16. 9  legitimate to analyze the development of relatively young polyarchies - to study the infant mortality rates of Third World democracy, if you will.  Just as pediatrics is far more essential to Third World medicine than is geriatrics, so must students of democratization sometimes adapt their analytic tools to an environment where the odds for successful democracy have frankly not been very favorable.”  Recent events in Latin America, ranging from the autogolpe cases to forced/unexpected presidential resignations in a number of countries, seem to bear out this point. In the past, democratization research seems to have gotten caught up in the enthusiasm of the current global democratic trend and ended up lacking the “theoretical equipment” to deal with reversals in those trends (Remmer 1997, 42-3).  For example, the work in the 1960s searching for preconditions was triggered by the second democratic wave.  Yet, by the time Lipset’s founding article in this area was published in 1959, the second wave was already 16 years old and only had 3 years remaining.  In fact, by 1958, the second reverse wave had already begun and would last through 1975.  The point is not the exact dates.  The point is that the bulk of the work on preconditions, aiming at explaining transitions to democracy took place during the second reverse wave.  Initially the reversion cases could be viewed as anomalies but at some point there were too many anomalies for theories to accommodate.  Research in this area then entered a chaotic period finally shifting to an examination of democratic breakdown.  Once again, almost all the research on breakdown actually occurs after the second reverse wave had ended and the third democratic wave had actually begun.  Transition cases were treated as anomalies until theory could no longer accommodate their rising numbers.  Again, the 10  literature was thrown into a period of chaos until shifting to an examination of democratic transitions (Remmer 1997, 42-3). On its face, these observations could probably be made about any research area in the discipline.  We almost always study phenomena that have already taken place and theory is rarely able to accommodate radical shifts in reality.  What gives this criticism teeth is the fact that people recognize that democratic reversions occur even during a democratic wave, yet little work in this area explicitly attempts to understand democratic reversions that have occurred during the third wave as a phenomena in and of itself.  The objective of the bulk of the research in this area is to explain transitions and consolidation.  This, despite the fact that third wave democratic reversions continue to occur and Diamond (2008) is arguing we are experiencing a global “democratic rollback” while Samuel Huntington (1997, 5) took the position over a decade ago that we may have already entered a third reverse wave. Should such a reverse wave fail to materialize, research on democratic reversions would still advance the literature.  Reversions have occurred and will likely continue to occur during the third wave.  Understanding why reversions occur and what factors mitigate the risk of breakdown will be useful to the democratization literature because the processes of democratic consolidation and democratic reversion are distinct.  Linz and Stepan (1997, 16) point out that breakdowns are, “related not to weaknesses or problems specific to the historic process of democratic consolidation, but to a new dynamic in which the democratic regime cannot solve a set of problems, a nondemocratic alternative gains significant supporters, and former democratic regime loyalists begin to behave in a 11  constitutionally disloyal or semiloyal manner.”  The point is that if different phenomena are at work in democratic reversion and consolidation events, and reversions are occurring, then it is important to understand what is occurring in these anomalous cases, rather than simply casting them into residual categories.  Linz and Stepan (1997, 32) argue our understanding will advance when the “democratic transition and democratic breakdown literatures [are] integrated into the overall literature on modern democratic theory.  From the perspective of such an integrated theory, the ‘breakdown of consolidated democracy’ is not an oxymoron.” The Limits of Consolidation Research  The bulk of research attention on democratization during the later 1990s dealt with issues surrounding the consolidation of democracy.  While the consolidation literature has advanced our knowledge of the democratization process, definitional and methodological problems point to democratic reversions as a more productive research focus for the dissertation.  Research on democratic consolidation finds its roots in Rustow’s (1970, 358-61) “habituation” phase.  The basis behind his argument is recognition that “the factors that keep a democracy stable may not be the ones that brought it into existence. (346)”  This recognition provides a foundation for the distinction between democratic transitions and democratic consolidations, a distinction that has informed a good deal of research.7  7 See, for example, Guillermo O’Donnell, Philipe C. Schmitter and Laurence Whitehead, eds., Transitions From Authoritarian Rule: Comparative Perspectives (Baltimore: The Johns Hopkins University Press, 12  The usefulness of democratic consolidation as a research concept has been questioned by a number of leading scholars.  This attack has been led in part by Guillermo O’Donnell (1997, 54) who recognizes that “it is high time for self-criticism.” While there is widespread recognition that consolidation involves more than just stability or endurance,8 one of the main problems with the concept itself is vagueness.  Schedler (1998, 92) argues, At this point, with people using the concept any way they like, nobody can be sure what it means to others, but all maintain the illusion of speaking to one and other in some comprehensible way.  While “democratic consolidation” may have been a nebulous concept since its very inception, the conceptual fog that veils the term has only become thicker and thicker the more it has spread through the academic as well as political world.  If  1986); Guillermo O'Donnell and Philippe Schmitter, Transitions From Authoritarian Rule: Tentative Conclusions About Uncertain Democracies (Baltimore: Johns Hopkins University Press, 1986); Terry Lynn Karl and Philipe C. Schmitter, “Modes of Transition in Latin America, Southern and Eastern Europe,” International Social Science Journal 143 (May 1991); James Malloy and Mitchell A. Seligson, eds., Authoritarians and Democrats: Regime Transition in Latin America (Pittsburgh: University of Pittsburgh Press, 1987); Larry Diamond, Juan J. Linz, and Seymour Martin Lipset, Democracy in Developing Countries (Boulder: Lynne Rienner, 1988); Enrique Baloyra, Comparing New Democracies: Transition and Consolidation in Mediterranean Europe and the Southern Cone (Boulder: Westview Press, 1987). 8 See, for example, Michael Burton, Richard Gunther, and John Higley, “Introduction: Elite Transformations and Democratic Regimes,” in John Higley and Richard Gunther, eds., Elites and Democratic Consolidation in Latin American and Southern Europe (Cambridge: Cambridge University Press, 1992), p. 7; J. Samuel Valenzuela, “Democratic Consolidation in Post-Transition Settings: Notion, Process, and Facilitating Conditions,” in Scott Mainwaring, Guillermo O'Donnell and J. Samuel Valenzuela, eds., Issues in Democratic Consolidation: The New South American Democracies in Comparative Perspective (Notre Dame: University of Notre Dame Press, 1992), 59; Doh Chull Shin, “On the Third wave of Democratization: A Synthesis and Evaluation of Recent Theory and Research,” World Politics 47 (October 1994), p. 144; Cynthia McClintock, “The Prospects for Democratic Consolidation in a “Least Likely” Case,” Comparative Politics 21:2 (January 1989), p. 133. 13  it is true that ‘(n)o scientific field can advance far if the participants do not share a common understanding of key terms in the field,’ then the study of democratic consolidation, at its current state of conceptual confusion, is condemned to stagnation.  The aspiring subdiscipline of ‘consolidology’ is anchored in an unclear, inconsistent, and unbounded concept, and thus is not anchored at all.9 The difficulties that result from this vagueness can be seen in the lack of consensus over measurements of democratic consolidation.  Examples of attempts to operationalize consolidation include survival through a second democratic election for the national leader (Power and Gasiorowski 1997, 132), survival through three consecutive democratic elections (McClintock 1993, 133), the “one turnover” test (Lawson 1993, 194), the “two turnover” test (Huntington 1990, 266-7),10 an elite settlement or convergence (Burton, Gunther and Higley, 1992, 13-5, 24-5),11 absence of a politically significant antisystem party or social movement (Gunther, Diamandouros, and Puhle 1995, 12-3), and survival through twelve years of democratic experience (Gunther,  9 Schedler quotes, Elinor Ostrom, “An Agenda for the Study of Institutions,” Public Choice 48 (1986), p. 4. 10 According to Huntington, “a democracy may be viewed as consolidated if the party or group that takes power in the initial election at the time of transition loses a subsequent election and turns over power to those election winners, and if those election winners then peacefully turn over power to the winners of a later election.” 11 The authors define elite settlements as “events in which warring factions suddenly and deliberately reorganize their relations by negotiating a compromise on their most basic disagreements” (p. 13).  The authors define elite convergence as “when some of the opposing factions in the disunified elite … discover that by forming a broad electoral coalition they can mobilize a reliable majority of voters, win elections repeatedly, and thereby protect their interest by dominating government executive power.  The elite convergence continues once successive electoral defeats convince major dissident and hostile elites … they must beat the newly formed dominant coalition at its own electoral game” (p. 24). 14  Diamandouros, and Puhle 1995, 12-3).  Approaches that examine longevity or crisis run the risk of providing a tautological explanation: the regime survives, ergo it must be consolidated (Bruton, Gunther and Higley 1992, 7; Gunther, Diamandouros and Puhle 1995, 154).  These indicators have also been criticized as atheoretical and ethnocentric. The various attempts to measure consolidation seem to be drawn from a list of desirable characteristics of Western democracies rather than driven by theory (O’Donnell 1992, 48; O’Donnell 1996, 161-163; Hanson 2001, 138). A general tendency in this portion of the literature has been to avoid the problem of specifying an operationalization and instead to approach consolidation as an ideal type (Gunther, Diamandouros, and Puhle 1996, 152-3; Gasiorowski and Power 1998, 743). Cases are the often classified according to the you’ll-know-it-when-you-see-it test.  Even the authors utilizing this approach recognize that use of ideal types “sometimes make it difficult to locate a precise dividing line between consolidated and unconsolidated democratic regimes” (Gunther, Diamandouros, and Puhle 1996, 152).  These approaches seem especially vulnerable to O’Donnell’s argument that democratic consolidation is a teleological concept (O’Donnell 1992, 54; O’Donnell 1996, 162-4; Schedler 1998, 95). The dissertation overcomes these difficulties by moving past consolidation.  In examining democratic reversion, the dissertation returns to one of the original issues driving the consolidation research – regime endurance.  Despite avoidance in the consolidation literature of an explicit examination of endurance, it remains an interesting and important question: why do some regimes endure while others experience reversions from democracy?  This is the core question driving the research in this dissertation. 15   Given the difficulties with the concept of democratic consolidation and the shortcomings of the literature, the dissertation instead turns to an examination of democratic reversion.  Schedler (1998, 103) argues that research should return to the concept’s original concern with democratic survival and that doing so would allow the problematic aspects of the concept to be replaced by “superior alternative concepts.”  The dissertation’s focus on the concept of reversion is an attempt to avoid the conceptual confusion and the operational specification difficulties surrounding democratic consolidation.  State of the Democratic Reversion Literature  The most explicit, comprehensive attempt to study democratic breakdown was undertaken in the four-book collection, The Breakdown of Democratic Regimes, edited by Juan J. Linz and Alfred Stepan and published in 1978.12  To date, this remains the most significant work on democratic reversion.  The theory section of this collection, The Breakdown of Democratic Regimes: Crisis, Breakdown & Reequilibration, written by Linz, drives the case study work in the later sections.  Linz concentrates on the importance of incumbent democratic leaders maintaining legitimacy in order to prevent  12 See Juan J. Linz, “Crisis, Breakdown, and Reequilibration,” in The Breakdown of Democratic Regimes, ed. Juan J Linz and Alfred Stepan (Baltimore: Johns Hopkins University Press, 1978); Juan J. Linz and Alfred Stepan, eds., The Breakdown of Democratic Regimes: Latin America (Baltimore: Johns Hopkins Press, 1978); Juan J. Linz and Alfred Stepan, eds., The Breakdown of Democratic Regimes: Europe (Baltimore: Johns Hopkins Press, 1978); Arturo Valenzuela, The Breakdown of Democratic Regimes: Chile (Baltimore: Johns Hopkins Press, 1978). 16  semiloyal and disloyal elements in society from undertaking actions to break down democracy.  He stresses the importance of the initial policy agendas of new democratic governments as a means of bolstering its legitimacy.  However, in the end, he points to the unsolvable problems most new regimes face in painting a rather bleak picture of their ability to survive.  In hindsight, while extremely important, his approach seems a bit deterministic and the predictions one might draw based on this approach seem overly pessimistic. As influential as this collection was, and remains, it left unresolved problems that the dissertation seeks to address.  In response to the extensive reliance on structural approaches (for example, economic conditions or institutional configurations) existing at that time in the literature, Linz and Stepan (1978, ix) utilize a process approach and explicitly orient the collection to examine the ways “actions or nonactions of incumbent democratic leaders contributed to the breakdown under analysis.”  They (1978, 19) also heavily rely on the notion of legitimacy, despite a recognition that “we have neither developed systematic indicators nor collected the data over time on the legitimacy of regimes that would test hypothesis of this type.” Obviously, there are three decades of cases to examine since the publication of the Breakdown volumes.  The European cases were all from the interwar period and the Latin American cases, other than Chile, were mainly pre-1960s.  The dissertation thus offers the potential to test their ideas against the last thirty years of cases.  Linz and Stepan (1978, vii) were clear about the fact that their work “precludes the highly abstract generalizing of ahistorical social scientific models of the type susceptible to computer 17  simulations and applicable to all past and any future cases” and hope for scholars “interested in developing more formal models (that) may build on our work and incorporate into their models the complex realities discussed here.”  It is the aim of the dissertation to push the literature in the direction of this call. More recently, Cohen (1994) undertook the most explicit attempt to apply a rational choice approach to the study of democratic breakdown.  Making use of a single shot Prisoner’s Dilemma game, Cohen argues actors in Brazil and Chile found themselves in a situation that prevented cooperation and led to the collapse of democracy. He (1994, 124) concludes, these successful tests of the PD game “make a good case for the use of intentional explanations in the study of large-scale political transformations such as the collapse of democracy.”  The work in this dissertation seeks to address some of Cohen’s shortcomings (discussed shortly) and further extend Cohen’s initial efforts to apply rational choice modeling to the issue of reversion. The dissertation also builds on recent efforts to understand autogolpes in Latin American.  An autogolpe occurs when the executive temporarily suspends the constitution, dissolves the legislature, and rules by decree until calling new legislative elections and a referendum on ratification of a system with more expansive executive powers (Cameron 1998b, 220).  The textbook case occurred in April of 1992 when Alberto Fujimori, the President of Peru, suspended the constitution, closed the congress and fired the country’s top judges.  While research has moved beyond the Peruvian case to examine events, for example in Guatemala and Russia, most of the autogolpe literature is limited to Latin American case studies (see, for example, Cameron 1998a).  The 18  dissertation incorporates the lessons from this form of reversion in an effort to introduce a framework capable of accounting for all forms of democratic reversion. Given the increasing concerns regarding democratic stagnation in a number of regions around the world, more attention has been paid to democratic reversion issues recently.  As one most influential works recently published in this area of study, Przeworski, et al. (2000, 36) examined 141 countries for at least some of the time between 1950 and 1990.  They concluded (2000, 269) that “even if democracies do occasionally spring up in poor countries, they are extremely fragile when facing poverty, whereas in wealthy countries they are impregnable.” And in terms of the key factors that trigger democratic reversions, they concluded (2000, 273) that, “ the probability that, once established, a democracy will survive increases steeply and almost monotonically as per capita incomes get larger.  Indeed, democracy is almost certain to survive in countries with per capita incomes above $4000.”  In comparison to this work, there are a number of value-added elements of the dissertation, including: expanding the analysis to include the post-Soviet cases, as well as the cases involved in the democratic stagnation of the late 1990s and the 2000s; drawing on both structural and process explanations in an effort to provide a more complete picture of democratic reversion; testing the theories involved against a universe of democratic reversion case studies; explicitly examining the role democratic uncertainty may play in reversion decision-making; and, challenging the conclusions in this work by attempting to overcome several methodological issues via the dissertation’s use of a multi-methodological approach. 19  In addition, there have been recent works more directly aimed at the issue of democratic reversion or breakdown.  For example, Mainwaring and Pérez-Liñán (2005) examine the issues of democratic transitions, breakdowns and erosions in Latin American from 1946 to 2003 in an effort mainly directed at explaining the evolution of democratization in the region since 1978.  They found that “structural and regime performance variables were weak predictors of both transitions and breakdowns” (2005, 34).  They do however point out that political variables, particularly party polarization, are important.  However, they note that one difficulty they face in their attempts to examine the post-1978 situation in Latin America was that, “there was only one breakdown after 1978 (Peru’s coup in 1992).  No model estimation was feasible for the post-Third Wave part of the sample because of the lack of variance in the dependent variable.  Hence it was not possible to test statistically whether the impact of some independent variables had changed over time” (2005, 38).  In comparison to this work, there are a number of value-added elements of the dissertation, including: expanding the scope of the analysis to include the whole world during the third wave period; the inclusion of numerous cases of democratic reversion during this period allowing for statistical tests of the independent variables in question; a more explicit effort to employ both structural and process approaches; and a demonstration that their methodological emphasis on examining democratic reversions in comparison only to ongoing democratic regime years does call into question some of the conclusions found in Przeworski et al. (2000).  20  The Competition between Structure and Choice  One difficulty that the democratization literature has encountered for quite some time is that scholars employing the two fundamental methodological approaches in this area have ended up talking past one and other rather than attempting to explicitly integrate these approaches in an effort to reach a deeper, more complete picture of democracy and change.  At the root, structural approaches argue that regime change is a function of resource and institutional constraints.  Process approaches see regime change as a function of actor choice driven by perceptions of preferences and relative strengths (Kitschelt 1992, 1028).  While some advocates of each of these approaches could be viewed as true believers, it is important to recognize that these are not really warring camps.  Some scholars, for example Adam Przeworksi, have moved between these two approaches while most others would freely admit there is a point to be made by research employing the alternative approach.  Rather, this is more of an issue that existing research does not fully specify their models.  Scholars employing structural approaches most often simply leave choice variables out of their models.  And while these works may even make reference to actor choice issues, they simply are not explicitly contained in the models they consider.    Likewise, scholars employing process or actor choice approaches most often simply leave structural constraints out of their models.  And while these works may even make reference to such constraints, they simply are not explicitly contained in the models they consider and this problem results in scholars talking past each other.  The dissertation attempts to at least partially address this difficulty by drawing on both structural and process explanations. 21  The roots of the structural literature can be found in writings from the 1960s and early 1970s concerning the search for preconditions.  The four major areas of preconditions research are modernization, political culture, historical, and external factors, all of which relied on the argument that certain preconditions were necessary for the emergence of democracy (Karl 1990, 2-4).  Structural explanations also predominated the subsequent study of democratic breakdown (Cohen 1994, 128).  Following the resurgence of democratic regimes during the third wave, the structuralists have attempted to explain the breakdown of authoritarian regimes and their transitions to democracy.13 For example, research on consolidation turned its attention to structure, examining political institutions14 and socioeconomic factors.15  The common foundation for  13 See, for example, Adam Przeworski, “Some Problems in the Study of the Transition to Democracy,” in Guillermo O’Donnell, Philipe C. Schmitter and Laurence Whitehead, eds., Transitions From Authoritarian Rule: Comparative Perspectives (Baltimore: The Johns Hopkins University Press, 1986); Edward C. Epstein, "Legitimacy, Institutionalization, and Opposition in Exclusionary Bureaucratic-Authoritarian Regimes," Comparative Politics 17 (October 1984); John Walton and Charles Ragin, “Austerity and Dissent: Social Bases of Popular Struggle in Latin America,” in William L. Canak, ed., Lost Promises: Debt, Austerity, and Development in Latin America (Boulder: Westview Press, 1989); Jonathan Hartlyn and Samuel A. Morley, “Bureaucratic-Authoritarian Regimes in Comparative Perspective,” in Jonathan Hartlyn and Samuel A. Morley, eds., Latin American Political Economy (Boulder: Westview Press, 1986); Mitchell A. Seligson, “Democratization in Latin America: The Current Cycle,” in James Malloy and Mitchell A. Seligson, eds., Authoritarians and Democrats: Regime Transition in Latin America (Pittsburgh: University of Pittsburgh Press, 1987). 14 See, for example, Juan J. Linz, “The Perils of Presidentialism,” Journal of Democracy 1:1 (1990a); Juan J. Linz, “The Virtues of Parliamentarism,” Journal of Democracy 1:4 (1990b); Juan J. Linz and Auturo Valenzuela, eds., The Failure of Presidential Democracy (Baltimore: Johns Hopkins University Press, 1994); Matthew Soberg Shugart and John M. Carey, Presidents and Assemblies (New York: Cambridge University Press, 1992); Scott Mainwaring, “Presidentialism, Multipartism and Democracy: The Difficult Combination,” Comparative Political Studies 26 (1993); Alfred Stepan and Cindy Skach, “Constitutional Frameworks and Democratic Consolidation: Parliamentarianism versus Presidentialism,” World Politics 46 (October 1993); Scott Mainwaring, “Party Systems in the Third Wave,” Journal of Democracy 9:3 (1998). 15 See, for example, Yi Feng, “Democracy, Political Stability and Economic Growth,” British Journal of Political Science 27 (1997), p. 392-4; David A. Leblang, “Political Democracy and Economic Growth: Pooled Cross-Sectional and Time Series Evidence,” British Journal of Political Science 27 (1997), p. 453. 22  structuralist is that regime change can be explained in terms of resource and institutional constraints, effectively holding actors’ preferences constant (Kitschelt 1992, 1028-1031). In contrast, process-driven explanations argue that academic focus should be placed on actor’s strength and preferences.16  For these analysts, democratization is seen as a process consisting of the breakdown of the old regime, a period of rule-making, and the installation and consolidation of the new democracy (Shin 1994, 143).  A key element of this approach is the uncertainty of the democratic game (Przeworski 1991, 12-14, 40- 50; Di Palma 1990, 40-43).  Democratic uncertainty is important because, as opposed to the (relatively) certain winners and losers of non-democratic regimes, all actors have an opportunity to win (and lose) under a democratic regime.  So, as the costs of maintaining  Also see for example, Ross E. Burkhart and Michael S. Lewis-Beck, “Comparative Democracy: The Economic Development Thesis,” American Political Science Review 88 (December 1994); Stephan Haggard and Robert R. Kaufman, The Political Economy of Democratic Transitions (Princeton: Princeton University Press, 1995); Uk Heo and Alexander C. Tan, “Democracy and Economic Growth: A Causal Analysis,” paper presented to the Annual Meeting of the Southern Political Science Association, (Norfolk, Virginia, November 5-8, 1997); Adam Przeworski and Fernando Limongi, “Modernization: Theories and Facts,” World Politics 49 (January 1997); Adam Przeworski, Michael Alvarez, Jose Antonio Cheibub and Fernando Limongi, “What Makes Democracies Endure?” in Larry Diamond, Marc F. Plattner, Yun-Han Chu, and Hung-Mao Tien, eds., Consolidating The Third wave Democracies (Baltimore: The Johns Hopkins University Press, 1997); Karen L. Remmer, “Democracy and Economic Crisis: The Latin American Experience,” World Politics 42 (April 1990), p. 327; Karen L. Remmer, “The Political Economy of Elections in Latin America, 1980-1991,” American Political Science Review 87 (June 1993), p. 405. 16 See, for example, John Higley and Richard Gunther, eds., Elites and Democratic Consolidation in Latin American and Southern Europe (Cambridge: Cambridge University Press, 1992); Scott Mainwaring, Guillermo O'Donnell and J. Samuel Valenzuela, eds., Issues in Democratic Consolidation: The New South American Democracies in Comparative Perspective (Notre Dame: University of Notre Dame Press, 1992); David Collier and Deborah Norden, “Strategic Choice Rational choice account of Political Change in Latin America,” Comparative Politics 24 (January 1992); Giuseppe Di Palma, To Craft Democracies (Berkeley: Univ. of California Press, 1990); Terry Lynn Karl, “Dilemmas of Democratization in Latin America,” Comparative Politics 23 (October 1990); Terry Lynn Karl and Philipe C. Schmitter, “Modes of Transition in Latin America, Southern and Eastern Europe,” International Social Science Journal 143 (May 1991); Guillermo O'Donnell and Philippe Schmitter, Transitions From Authoritarian Rule: Tentative Conclusions About Uncertain Democracies (Baltimore: Johns Hopkins University Press, 1986). 23  the certainty of the old regime increase, process-driven explanations argue, the attractiveness of the uncertainty of democracy increase (Przeworski 1991, 12-14, 40-50; Di Palma 1990, 40-43). The debate between structural and process-driven explanations has, to an extent, cast these approaches into two different camps.  This is a mistake.  Proponents of process-driven explanations make a good point when they argue that pure structure leaves no room for political choice by actors in the process.  Likewise, structural proponents make a good point when they argue that process explanations leave little room to accommodate outside, environmental factors that go beyond and, potentially constrain, actor choice.  These approaches go astray when they argue for the primacy or exclusivity of their approach.  Structure does a good job at explaining why a case reaches the point where some sort of regime change comes into play.  Process does a good job a explaining the timing and outcomes of cases once they are in play.  These camps should not be viewed in opposition to one and other but rather it should be seen that each approach can inform the other in arriving at a more complete picture of regime change.  The dissertation attempts to at least partially address this problem by drawing on both structure and process explanations. The Plan  The dissertation describes a framework for understanding democratic reversion based on the notion that actors make decisions regarding regime support based on pocketbook issues.  As such, a multi-method approach is used to examine the relevance of four of the most widely employed groups of structural considerations as they relate 24  specifically to the idea of democratic reversion, a rational choice account based around the importance of democratic uncertainty is laid out and finally a single, in-depth case study of the Philippines is explored as a means of testing the overall framework against a case that where democracy persists despite widespread challenges. In Chapter 2, the theoretical orientation as well as the data and methods employed in the dissertation are laid out.  A theoretical framework for analyzing democratic reversion based on existing approaches to studying democratization while attempting to incorporate the notion of democratic uncertainty is described.  The baseline argument uniting these various theoretical elements is that actors are concerned with the ability of democratic regimes to deliver an acceptable level of goods.  When regimes fail to deliver goods at an acceptable level, actors withdraw their support for the regime and the potential for a democratic reversion increase.  The chapter proceeds to describe the theoretical importance of economic development and performance, political institutional configuration, vulnerability to international pressure, the assertion of civilian control over the military, and the incorporation of democratic uncertainty into actor decision-making around regime support. Following this, the choices around data and methods are discussed.  The foundation of the methodological approach pursued in the dissertation is the importance of methodological pluralism.  The four chapters addressing the basic structural constraints important to democratic reversion each employ three methodological approaches.  First, a cross-national, time-series analysis is undertaken which examines the dependent variable is explored from four separate perspectives ranging from the more general regime type differences to the more specific issue of reversions from democracy. 25  This analysis is followed by an aggregate analysis of thirty cases of democratic reversion. Finally, a mini-analysis of each of the thirty cases as they relate to the chapter’s subject matter is undertaken.  Next a rational choice argument for the incorporation of democratic uncertainty into reversions decision-making is advanced.  Finally, the framework is tested against a single chapter-length case study of a non-reversion case, the Philippines. In Chapter 3, the role of economic issues as a structural constraint on actors considering democratic reversion is explored.  The research in this chapter examines the level of economic development, changes in economic growth, the level of inflation and changes in inflation.  The analysis confirms a strong influence which both the level of economic development and changes in economic growth exert on democratic reversion. The aggregate case analysis and the individual case-level analyses both draw into question these connections and caution against overstating the importance of the role economic issues play in democratic reversion. In Chapter 4, the role of the structure of the political system as a structural constraint on actors considering democratic reversion is explored.  The research in this chapter examines the influence of presidential and parliamentary systems, plurality versus proportional representation electoral rules, legislative fractionalization and the tenure of the regime.  The findings indicate that while the structure of the political system is clearly relevant to democratic reversion, it is difficult to support the position that it is a determinant factor.  The analysis demonstrates that the manner in which you examine the dependent variable drives a lot of the results.  When the dependent variable is based on 26  difference in regime type, the political institutions approaches are significant and explain a good deal of the variance.  When the analysis shifts to an examination of reversion, the models remain significant but explain very little variance.  While the aggregate case analysis seems to confirm this problem, the case level analysis points to the important role the structure of the political system plays in mediating the evaluation of democratic uncertainty. In Chapter 5, the role of international factors as a structural constraint on actors considering democratic reversion is explored.  The research in this chapter examines a country’s vulnerability to reversion and pressures to maintain democracy based on its dependence on trade and aid as well as its dependence on fuel imports and exports.  The analysis points to a number of significant relationships however none of these provide much purchase for explaining variance of the dependent variable.  The aggregate case analysis points to no substantial role for trade or aid vulnerability.  The individual case analyses however point to the potentially important role international actors may play in democratic reversion.  In some cases this role supports democracy while in others it supports reversion.  The role of Russia over the post-Soviet states is an example of the later.  The role of colonial powers over their former colonies is mixed.  The role of the United States seems to depend on time and the strategic importance of the country. These findings suggest that when international influences are linked with an understanding of the perceptions of democratic uncertainty in a country, we may arrive at a better understanding of the potential for democratic reversion. 27  In Chapter 6, the role of the military as a structural constraint on actors considering democratic reversion is explored.  The implications of attempting to assert civilian control over the military are examined by looking at changes in force levels, changes in defense spending, and the existence of a civilian defense minister.  The findings in this chapter draw into question the conventional wisdom with regards to asserting civilian control over the military.  While the analysis exploring difference between regime types confirms the idea that civilian governments should assert control over the military, the aggregate case analysis as well as the thirty individual case studies point to a conclusion that attempts to assert such control may end up triggering, not preventing, democratic reversion. In Chapter 7, the idea of democratic uncertainty is introduced.  The flaws in the recent interpretation of this concept are discussed and a new interpretation which addresses the existing concerns is introduced.  This interpretation relies on a more limited conception and set of outcomes.  An argument is introduced that actors make decisions about regime support by viewing the system’s structural constraints in light of their evaluations of democratic uncertainty.  This allows them to consider not just their present circumstance but also consider the potential for gaining future power by utilizing the uncertainty inherent to the democratic rules of the game. In Chapter 8, the case of the Philippines is examined. While chapters 3-6 examine case studies of democratic reversion, this chapter explores a case where a reversion did not occur despite demonstrating vulnerabilities to all the structural constraints outlined in the dissertation.  Following the transition to democracy in 1986, the Philippines faced a 28  strong and independent military, a number of economic crises, international influences not always favoring democracy, and difficulties regarding presidential power, shifting electoral rules and legislative fractionalization.  The chapter’s case analysis points to a variety of efforts to bolster democratic uncertainty, including constitutional reforms, efforts to accommodate the military, very public anti-corruption efforts, rhetorical employment of the negative elements of the authoritarian legacy, turnover in party control of the executive and legislature, as well as the role played by the judiciary and the Church.  While it is recognized that each of these measures are important in their own right, they also contributed to maintaining a level of democratic uncertainty in the Philippines that allowed the democratic regimes to persist in the face of numerous crises that held the potential for triggering democratic reversion. Chapter 9 concludes the dissertation by reviewing the results of the individual chapters, drawing some general conclusions and suggesting some future avenues for research suggested by the dissertation.  29  Chapter 2: Theoretical Orientation and Methodology 30  Introduction  This chapter introduces a framework for explaining the reversion from democracy which relies on the notion that actors base their decision making around support for democratic regimes on traditional pocketbook issues.  It draws on a number of the most widely employed theories in the study of democratization in order to examine them in the specific context of democratic reversions.  The concept of democratic uncertainty is introduced as an additional theoretical element for consideration.  The baseline argument uniting these various theoretical elements is that actors are concerned with the ability of democratic regimes to deliver an acceptable level of goods.  When regimes fail to deliver goods at an acceptable level, actors withdraw their support for the regime and the potential for a democratic reversion increase.  These actors are considered as mainly interested in economic goods.  So the chapter begins with an explanation of the importance of both the level of economic development as well as economic performance. Countries at higher levels of economic development are less vulnerable to reversion.  As well, countries with higher levels of economic performance (in terms of economic growth and inflation rates) are less vulnerable to reversion.  A country’s political institutional configuration is also relevant to the issue of democratic reversion.  Here it is argued that when governments that fall into legislative gridlock the country is more vulnerable to democratic reversion as the government’s ability to address pocketbooks issues is perceived as compromised.  In this regard, executive structures, the political party systems and electoral rules are taken into consideration and presidential systems, systems 31  with higher party fractionalization, and systems employing proportional representation voting are seen as more vulnerable to democratic reversion. Further, a country’s vulnerability to international pressure is considered.  As international pressure can directly affect pocketbook issues, countries that are less vulnerable to such pressures are more likely to experience democratic reversions.  As well, as the military is the most common veto player in a regime change situation, the pocketbook issues of the military are considered.  While democratic regimes are often counseled to assert civilian control over the military, this is often accomplished in a manner that affects the military’s pockets (its budget or personnel levels).  As such, the countries that confront the military in an effort to assert increasing civilian control are more vulnerable to democratic reversion. Finally, the idea of democratic uncertainty is introduced as a means to further understanding what levels of goods are acceptable as well as what time horizon actors consider when evaluating their support for democratic regimes.  When actors perceive higher levels of democratic uncertainty, they are more willing to tolerate a lower level of goods and take a longer time horizon into account when evaluating their regime support. So countries with lower democratic uncertainty are more likely to experience democratic reversion.  The chapter concludes with an explanation of the data and methods employed in the dissertation. Economic Influences  The question of if and how economic issues influence democracy (and vice versa) has been debated in comparative politics for at least the last five decades.  There are two 32  main considerations addressed in the democratization literature.  The first explores the relationship between the level of economic development and regime type.  The second examines the implications of poor economic performance on regime type.  Some of the earliest research in this area of study was concerned with a search for preconditions to democracy.17  These early efforts to study the connections between economic issues and regime were greatly expanded in the 1990s.18  More recently, Przeworski et al. (2000) “hit the field of political development like a bolt of lightening … offering evidence that the exogenous theory (development makes democracies, once established, less likely to fall to dictatorships) holds and the endogenous one (development increases the likelihood that poor countries will undergo a transition to democracy) fails” (Boix and Stokes 2003, 517).  While a number of scholars have since risen to the defense of endogenous democratization, the notion that “the dynamics of achieving democracy and sustaining it may not be the same” has gained widespread acceptance in the literature (Biox and Stokes 2003, 545).  As such, this dissertation tests the existing economic development and economic performance theories by examining both large-n statistical tests as well as a smaller group of cases consisting of the universe countries experiencing a democratic reversion during the third wave.  17 See, for example, Seymour Martin 1959; Huntington 1968; Almond and Verba 1963; Moore 1966; Dahl 1971; Cutright 1963; Binder 1971. 18 See, for example, Feng 1997, 392-4; Leblang 1997, 453.  Also see for example, Burkhart and Lewis- Beck 1994; Haggard and Kaufman 1995; Heo and Tan 1997; Przeworski and Limongi 1997; Przeworski, Alvarez, Cheibub and Limongi 1997; Remmer 1990, 327; Remmer 1993, 405. 33  There has long been a presumed connection between economic development and regime type.  As the bulk of countries in the world, especially since WWII, have tended to group in the higher developed/democracy and lower developed/non-democracy quadrants, many assumed that there was a causal connection between the two.  The debate in the late 1950s and 1960s revolved around the notion that non-democratic forms of government were necessary in order to generate the economic development necessary to sustain democracy over the longer term.  From the modernization perspective, a country must undergo economic and social transformations before a sustainable form of democracy can emerge.19  Almost four decades later, Przeworski and Limongi (1997) and Przeworski et al. (2000) stood this notion on its head by arguing that we should be examining two different questions and recognizing that there were two different factors at play.  They termed the modernization approach as an endogenous theory.  This theory is employed to explain why countries made a transition to democracy.  Their research failed to support this approach.  On the other hand, when they turned to the question of why democracies, once established, endure, they found that the exogenous theory, that higher levels of economic development make democracies less likely to fail, was supported by the data. The dissertation thus examines the relationship between the level of economic development and democratic reversion by hypothesizing that countries at higher levels of economic development are less likely to experience democratic reversion than countries at lower levels of economic development.  The assumption behind this approach is that in  19 See, for example, Przeworski et al. 2000, 2-3; Huntington and Nelson 1976, 23; Lipset 1963, 31. 34  wealthy countries the income actors receive is sufficiently high to deter reversions as the marginal increase in income is not sufficient to justify the risks and costs associated with a reversion move.  On the other hand, in poor countries, given the low levels of income, the relative difference in incomes between the different regime types is sufficiently large to take on these risks and costs (Przeworski and Limongi 1997, 166). One of the interesting findings Przeworski (1996, 49) has repeatedly pointed out is that democracy is “certain to survive” above $6,000 per capita income.  Given that this finding has held over the subsequent decade, one wonders about the countries under that threshold.  What, if anything, differentiates reversion from non-reversion cases?  This leads to a secondary question of the relationship between economic performance and regime endurance.  A variety of scholars have argued there is a connection between poor economic performance and regime change.20  This is the basic “pocketbook” voter hypothesis applied to regime support.  Here the idea is that actors will support a democratic regime as long as it is performing well.  In this case, performance is judged by the economic health of the country on an annual basis.  The dissertation thus hypothesizes that countries experiencing positive economic performance are less likely to experience democratic reversion than countries with negative economic performance.  As will be discussed in greater detail in the subsequent chapter addressing these matters, we will examine economic growth and inflation as proxies for economic performance.  20 See, for example, Diamond 1999, 77-93; Diamond and Linz 1989, 44-46; Gasiorowski 1995; Geddes 1999; Haggard and Kaufman 1995; Lipset et al. 1993; Przeworski et al. 2000. 35  Political Institutional Influences  The democratization literature has also experienced a substantial debate on whether institutional configurations play a role in regime stability and, if so, which are most conducive to democracy.  These debates include important exchanges on executive structures, electoral rules and party structures.  The issues surrounding the structure of the executive have largely revolved around a discussion of the relative merits of presidential and parliamentary systems while the issues surrounding party structure address the distribution of power across the party system.  As one of the central themes of the dissertation involves an examination of pocketbook issues as they relate to regime support, the structure of political institutions is important as it affects the ability of the current government to at least appear to be addressing core pocketbook concerns of actors.  While it is important not to cast aside the nuance of the various institutional issues, one way actors may evaluate the relative merits of regimes is to look at the ability of the government to pass legislation addressing their core concerns.  As such, we argue that when governments appear to be gridlocked, actors concerned with basic pocketbook issues may withdraw their support for the regime. The first political institution issue addresses the structure of executive power in the context of the debate between presidential and parliamentary systems.21  A large portion of the literature over the last two decades points to the idea that presidential systems are more prone to democratic reversion.  Valenzuela (2004, 15) points to the  21 See, for example, Linz 1990a; Linz 1990b; Linz and Valenzuela 1994; Shugart and Carey 1992; Mainwaring 1993; Stepan and Skach 1993; Mainwaring 1998; Power and Gasiorowski 1997; Hadenius 1994; Przeworski et al. 1996) 36  problem of competing legitimacies in presidential systems as the executive and legislature can lay claim to distinct electoral mandates.  Linz (1990, 1994) argues that zero sum elections and dual legitimacy undermined the ability of a president and the legislature to compromise.  When the executive and legislature are unable to compromise and lay claim to distinct electoral mandates, the potential for legislative gridlock increases.  Parliamentary systems, on the other hand, fuse executive and legislative powers reducing the issues of competing legitimacies and increasing the likelihood of cooperation between the executive and legislature and thus reducing the potential for legislative gridlock. A second consideration lies in the differential safety valves of the two systems.  In a presidential system, it is usually quite difficult for the legislature to remove an unpopular or uncooperative president from power.  Likewise, in most presidential systems, the president’s ability to dissolve the legislature in the face of a crisis is more restricted or nonexistent.  In a parliamentary system, on the other hand, a prime minister may dissolve parliament while the parliament may employ a vote of no confidence. Given these safety valves, Valenzuela (2004, 16) argues that a crisis of government in a parliamentary system rarely becomes a crisis of regime.  As such, the dissertation hypothesizes that presidential systems are more likely to experience democratic reversions than are parliamentary systems. The second political institution issue explores political party structure in light of legislative fractionalization.22  While some early scholars (Dahl, 1971) took the position  22 See, for example, Duverger 1954; Lipset 1960; Dahl 1971; Lijphart 1977; Midlarsky 1984. 37  that a two party system was essential to democratic stability, others have advocated more moderate levels of fragmentation (Dishkin et al, 2005).  Few however take the position that high fractionalization is a positive force for democratic stability.  At very low levels of fractionalization, the potential for legislative gridlock should be lower as a single party is more likely to be able to push through its legislative agenda.  As the legislature becomes increasingly fragmented, however, the number of players necessary to pass legislation is more likely to rise, increasing the potential for gridlock.  The dissertation thus hypothesizes that countries with higher levels of legislative fractionalization are more likely to experience democratic reversion than countries with lower levels of legislative fractionalization.  Further to this point, the issue of electoral rules is examined in the context of the debate between plurality and proportional representation systems.23 While some argue that proportional systems may increase the fragility of the party system, others argue they increase representation.  Likewise, plurality systems are viewed as creating zero sum contests and reducing representation (Diskin et al, 2005).  Given the emphasis on legislative gridlock, the dissertation hypothesizes that proportional representation systems are more likely to experience democratic reversion than plurality systems. International Influences  International influences are critical to the democratization process.  For quite some time, the literature was dominated by the consideration of domestic issues.  In  23 See, for example, Blais 1991; Blais and Dion 1990; Hadenius 1994; Lardeyret 1991; Lijphart 1994. 38  1995, Karen Remmer (106-7) reminded us that “attempting to understand national politics in isolation from international forces is likely to prove particularly futile, if not counterproductive.”  In response to this shortcoming, attention to international factors has substantially increased.  If we believe that pocketbook issues are behind decisions surrounding democratic reversion, then an examination of international factors is highly relevant.  Countries do not operate in a vacuum.  Decisions regarding regime change will engender a response from a variety of sources outside of the country.  In recent years, there has been a substantial push for actors within the international community to encourage democratization and discourage democratic reversion.  The international responses to democratic reversion have the potential to substantially affect pocketbook issues.  Therefore, actors considering their support for a reversion move have to account for possible responses including economic sanctions as well as political and military pressure to restore the democratic regime.  The dissertation hypothesizes that countries with higher levels of vulnerability to international pressure are less likely to experience democratic reversion than countries with lower levels of vulnerability. All countries are vulnerable, to a greater or lesser extent, to pressure nondomestic actors may attempt to exert.24  Given our discussion of pocketbook issues, it is important to consider a country’s vulnerability to economic pressures or threats.  The most obvious issues are the extent to which countries engage in international trade and their level of reliance on international aid.  It is important to remember that while the international  24 For the sake of convenience, I am terming nondomestic actors as international actors.  These may (at least) include other states, international organizations and nondomestic NGOs. 39  community, in general, has become more supportive of democracy, such support is far from universal.  It is easy to imagine circumstances where the response by some international actors would be supportive of regime change.  As such, to the extent possible, it is relevant to consider not just vulnerability to pressure but also the form of expected pressure (pro-democratic or pro-reversion). Military Influences  Civilian control over the military is one of the most widely discussed issues in the democratization literature.  The problem inherent to the relationship between the civilian government and the military, the civil-military problematique, is that the institution created to protect the state is also the institution with sufficient power to threaten the state (Feaver 1999, 214).  There is little doubt that the military is a crucial actor throughout the democratization process.  One of the most basic propositions in the democratization literature regarding the military addresses its veto player role (see, for example, Needler 1975; das Gupta 1978).  In terms of the specific issue of the reversion from democracy, most often, the military (or at least some element of the military) plays the critical role. When they are not the critical actor, they usually provide a substantial supporting role. And while there are instances where the military does not instigate the reversion, it is highly unlikely such a reversion may occur without at least passive acquiescence on the part of the military.  At minimum, the military must agree not to actively oppose a reversion move if it is to succeed. Given the important role of the military, the issue becomes how to keep the military in the barracks.  The most common suggestion involves the civilian government 40  asserting control over the military (Foster 2005, 91; see also, Needler 1975; das Gupta 1978; Pion-Berlin and Trinkunas 2005; Rice 1992; Kohn, 1997).  In this view, a lack of civilian control over the military is viewed as contributing factor to democratic reversion. The leaders of newly democratic regimes are thus counseled to assert control over the military.  The argument concludes that “democracy is not possible without civilian control of the military” (Foster 2005, 96).  The difficulty with this position is that it seems more suited as the answer to a different question.  If the question is how do we “deepen” democracy or how do we know when a democracy is “consolidated,” then a discussion of how to assert civilian control over the military should be front and centre. However, we are concerned with the survival of democracy not the enhancement or deepening of democracy.  And it is important to bear in mind that these are distinct issues.  While successfully confronting the military may demonstrate the depth of democracy, this dissertation takes the position that the act of confronting the military puts the regime at risk.  If this institution is in fact the most likely to threaten the existence of the regime and if the overriding concern is avoiding a reversion from democracy, then it seems far more prudent to avoid such confrontation.  As such, the dissertation hypothesizes that regimes attempting to assert higher levels of civilian control over the military are more likely to experience a reversion from democracy than regimes that assert lower levels of such control. The two ways in which it is possible to examine civilian control are to look at resources and command structures.  The most basic measures of the resources of the military are budget and force levels.  Advocates of asserting greater civilian control argue that democratic regimes should reduce military spending (see, for example, Haggard and 41  Kaufman 1995, 114-5; Gasiorowski and Power 1998, 746) and reduce force levels (see, for example, Huntington 1995, 12; Kohn 1997, 145; Gasiorowski and Power1998, 746; Diamond 1999, xxxi).  The competing point of view is that confronting the military runs the risk of threatening its corporate interests and triggering a reaction that could result in a democratic reversion.  This later position is in keeping with the pocketbook issue focus previously discussed in this chapter.  While the military may be concerned with the state of the economy or the functionality of the legislative process, they are likely to focus on how these issues directly impact their bottom line.  This point of view has been articulated by the toys-for-boys advocates who take the position that budgets should be maintained or even increased in order to keep the military happy (Sorenson 2007, 102; Huntington 1991, 252).  As such, the dissertation hypothesizes that countries that reduce military spending/force levels are more likely to experience reversions from democracy than countries that maintain or increase military spending/force levels. The notion of the importance of a civilian defense minister seems to be taking on gold standard status as an indication of civilian control over the military (see, for example, Fitch 1998, 37-8; Kohn 1997, 150).  In the case of countries that have experienced a recent transition to democracy, the idea is that the military “carries the burden of loyalty to the previously autocratic government” (Kohn 1997, 150).  As such a civilian defense minister serves to assert control over these loyalty issues and clarifies the chain-of-command (Huntington 1991, 252).  On the other hand, the military is quite often intensely concerned with protecting its corporate interests and it is not difficult to see how they could view a civilian, appointed by elected officials, tasked with running the military as a threat to these interests.  As such, the dissertation hypothesizes that countries 42  with civilian defense ministers are more likely to experience a reversion from democracy than countries with military defense ministers. Democratic Uncertainty  Research on democratization has long accepted the importance of the notion of uncertainty.  O’Donnell and Schmitter (1986, 66) brought attention to the idea that authoritarian breakdowns trigger a period of uncertainty that may result in a variety of regimes, both democratic and nondemocratic.  Przeworski (1986 and 1991) introduced the argument that there is a form of uncertainty that is particular to democracy.  Actors in a democracy subject their interest to competition and hence “institutionalize uncertainty” (Przeworski 1991, 14).  The point of this form of uncertainty is that while actors may know what is possible or likely to result, they do not know, with certainty, the outcome of elections.  This notion has been employed as a theoretical tool for answering the question of why democratic losers accept defeat rather than withdrawing their support for the regime.  Rather than simply considering the payoffs they will receive as losers, actors also consider the potential for improving their situation due to the uncertainty of the outcome of future elections.  The dissertation hypothesizes that when actors perceive a high level of democratic uncertainty there is less risk of a reversion from democracy than when actors perceive a low level of democratic uncertainty. One of the difficulties with incorporating democratic uncertainty into an analysis of the reversion from democracy is there is a great deal of conceptual confusion surrounding the idea of uncertainty in the democratization literature.  As a result, the theoretical foundations of the concept have recently come under fire (see, for example, 43  Alexander 2002 and Schedler 2001).  To address such concerns, the dissertation narrows the application of democratic uncertainty to an evaluation of the extent to which a democratic regime creates the opportunity to legally recruit followers and to periodically convert that support into representation.  As such, the idea of democratic uncertainty serves as a mediating influence between the structural stimuli that may trigger a democratic reversion and the decision-making process and choices actors make in considering whether to support a reversion.  When making such decisions, actors do not limit themselves to an evaluation of the goods they currently (or recently) receive, rather they also consider the potential for future change, and hence improvement, given the level of perceived democratic uncertainty in a country.  So when actors perceive a high level of democratic uncertainty they are more tolerant of a set of benefits that is lower than their preferred outcome because they perceive a higher potential for future change via the rules of democracy.  On the other hand, when the perceived level of democratic uncertainty is low, they may require a higher level of goods. The inclusion of democratic uncertainty into the analysis of reversion is an important tool for evaluating the pocketbook issues previously discussed.  Most research in this area of study that employs statistical analyses across time and/or space, rely on one year lags.  So, actors make decisions regarding regime support based on the goods they received in the previous year.  As discussed in the next section, this is a troubling assumption given the random distribution of reversion dates across any given year.  In addition, the effort to include the concept of democratic uncertainty into the analysis of reversion draws attention to the idea that these lagged approaches leave no real room for actors to consider the longer-term benefits of democracy.  From this perspective, regimes 44  are simply mechanisms that deliver goods.  If they perform well, regimes usually receive the level of support necessary to survive.  When they underperform, support is withdrawn.  Actors are thus regime-neutral and are not allowed to consider how different regime types may change the level of goods they expect to receive in the future. Including the idea of democratic uncertainty into the analysis of democratic reversion remedies this shortcoming. Data and Methods  The foundation of the methodological approach pursued in the dissertation is the importance of methodological pluralism.  Rather than relying on a single methodological approach, this research makes use of a number of the most commonly employed approaches to the study of democratization.  To begin, each of the “structural constraint” chapters employs three basic methodological approaches.  First, a cross-national, time series analysis examines the relationship between the independent variable in question and democratic reversion.  The purpose of this approach is to replicate existing quantitative work on regime change and then to narrow down the analysis to the particular issue of democratic reversion.  This is accomplished by analyzing the dependent variable from four different perspectives.  The data for the dependent variable for this portion of the analysis are drawn from Freedom House, one of the most widely employed databases of regime type.  Second, an aggregate analysis of thirty cases of democratic reversion is undertaken.  These thirty cases represent one universe of cases based on the Polity IV date which is the other most widely employed databases of regime type.  The purpose of this step is to test the relevance of the first-stage quantitative 45  findings by exploring the general characteristics of this group of cases.  Third, an analysis of each of the thirty cases is undertaken as they specifically relate to the independent variable being addressed in the chapter.  The purpose of this step is to examine the findings of the previous sections in the particular context of each of the thirty individual cases.  Next a rational choice account of democratic reversion is introduced as a means of drawing attention to the importance of the concept of democratic uncertainty.  Finally, the framework is tested against a single chapter-length case study of the Philippines. Rather than relying solely on the thirty cases democratic reversion, this last approach examines a case where the structural constraints are indicative of a situation that should result in a democratic reversion, however democracy persists.  The purpose of this portion of the analysis is to examine the relevance not only of the independent variables but also to test the relevance of the rational choice explanation laid out in the chapter on democratic uncertainty.  The use of a wide assortment of methodological approaches is employed in an effort to put the existing literature through a vigorous test and to increase the confidence in the dissertation’s findings. Cross-National, Time Series Analysis  Each of the “structural constraints” chapters begins with a cross-national, time- series analysis.  In these sections, a “country year” is the unit of analysis.  This means that every year that a country is in existence during the period under examination is treated as an independent and discrete entity.  The data for the dependent variable for these portions of the analysis are drawn from the Freedom House data set (also referred 46  to in the literature as Gastil) consisting of the scores reported in The Comparative Survey of Freedom beginning in 1972 and updated on an annual basis. While this data has been used extensively in social science research (Heo and Tan 1997, 7; Bollen 1993, 1210-1211; Leblang 1997, 457; Burkhart and Lewis-Beck 1994, 904), a commonly expressed concern is whether it provides reliable indicators of change. One way to test this is to examine the Freeedom House data in relation to other widely used data sets.  Heo and Tan (1997, 8) found that the Freedom House and Arat data sets are highly correlated.  Burkhart and Lewis-Beck (1994, 906-7) compared Freedom House to Bollen’s “uncontaminated” 1980 index and found it highly correlated, concluding little error exists in the Freedom House index and that it does an “excellent job of tapping the underlying real variable of democracy.” The various time series data sets that examine regime type prove to be very similar in terms of regime classification.  Przeworski and Limongi (1997, 179) argue, “from a practical point of view, alternative measures of democracy generate highly similar results.  The dimensions used to assess whether or to what extent a particular regime is democratic seem to make little difference.”  Further, Inkeles (1990, 5-6; as quoted in Przeworski et al. 2000, 56) argues, “the indicators most commonly selected to measure democratic systems generally form a notably coherent syndrome, achieving high reliability as measurement scales…A testimonial to the robustness of the underlying common form and structure of the democratic systems is found in the high degree of agreement produced by the classification of nations as democratic or not, even when democracy is measured in somewhat different ways by different analysts.” 47  One concern, not commonly voiced, is that the manner in which these data are employed throughout the literature may not actually represent the dependent variable being discussed in most cases.  To address this concern, in Chapters 3 - 6, the dependent variable is examined from four different perspectives.  The data for this section are drawn from the Freedom House annual ratings which are based on adding the “civil liberty” and “political rights” scores a country received in any given year.  Both of these categories are rated 1 to 7 (1 being most democratic and 7 being least democratic).  Thus every country in the world ends up with a 2 to 14 rating for every year beginning in 1972.  The first, most general approach to analyzing the dependent variable, attempts to make use of all the variance contained in these ratings; running the analysis based on the dependent variable ranging from 2 to 14.  This perspective on the dependent variable examines the relationship between the independent variables and the various gradations of regime type. Were this approach to work, it would provide us with information regarding the correlation between regime type and the independent variables in question.  Such an analysis does not examine the movement between the different levels of regime type and specifically, it does not tell us anything about the reversion from democracy.  In addition, the amount of missing data for many of the independent variables renders judgments about even these questions highly problematic. In order to overcome this problem, the analysis then turns to a less finely divided approach to the dependent variable.  Freedom House aggregates the scores into three basic ratings: Free (2-5), Partly Free (6-10) and Not Free (11-14).  Approaching the dependent variable from this perspective overcomes some, but not all, of the problems with missing data/cells.  It does not, however, address the objection that while such an 48  analysis may tell us something about the relationship between issues related to the independent variables and the three aggregate levels of regime type, it does not address the movement between the different regime types. To address this shortcoming, the analysis turns to a specific examination of the reversion from democracy.  To accomplish this, countries receiving a Free rating in any given year are considered democratic.  When they receive a Partly Free or Not Free rating, they are considered as not democratic.  Thus, a democratic reversion is indicated for the year in which a country that received a Free rating the previous year, receives either a Partly Free or a Not Free rating.  From 1972 through 2003, the Freedom House data indicate 56 such events (see Table 1).  So the third analysis in this section examines these 56 country-years in comparison to all the remaining country-years.  This approach explicitly addresses the problems associated with attempting to draw conclusions about democratic reversion based simply on an analysis that examines the difference between basic regime types.  Rather, it specifically compares the instances of democratic reversion with country-years where no reversion occurred. The difficulty with this approach is that all non-reversion country-years (both democratic and non-democratic) are treated as relevant.  Such an analysis explores the question of whether there is something unique about democratic reversion country-years. For our purposes, however, it is necessary to undertake one final approach.  In this last step, only democratic country-years are compared to democratic reversion country-years. The reason for this is that while it may be interesting to understand what makes a country that experiences a democratic reversion different than all the countries (country-years) 49  that do not experience such a reversion, what we are explicitly interested in is why countries experience democratic reversions.  In order to experience such a reversion, it is necessary to first be democratic; hence the exclusion of all non-democratic country-years from the analysis.  To be clear, in this final stage of the analysis, all countries that have never experienced democracy (received a Free rating) are excluded.  If a country moves back and forth between regime types, only the democratic years and the first non- democratic year (the democratic reversion event) are included.  Of course, countries maintaining a constant democratic rating across the time period under consideration are also included. Case Analysis  As a function of the stated desire to pursue a methodologically diverse approach to the examination of democratic reversion, the examination of the “structural constraints” chapters (3 - 6) does not stop with the quantitative analysis just described. Rather, in each of these chapters, the quantitative analysis is followed by a more in-depth analysis of thirty cases of democratic reversion (See Table 2).   These cases were drawn from the Political Instability (State Failure) Problem Set: Internal Wars and Failure of Governance, 1955-2004 (principle investigators: Monty G. Marshall, Ted Robert Gurr and Barbara Harff).  The Political Instability Task Force project was originally housed at the University of Maryland under the direction of Ted Robert Gurr and has since moved to the University of Maryland under the direction of Monty G. Marshall (for more information, please see:  http://globalpolicy.gmu.edu/pitf/).  In order to select the specific cases to investigate for this portion of the analysis, the Adverse Regime Changes section 50  of the project is utilized.  As the PITF Problem Set Codebook (http://globalpolicy.gmu.edu/pitf/pitfcode.htm) indicates, the main criteria used to identify adverse regime changes is a six or more point drop in the value of a state’s Polity IV index score over a period of three years or less.  The Polity index is a measure of the institutionalized regime authority characteristics of the central state.  The index scale ranges between negative ten (fully institutionalized autocracy) and positive ten (fully institutionalized democracy).  The six point drop standard indicates that a substantial change has occurred in the authority characteristics of the regime as it is associated with qualitative changes in the openness of executive politics or general political competitiveness.  Borderline cases (within two points of the threshold) were all individually reviewed by the PITF team. At this point, cases are divided according to three variables that tap into different aspects of adverse regime change or collapse.  For the purposes of this analysis, the Magnitude Scale 2 (MAGCOL): Collapse of Democratic Institutions is examined.  This is a four point scale that identifies situations in which democratic or quasi-democratic institutions are weakened or replaced by autocratic political institutions.  Here, the cases coded as 1 (abrupt or disruptive transitions within autocratic political systems) and 2 (democratic or quasi-democratic institutions continue to exist but in circumstances of violent challenge and weakening of central authority are excluded.  Instead, the analysis included only cases coded 3 (a quasi-democratic polity is forcefully replaced by violence or threat of violence by an autocratic political system) and 4 (a fully democratic polity is forcefully replaced by violence or threat of violence by an autocratic political system).  A democratic regime is one whose Polity score is greater than +5 and a quasi-democratic 51  regime is one with a Polity score less than +6 and greater than 0.  These selection criteria result in thirty instances of democratic reversion in twenty seven countries (there were two reversions in Comoros, Haiti and Pakistan).  Nine of the cases were MAGCOL 4 - reversion from ‘full’ democracy, while the remaining twenty one cases were MAGCOL 3 – reversion from ‘quasi’ democracy (see Table 3 for the PIFT description of the reversion event). In the aggregate case analysis section of the “structural constraints” chapters, the behavior of the independent variables is examined across the thirty cases.  Rather than simply relying on the quantitative analysis at the beginning of these chapters, those finding are examined in light of the thirty cases.  The first portion of this examination explores the thirty cases as a single set.  So the level and movement of the independent variables is mapped against an aggregate analysis of the cases.  The second portion of this examination undertakes an analysis of each of the thirty individual cases in light of the structural constraint being considered.  To accomplish this, an in-depth study of each of the cases was undertaken.  The foundation for each of these thirty studies was based on a variety of commonly employed country survey sources including Europa World Plus, Keesing’s World News Archives, Freedom House’s Freedom in the World, the CIA World Factbook, and Facts on File (for an example of research employing these data, see Power and Gasiorowski 1998, 137).  Additional secondary sources were consulted as needed for each individual case. It is notable that the dissertation draws the dependent variable data from the Freedom House and Polity IV datasets.  These datasets represent the two most widely 52  employed cross-national, time series databases of regime type.  As previously indicated, while there is a substantial debate as to the superiority of one source over the other, there is also a very high correlation between the two databases.  It is important to note that this high correlation is in part due to agreement on the non-controversial cases.  There is a much more disagreement about the transition cases (see, for example Ward 2002, 47-49). This is a problem all democratization research that examines a variety of cases across time and space encounters.  The result is that the lists of cases of democratic reversion (or any form of transition) end up being quite different.  As there is no set of reversion cases that has achieved widespread agreement, scholars are left to pick and choose. When selecting a list of cases, scholars encounter a number of challenges.  One method of selecting the cases is to decide where you stand on a variety of debates such as (for example) minimalist versus maximalist definitions as well as dichotomous versus scaled indicators and then attempt to select the dataset that best reflects your positions. The resulting research does end up getting caught up in these prior debates and your findings are only as good as your ability to defend those selection criteria.  More likely, one’s ability to defend such criteria is not as relevant as the initial perspective on these issues that the reader brings to the article.  A separate problem is that all lists drawing cases across time and space which are sorted on an “essentially contested” concept such as democracy will face criticism from regional or country experts for their inclusion or exclusion of cases.  No such list is ever going to be perfect and scholars with more in- depth knowledge of regions and countries are always going to be able to pick apart these cross time and space lists.  A common response to this is to simply create your own unique list that attempts to address as many of these concerns as possible.  Of course, this 53  opens one’s research to criticism from other cross-national scholars that the findings of such research is of limited by the peculiarities of the selection criteria as well as a more general adaptation of such a list. In response, the dissertation draws on both the Freedom House and the Polity data.  These are the two most widely employed data sources for this type of case selection.  The notion behind employing two different lists of cases is that it should increase our confidence in the findings of the research.  There is an ongoing dispute in the literature over the definition of democracy and this dispute influences the case lists generated by the data sources.  Freedom House employs a more maximalist definition of democracy.  The effect of this is to identify “cleaner” cases of democracy.  The cost is that “grey zone” cases are more likely to be excluded from the analysis as the requirements for being identified are greater.  The Polity data employs a more minimalist approach to democracy.  The effect of this approach is the inclusion of a greater number of “grey zone” cases.  Employing both data sets allows us to explore the idea of democratic reversion from two slightly different perspectives.  The analysis relying on the Freedom House data in effect sets a higher threshold on democracy while the Polity data set a higher threshold on reversion.  Both of these approaches are important to our understanding of democratic reversion. In order to understand this position, the manner in which the data are employed to generate the lists of cases must be more fully explored.  While the discussion of when we should consider a country to have moved from a nondemocratic to a democratic regime is highly developed, the discussion of when a case is considered to have moved from a 54  democratic to a nondemocratic regime is highly undeveloped.  And the observation that the later should be obvious from the former ends up not being so obvious.  The practical effect of this, in terms of generating a list of cases, can be seen in the issue of magnitude. When we attempt to analyze regime type employing a larger scale measure (such as the Freedom House 2-14 scale), the problem of missing data really begins to interfere with our ability to undertake meaningful analysis (there are too many empty cells).  As a result, most analysis aggregates these scales into some kind of a dichotomous variable (which is how the list of cases is generated).  In our analysis of the Freedom House data, a reversion is considered to have taken place when a country’s aggregate categorization moves from Free to Partly Free or Not Free.  These are admittedly fairly blunt categorizations and should be used with caution.  Still, others have employed this approach and it does generate a fairly reasonable list of cases.  The problem is the magnitude of change is washed out.  One advantage of PITF’s MAGCOL list is that it relies on a measure of magnitude to generate the list of cases.  A country has to experience a six point drop on the Polity scale (which ranges from -10 to +10) in order to be coded as a reversion.  The advantage of this is that the cases generated for this list have experienced a move of substantial magnitude. The difficulty with this approach is that a lot of potential cases that are washed out.  Rather than attempting to rescale the Freedom House data to account for magnitude, the decision here was to employ the MAGCOL approach but to relax the criteria for being considered a democracy by including both “full” and “quasi” democratic cases. The practical effect of this decision is that the case list that is generated includes some controversial cases.  However, if we are interested in why countries experience reversions 55  from democracy, the tradeoffs inherent in this decision should strengthen the dissertation’s findings.  It is exactly the countries that are hovering near one side or the other of any definitional division of democracy that are the most likely to experience regime problems.  If these are the cases that are most at risk of a reversion from a more democratic regime to a less democratic regime, it seems highly relevant to discover if the explanations of regime change are relevant not just to reversions of solidly democratic regimes but also to countries that have made substantial democratic progress yet remain controversial classifications.  This should also serve to increase the relevance of the findings in this research.  Policymakers are much less concerned with the nuance surrounding the discipline’s definitional debates and more concerned with the applicability of research to the problem of democratic stagnation that we are currently witnessing.  One potential solution to the most controversial cases would be to simply exclude them from the list.  The difficulty with this approach, as previously mentioned, then becomes that the researcher is in effect creating a custom list with custom criteria which will generate findings that will be seen as particular to those choices.  Given that the list generated from these choices will be used for the case studies side of the analysis, it seems cleaner to include all of the cases and simply address any oddities in the individual case analyses. So the argument behind the decision to employ these different data sources to conduct the analysis for two of the four basic methodological approaches employed in the dissertation is that our findings should be more robust.  Rather than relying on a single data source, we can use these different approaches as a means of undertaking an examination of two slightly different approaches to the idea of democratic reversion.  The 56  Freedom House approach sets a higher threshold for a regime to be considered as democratic while the PITF/Polity approach sets a higher threshold for regime to be considered as having experienced a reversion.  The dissertation’s findings from the analysis utilizing these two approaches should thus be somewhat less vulnerable to criticism based on the methodological baggage carried by one or the other of these two data sources.  This choice is consistent with the attempt to examine the issue of democratic reversion by employing a variety of methodological approaches and results in a more robust exploration of the dependent variable than is commonly found in the literature. Rational Choice Analysis  The next methodological approach employed is rational choice analysis.  In Chapter 7, a rational choice approach is introduced as a means of drawing attention to the importance of the concept of democratic uncertainty.  While this idea has enjoyed a good deal of theoretical attention, it is usually set aside when research turns to some form of empirical analysis (Schedler 2001, 5).  While it is beyond the scope of the current project to attempt to operationalize and test the relevance of democratic uncertainty in a cross- national, time-series analysis, the work done in this chapter pushes the literature forward by explaining how the inclusion of this concept in a rational choice account of democratic reversion changes considerations actors may hold about their level of satisfaction with the goods received, their time horizons for evaluating the benefits they receive under a democratic regime as well as how they evaluate risk and cost.  Additionally, the rational 57  choice account introduced in this chapter draws attention to the theoretical difficulties of one year lags commonly employed in empirical research on democratization. The Philippines  The final methodological approach was to put the entire theory to the test against a single, in-depth case study.  One potential critique of the two sets of analyses based on the thirty cases of democratic reversion is that it is based on a Most Different Systems (MDS) analysis.  In this approach, a set of very different cases are selected based on their common score on the dependent variable.  The two different analyses of thirty case employed in the previous section is based on an MDS approach because it looks at the universe of democratic reversion cases during the time period under consideration. While this is one of the most commonly employed research methods in Comparative Politics, it could be open to a criticism based on the position the MDS approach cannot account for the action of cases that score differently on the dependent variable.  In particular, the extensive analysis of the thirty cases of reversion is open to the critique that we are uncertain how non-reversion cases work.  While this is offset by the prior quantitative analysis, it does provide the opportunity to test the framework against a single case. So, in response, an in-depth case study of the Philippines was undertaken, concentrating on the period following the transition to democracy in 1986.  As will be discussed in more detail in the chapter devoted to this case, the Philippines case was selected because the structural constraints pointed to a case that should have experienced multiple democratic reversions, yet democracy persisted.  Thus the purpose of this 58  chapter is twofold.  First, it provides a case study examination of the structural constraints.  And second, it allows for an in-depth examination of the relevance of democratic uncertainty in the decision-making process surrounding regime support calculations. It is worth noting that it is difficult, if not impossible, to arrive at a satisfactory set of cases that have not experienced democratic reversion that can be compared in an aggregate manner to a set of cases that experienced a reversion.  The basic problems are how to select the list of countries and how to select the year (or specific time period) to study.  One approach to such an undertaking is to take a random sample of democratic countries during a particular year and compare them to the reversion cases.  Dishkin, Dishkin and Hazen (2005), for example, compared their 30 cases of collapsed democracies to a group of 32 non-European and non-highly developed cases.  In order to examine their independent variables, they looked at the year prior to the collapse for the democratic reversion cases and 1998 for the democratic cases.  It is troubling that they provided no justification, at all, for the selection of the year 1998.  Since the reversions take place across time, there is certainly the (unexplored) possibility that events common to 1998 could bias the results. Another approach would be to treat every country year as a distinct entity (a la Przeworski et al. 2000).  One difficulty with this approach is that it is pretty much impossible to consider the nuance of many thousands of country years.  One could also look at paired comparisons of cases.  This still leaves unanswered the issue of how to select the ongoing democracy case.  Efforts to select by geography, similar stage in 59  development, by year, etc., all present difficulties.  A regional Most Similar Systems approach, attempting to hold constant as many independent variables as possible is another potential approach.  Again however, the issue of a baseline for comparison (year, stage, location, etc) is left unanswered.  Given the difficulties with any approach to comparing sets of cases, the decision to study all the reversion cases using an MDS approach and then to use a single, in-depth study across two decades of potential democratic reversion triggers seems reasonable. Conclusion  This chapter introduces a framework for explaining the reversion from democracy based on the notion that actors base their decision making around support for democratic regimes on traditional pocketbook issues.  Actors are concerned with the ability of democratic regimes to deliver an acceptable level of goods.  When regimes fail to deliver goods at an acceptable level, actors withdraw their support for the regime and the potential for a democratic reversion increase.  These actors are considered as mainly interested in economic goods.  So the chapter begins with an explanation of the importance of both the level of economic development as well as economic performance. Countries at higher levels of economic development are less vulnerable to reversion.  As well, countries with higher levels of economic performance (in terms of economic growth and inflation rates) are less vulnerable to reversion.  A country’s political institutional configuration is also relevant to the issue of democratic reversion.  Here it is argued that when governments fall into legislative gridlock the country becomes more vulnerable to democratic reversion as the government’s ability to address pocketbooks issues is 60  perceived as compromised.  In this regard, executive structures, the political party systems and electoral rules are taken into consideration and presidential systems, systems with higher party fractionalization, and systems employing proportional representation voting are seen as more vulnerable to democratic reversion. Further, a country’s vulnerability to international pressure is considered.  As international pressure can directly affect pocketbook issues, countries that are less vulnerable to such pressures are more likely to experience democratic reversions.  As well, as the military is the most common veto player in a regime change situation, the pocketbook issues of the military are considered.  While democratic regimes are often counseled to assert civilian control over the military, this is often accomplished in a manner that affects the military’s pockets (its budget or personnel levels).  As such, the countries that confront the military in an effort to assert increasing civilian control are more vulnerable to democratic reversion. Finally, the idea of democratic uncertainty is introduced as a means to further understand what levels of goods are acceptable as well as what time horizon actors consider when evaluating their support for democratic regimes.  When actors perceive higher levels of democratic uncertainty, they are more willing to tolerate a lower level of goods and take a longer time horizon into account when evaluating their regime support. So countries with lower democratic uncertainty are more likely to experience democratic reversion. The chapter concludes with an explanation of how the empirical approach pursued in the dissertation relies on a belief in the importance of methodological pluralism.  The research makes use of a wide variety of methodological approaches in an 61  effort to undertake a rigorous test of the issues under examination as well as to increase the confidence in the ultimate findings.  The structural constraints chapters are examined though a combination of cross-national, time-series analysis, an aggregate analysis of a set of thirty reversion cases, and a set of analyses of those case studies in light of the topic addressed in a given chapter.  A rational choice account of reversion is introduced which draws attention to the importance of the notion of democratic uncertainty.  Finally, the framework is tested against a single, in-depth case study.  The use of a wide assortment of methodological approaches is thus employed in an effort to put the existing literature through a vigorous test and to increase the confidence in the dissertation’s findings.  62  Table 2.0.1: Cases Losing “Free” Rating (Freedom House, 1973 – 2003)  Antigua and Barbuda 1991 Argentina 1974 Argentina 2001 Bangladesh 1993 Bolivia 2003 Brazil 1993 Burkina Faso 1980 Chile 1973 Colombia 1989 Cyprus (Greek) 1974 Dominican Republic 1974 Dominican Republic 1993 Ecuador 1996 Ecuador 2000 El Salvador 1976 Estonia 1992 Fiji 1987 Fiji 2000 The Gambia 1981 The Gambia 1994 Ghana 1982 Grenada 1979 Guatemala 1974 Guyana 1973 Honduras 1993 Honduras 1999 India 1975 India 1991  Latvia 1992 Lebanon 1975 Malawi 1999 Malaysia 1974 Maldives 1973 Mali 1994 Malta 1982 Mauritius 1978 Nepal 1993 Nigeria 1984 Papua New Guinea 1993 Papua New Guinea 2003 Peru 1989 Philippines 1990 Seychelles 1977 Slovakia 1996 Solomon Islands 2000 Sri Lanka 1982 Suriname 1980 Suriname 1989 Thailand 1976 Thailand 1991 Trinidad and Tobago 2001 Turkey 1980 Vanuatu 1982 Venezuela 1992 Venezuela 1999 Zambia 1993 Source: Freedom House, Freedom in the World, 1972-2004 63  Table 2.0.2: Democratic Reversion Cases Adverse Regime Change Data Set: Political Instability (State Failure) Problem Set, 1975-2004  Albania   1996 Armenia    1996 Azerbaijan   1993 Belarus   1995 Burkina Faso   1980 Burundi   1993 Cambodia   1997 Central African Republic 2003 Comoros   1976 Comoros   1999 Congo-Brazzaville  1997 Fiji    1987 The Gambia   1994 Ghana    1981 Guinea-Bissau   2003 Guyana   1978 Haiti    1991 Haiti    1999 Iran    2004 Niger    1996 Nigeria   1983 Pakistan    1999 Pakistan   1977 Peru    1992 Sierra Leone   1997 Sudan    1989 Thailand   1976 Turkey    1980 Uganda   1985 Zimbabwe   1987  Source: Monty G. Marshall, Ted Robert Gurr and Barbara Harff, Political Instability (State Failure) Problem Set: Internal Wars and Failure of Governance, 1955-2004, http://globalpolicy.gmu.edu/pitf/ 64  Table 2.0.3: Democratic Reversion Event Descriptions Adverse Regime Change Data Set: Political Instability (State Failure) Problem Set, 1975-200425  Albania 1996 Third post-communist parliamentary elections are marked by bloody police repression and electoral fraud.  President Berisha uses intimidation, violence and fraud to consolidate his political power.  Armenia 1996 President Ter-Petrossian suspends the country's most influential opposition party. Electoral malpractice and government intimidation tarnish subsequent legislative and presidential elections.  Azerbaijan 1993 Post-Soviet democratic transition undermined by government instability, armed insurrection and fraudulent presidential and legislative elections.  Belarus 1995 A bitter, political feud between President Lukashenka and the Supreme Council (legislature) culminates in a popular referendum. Voters strongly approve the President's initiative and institute constitutional changes that strengthen the executive office  25The descriptions of the reversion events for each case can be found in the Adverse Regime Change Problem Set excel file located at  http://globalpolicy.gmu.edu/pitf/pitfpset.htm 65  (allowing the President to rule by decree), disband the Supreme Council, and set up a new bicameral legislature subordinate to the executive.  Burkina Faso 1980 Leader of former military regime, President Lamizana, elected as head of civilian government.  Subsequent economic crisis and labor unrest triggers military coup and suspension of constitution.  Burundi 1993 Opposition forces win first multiparty elections, ending longstanding rule by Tutsi minority. Coup by Tutsis officers aborts transition to democracy as ethnic clashes escalate to civil war. Subsequent attempt at multi-ethnic civilian government falls in second Tutsi coup.  Cambodia 1997 Hun Sen ousts coalition partner and ends fractious coalition government installed following UN-supervised elections in 1993. Hun Sen consolidates power in a new coalition.  Central African Republic 2003 Forces loyal to Gen. Bozize succeed in ousting government of elected-President Patasse while he is out of the country, having failed in several earlier attempts.  66  Comoros 1976 Twenty-eight days after the declaration of independence, a coalition of six political parties known as the United National Front overthrew the Abdallah government, with the aid of foreign mercenaries. After the coup, a three-man directorate took control.  Comoros 1999 Army Chief of Staff Col. Assoumani Azzali leads April 30, 1999, coup that dissolves constitution and government. Promised transition to new elections based on Antananarivo agreement does not materialize.  Congo-Brazzaville 1997 Transition to democracy ends when Sassou-Nguesso ousts President Lissouba after five months of fighting.  Fiji 1987 A parliamentary election in 1987 brought the Indian party to power. The elected government was ousted in a military coup and Lt. Col. Sitiveni Rabuka assumed control.  The Gambia 1994 Long-standing multiparty system, dominated by President Dawda, is overthrown in military coup.   67  Ghana 1981 Limann's People's National Party (PNP) began the Third Republic with control of only seventy-one of the 140 legislative seats; the percentage of the electorate that voted had fallen to 40 percent. Unlike the country's previous elected leaders, Limann was a former diplomat and a noncharismatic figure with no personal following. As the country’s economy continued to decline and widespread strikes threatened to shut down the government, Jerry John Rawlings led a successful coup on and established personalistic rule backed by the AFRC.  Guinea-Bissau 2003 New elected-government of President Yalla and former-opposition parties is beseiged by challenges and continuing instability. Armed forces led by Gen. Seabre oust Yalla and establish junta to rule country until new elections are held.  Guyana 1978 Political domination of the black-based PNC consolidated with the abrogation of the democratic constitution and the use of electoral fraud.  President granted unlimited powers in new one-party state.  Haiti 1991 Populist priest Jean-Bertrand Aristide elected president by a large majority, is unwilling to govern within the political system and alienates the elite and foreign community.  He is overthrown and replaced by a military supported puppet government. 68   Haiti 1999 Following two years of stalemate between the executive and the opposition-led legislature, President Preval dissolves the legislature and rules by decree. The President uses unchecked executive power to ensure electoral victory for his party, Fanmi Lavalas, in 2000 legislative and presidential elections.  Iran 2004 Under increasing internal and external pressures related to US military operations in neighboring Afghanistan and Iraq, the theocratic Council of Guardians prohibits reformist candidates from standing for election.  Niger 1996 Military coup overthrows democratically elected government and suspends the 1992 constitution.  Coup leader, Col. Ibrahim Mainassara Barre, elected president in seriously flawed elections.  Nigeria 1983 Ethnic competition, widespread corruption and electoral malpractice weaken the democratic institution of the Second Republic.  Successive military coups bring to an end the Second Republic and expand the role of the armed forces in the political arena.   69  Pakistan 1999 Gen. Musharraf leads military coup on October 12, 1999; arrests democratically-elected Prime Minister Sharif, suspends consitution, dissolves parliament, and imposes rule by National Security Council.  Pakistan 1977 Democratic government overthrown in military coup as political violence escalates in the wake of surprise parliamentary elections.  General Zia dissolves the legislature, arrests politicians and declares martial law.  Peru 1992 President Fujimori, backed by the military, dissolves Congress and suspends the constitution.  Sierra Leone 1997 Mutinous soldiers side with RUF guerrillas to overthrow President Kabbah. Junta is defeated by ECOMOG in February 1998 but violence continues. Brokered peace agreement is reached between Kabbah government and RUF forces in May 2001 and State of Emergency is lifted in March 2002.     70  Sudan 1989 Military overthrows democratic government after attempts to reduce the influence of religion in politics.  Legislature dissolved and non-fundamentalist parties banned as an Islamic state is established.  Thailand 1976 Persistent guerrilla insurgency and open warfare between leftist students and rightist paramilitary groups triggers military coup.  Military establishes a hardline civilian government that restricts political liberties and civil rights.  Turkey 1980 Parliamentary instability and widespread social unrest triggers military coup.  Political activity banned as the military lays the groundwork for the restoration of democracy under military supervision.  Uganda 1985 An army brigade composed mostly of ethnic Acholi troops took Kampala and proclaimed a military government, replacing President Obote, who had been elected in 1980 but had failed to accommodate or contain Museveni's popular National Resistence Army insurgency.    71  Zimbabwe 1987 Ethnic tensions and crackdown on political opposition weakens Zimbabwe's fragile democratic institutions.  Merger of ZAPU with ruling ZANU effectively establishes a single party system.      72  Chapter 3: Economic Issues and Democratic Reversion 73  Introduction One of the most widely accepted notions in democratization is that a regime’s economic situation is critical to the survival of democracy.  It is this understanding that undergirds the importance of a framework that relies on the importance of pocketbook issues.  Yet, when we examine the instances of democratic reversion, we see cases like Armenia which in the three years leading up to the reversion experienced economic growth rates of 7.9%, 9.1% and 7.5%.  While cases such as this may not be the norm, twelve of the twenty nine case studies examined in this chapter experienced positive economic growth during the year the reversion year and thirteen experienced positive growth the year before the reversion.  Likewise, nine of the twenty six cases had inflation rates under 10% during the reversion year.  And while none of our thirty cases were among the most economically developed countries in the world, four had GDP per capita levels above $1500 and nine more were above $1000.  Despite findings in this chapter of significant relationships between democratic reversion and economic development, economic growth and the consumer price index, we find these case study results troubling. The question of if and how economic issues influence democracy (and vice versa) has been debated in comparative politics for at least the last five decades.  Some of the earliest research in this area of study was concerned with a search for preconditions to democracy.26  These early efforts to study the connections between economic issues and  26 See, for example, Seymour Martin 1959; Huntington 1968; Almond and Verba 1963; Moore 1966; Dahl 1971; Cutright 1963; Binder 1971. 74  regime were greatly expanded in the 1990s.27  More recently, Przeworski et al. (2000) offered evidence that supported the exogenous theory (development makes democracies, once established, less likely to fall to dictatorships) but not the endogenous one (development increases the likelihood that poor countries will undergo a transition to democracy) (Boix and Stokes 2003, 517).  While a number of scholars have since risen to the defense of endogenous democratization, the notion that “the dynamics of achieving democracy and sustaining it may not be the same” has gained widespread acceptance in the literature (Biox and Stokes 2003, 545).  In addition, a variety of scholars have argued there is a connection between poor economic performance and regime change.28 This chapter endeavors to explore these debates as they relate to democratic reversion.  In particular, we examine the hypotheses that democratic regimes at lower levels of economic development are more likely to experience reversion, and that democratic regimes with poor economic performance are more likely to experience reversion.  To accomplish this, three different methodological approaches are employed. First, a cross-national, time-series analysis of economic development and regime type is undertaken.  Working from a general examination of development and regime type down to a specific analysis of reversion of democracy in democratic regimes, allows us to shift the focus from general associations between development and regime type to the specific  27 See, for example, Feng 1997, 392-4; Leblang 1997, 453.  Also see for example, Burkhart and Lewis- Beck 1994; Haggard and Kaufman 1995; Heo and Tan 1997; Przeworski and Limongi 1997; Przeworski, Alvarez, Cheibub and Limongi 1997; Remmer 1990, 327; Remmer 1993, 405. 28 See, for example, Diamond 1999, 77-93; Diamond and Linz 1989, 44-46; Gasiorowski 1995; Geddes 1999; Haggard and Kaufman 1995; Lipset et al. 1993; Przeworski et al. 2000. 75  issue of development and reversions from democracy.  This portion of the analysis leads us to conclude that we may safely reject the null hypothesis that there is no relationship between economic development and democratic reversion.  In fact, we found a strong relationship between both economic development (GDP per capita) and democratic reversion as well as between economic growth (GDP per capita growth) and democratic reversion. We then turn to the second methodological approach involving an aggregate analysis of thirty contemporary cases of democratic reversion.  Relying on a basic Most Different Systems analysis, we hold the dependent variable constant and look for consistent effects from independent variables across cases.  Based on this analysis, the conclusion reached is that we should not reject the null hypothesis.  The economic effects widely vary across the cases making it difficult to conclude there is an important relation to democratic reversion.  This conclusion is further confirmed by the final methodological approach in which an analysis of the individual case narratives is undertaken.  It is important to recognize that none of the reversion cases involve countries with highly developed economies.  To that extent, we can simply confirm the findings of Przerowski et al. (2000) that at high levels of economic development, democracies to date are immune to democratic reversion.  So while it is important to recognize that economic development issues may play a role in democratic reversion, we strongly caution against overstating the extent of the role played by economic performance issues. 76  These findings are important for understanding the relevance of the main theoretical position advanced in the dissertation.  Research examining the interaction between economics and regime usually has a structural orientation.  As such, it assumes actors respond mainly to economic stimuli and does not explicitly address the elements more commonly associated with process/choice approaches.  The basic findings of this chapter are at odds with the position that the economic situation of a case predicts reversion.  Given this, the position taken by Mainwaring and Pérez-Liñán (2005, 30) that the limited performance of structural models in Latin America suggests that scholars were right to emphasize process and choice, seems highly relavant.  While the findings in this chapter indicate a role for economic inputs into the decision-making process, given the ambiguity of the results, a consideration of the situation that moves beyond the one year time horizon that almost all cross-national, time-series research relies upon (see, for example, Przeworski and Limongi 1997, 169) seems warranted. Cross-National, Time-Series Analysis This section of the chapter examines the relationship between economic issues and democratic reversion during the third wave of democratization.  So the time period under consideration is 1972 through 2003.  The data for the dependent variable are drawn from Freedom House’s Freedom in the World.  Data for the independent variables are drawn from the World Bank’s World Development Indicators. As to the independent variables, in this chapter we are concerned with the economic influences on democratic reversion.  So in this section of the chapter, we examine three basic variable groupings.  The first addresses the level of economic 77  development.  Drawing on the work that has taken place on democratization over the last four decades, the two main variables considered are Gross Domestic Product (GDP) per capita and GDP per capita growth.  The second group examines inflation.  The main variable considered here, also drawing on a long history in the literature, is the annual inflation rate.  Two secondary variables, consumer price index and food price index, are also considered.  The third group explores issues around what we consider to be more social economic issues.  The main variables under consideration here are literacy and infant mortality.  As previously discussed, all data for these independent variables are drawn from the World Bank’s World Development Indicators. Dependent Variable 1  We begin with an examination of the regime type variables utilizing the Freedom House indicator that provides the 2 to 14 rankings for each country-year (see Table 1). Turning first to the relationship between economic growth and regime type, we find that GDP per capita is significant (0.000), in the expected direction and does a decent job of explaining the variance in the dependent variable (pseudo R2 = 0.109).  This would allow us to reject the null hypothesis that there is no relationship which is consistent with most of the existing literature in the field.  Interestingly, when we turn to the GDP per capita growth variable, the results do not allow us to reject the null hypothesis (p=0.871).  So while the levels of democracy and economic development are positively correlated, annual changes in such growth do not appear to be correlated with regime type. Turning to the question of inflation, we found a significant (0.000) relationship in the expected direction between Consumer Price Index (CPI) and regime type.  So, higher 78  levels of the CPI are associated with less democratic regime scores.  The psuedo R2 score (0.026) however, points out that this variable does not hold a great deal of explanatory power.  Interestingly again, the relative changes in the economic measure (in this case, inflation) is not significant (0.996).  So while different levels in the price of a basket of goods is correlated with the different levels of regime type, annual changes in the price of this basket do not appear related to various levels of regime type. Finally, we turn to an exploration of the social economy and its relationship to regime type.  Here again, the results are mixed.  On the one hand, there is no significant relation (0.140) between Infant Mortality and regime type.  On the other hand, there is a significant relation (0.000) between literacy and regime.  The relation holds for both adult and youth literacy rates.  The pseudo R2 scores suggest some level of explanatory power for both adult (0.167) and youth (0.211) rates. Taking these results together allows us to reject the null hypothesis that there is no relation between economic issues and regime type.  While it would be desirable to analyze these variables together in a single model, such attempts are thwarted by the extensive amount of missing data.  When attempting to test a model considering more than one independent variable, the percentage of dependent variable levels by subpopulation with zero frequencies rises to such a high percentage that it is impossible to run the data.   79  Dependent Variable 2  Given the problems with attempting an analysis in which the dependent variable is divided into thirteen categories, we turn to an examination that is based on dividing the dependent variable into three categories (see Table 2).  Starting again with the issue of economic development, we find that the level of economic development (GDP per capita, p=0.000) and annual changes in economic development (GDP per capita growth, p=0.000) are both significant.  We can thus reject the null hypothesis that there is no relationship between economic growth and regime type.  The problem is that the pseudo R2 has substantially declined from the prior analysis.  Neither GDP per capita (0.047) nor GDP per capita growth (0.009) appear to be particularly important explanatory tools for explaining the difference between the three regime types. The findings with regards to  inflation are similar.  As with the first cut at the dependent variable, CPI is significant (0.000) but does not demonstrate a substantial explanatory role (Psuedo R2 = 0.020).  Again, as with the GDP findings just discussed, annual changes in CPI (Inflation) do play a significant role (p = 0.029) but as with CPI, the explanatory power is limited (pseudo R2 = 0.002).  It is interesting to note that in both cases, the annual change measure achieves significance as we lower the dependent variable from thirteen to three units. Finally, the examination of the social economy data also demonstrates more relevant measures.  In this case, both Infant Mortality Rates (0.000) and Literacy Rates (0.000 for both adult and youth) are significant.  Not only are we able to reject the null hypothesis that there is no relation between Infant Mortality and regime type, but the 80  pseudo R2 (0.390) leads us to believe this measure holds substantial explanatory powers. The findings on Literacy rates, on the other hand, are quite similar given either approach to the dependent variable.  The pseudo R2 for adult literacy (0.206) is a bit higher using this approach to the dependent variable, while it is a bit lower for youth literacy (0.164). Overall, the findings for the first two approaches to measuring the dependent are quite similar.  The most important difference is that the measures of annual change (GDP per capita growth and Inflation) are not significant when we employ a measure of regime that employs the full range of variance allowed by the Freedom House measures.  They are significant, however, when the regime type measure is aggregated into three categories.  As mentioned, it is likely at least part of this result is due to the extremely high percentage of empty cells that are produced when employing the more finely detailed approach to regime type. Dependent Variable 3  While these findings serve to confirm a large body of research on the relationship between economic issues and regime type, they still leave us in the dark as to any relationship between economic issues and democratic reversion (see Table 3).  To overcome this, we first turn to an analysis of the dependent variable that attempts to discern if there is something unique about country-years in which a democratic reversion take place as opposed to the remaining universe of country years.  In this portion of the analysis, all cases of a country moving from a Free rating to either Partly Free or Not Free are coded a 1 the first year the score changes.  All other country-years are coded 0. 81  Turning to the analysis of economic growth, both GDP per capita (0.005) and GDP per capita growth (0.009) are significantly related to democratic reversion country- years.  Given this, we may reject the null hypothesis that there is no difference between the model without the independent variables and the model with the independent variables.  Again, however, we see pseudo R2 scores that are quite low (0.011 and 0.010 respectively).  Any implications we attempt to draw based on these scores is confirmed by both low betas and related low odds ratio scores.  In the case of the level of economic development, the beta (0.000) and odds ration score (1.000) indicate that while the relationship is significant, movement of one unit of the independent variable results in no discernable change in the dependent variable.  In terms of GDP per capita growth, the beta, -0.046 (SE =.016) indicates some small role.  The odd ration score (.995) supports the claim that each one unit increase in the independent variable decrease the odds of democratic reversion 0.5%. Turning to inflation, neither CPI (p=.705) nor Inflation (p=.317) are significant. The social economy explanations fare no better.  Neither Infant Mortality Rate (.260) nor the Literacy Rates (adult =.363, youth = .585) are significant.  We are thus unable to reject the null hypothesis.  The lesson we can draw from this portion of the analysis is that the only variable that has both a significant relationship and one that moves the odds of a democratic reversion taking place is GDP per capita growth.   82  Dependent Variable 4  As previously discussed, there is a possibility that potential relationships are being obscured by the inclusion of too many kinds of cases in the dependent variable.  So the final iteration of this section of the data analysis examines the dependent variable where all of the democratic reversion scores remain the same as in Dependent Variable 3, but the non-reversion cases are substantially narrowed (see Table 4).  Any country-year not scored as a democratic reversion that receives a Partly Free or Not Free rating is excluded from the analysis.  We are thus left comparing ongoing democratic countries (country- years) with the democratic reversion years. In terms of the relationship between democratic reversion and economic development, the findings are significant and demonstrate a greater importance than was illustrated in the analysis of Dependent Variable 3.  In terms of the role of the level of economic development, GDP per capita, the relationship is significant (0.000).  More importantly, unlike the previous analysis, when examining this iteration of the dependent variable, the beta score, -0.002 (SE = .000), exceeds zero and is in the direction expected. The odds ratio (.998), at first appears small until one remembers the interpretation of this score is based on a one unit move in the independent variable.  In this case, a one unit move in the independent variable translates into a $1 change in GDP per capita.  So the odds ration score tells us that for every $1 increase in GDP per capita, the odds of a democratic reversion decrease 0.2%.  Likewise, GDP per capita growth is also significant (0.000).  As well, the beta, -0.097 (SE = .023) is larger than when the dependent variable included all the non democratic cases in the analysis.  The odds ratio (.908) indicates that 83  for each 1% increase in GDP growth, the odds of a democratic reversal decrease 9.2%. This is an important finding as it confirms the relevance of both economic development and economic growth specifically when comparing country years for ongoing democracies to the years in which countries experience a democratic reversion. When we examine inflation, no significant relation (p=.162) is found with democratic reversion (which is the same finding reported in the analysis of dependent variable 3).  However, the relation between CPI and reversion is significant (.036).  The beta score, 0.040 (SE = .019), is in the expected direction.  And the odds ratio, 1.041, indicates that for each 1 unit increase in CPI (which is measured using a baseline 2000 = 100), there is a 4.1% increase in the likelihood of democratic reversion. Unfortunately, we are unable to draw any conclusions about the relationship between social economic factors and democratic reversion due to the extensive amount of missing data.  In the case of Infant Mortality and both Literacy Rate indicators, the data for at least 75% of the 3,424 country-years are missing.  Any attempts to draw conclusion based on such data is nonsensical.  This conclusion was born out when we ran the analysis and found that, based on the 25% of the data that exists, there is in fact a positive relationship between all three variables and democratic reversion (Infant Mortality = .000, Adult Literacy = .024 and Youth Literacy = .000).  The odds ratios would lead us to believe that for each 1 unit increase in the Infant Mortality Rate (per 1,000 live births), the odds of a democratic reversion increase 3.5%.  While that makes intuitive sense, the odds ratio of Adult literacy would lead us to believe that for each 1% increase in Adult literacy the odds of a democratic reversion increase 121.3% while each 1% increase in 84  Youth literacy rate increase the odds of a democratic reversion a whopping 231.1%. While results like these might persuade anti-democratic forces to invest heavily in education in order to achieve regime change, we have a suspicion this strategy is unlikely to produce results of such a magnitude.  If nothing else, this provides a concrete example of the problems of attempting to run sophisticated statistical analyses with extensive amounts of missing data (as will be discussed in more detail later in the chapter). Discussion  The bottom line with regards to the strategy of undertaking a statistical analysis that begins by including every country in the world for the time period of 1972 through 2003, and examining the most typical concepts and indicators around the relationship between economic issues and regime type is we have demonstrated the importance of the relationship between economic development and reversion from democracy.  The findings are both interesting and important to the democratization literature.  As stressed earlier, too often in this field of study, researchers are interested in the relationship between regime type and economic development issues.  For the most part, their data only allows them to make claims about differences in regime type and how various economic development indicators are related to one type of regime or another.  However, they often impute conclusions about movement between different regime types, making claims about the conditions under which, for example, a democratic regime will fail.  The problem with such conclusions is, this is not what their dependent variable is measuring. In the analysis conducted for this chapter, we are able to replicate these findings via an examination of Dependent Variables 1 and 2.  While these analyses demonstrated 85  significant relationships, they also demonstrated that the net effects of such relationships were usually fairly marginal.  The next step in our analysis overcomes the objection just outlined by explicitly examining reversion cases.  While there is no important economic effects to be found when examining reversions against the universe of non-reversions, when the dependent variable is narrowed to measure what we are claiming to be of theoretical importance (that is the question of whether economic development is related to the possibility of a democratic country experiencing a reversion from democracy), the significant and relevant effects of the level of economic development as well as change in the level of economic development are demonstrated. The natural next step in the process would be to test a model including GDP and GDP growth and include a variety of control variables.  Unfortunately, there is so much missing data, that running such an analysis, in any manner that makes sense to us, is simply impossible.  Attempts to include even a few control variables into such a model are met with the warning that it is impossible to complete such an analysis due to the extensive amount of missing cases (usually well over 95%).  In response to this problem and additionally in pursuit of the objective of employing multiple methodological approaches in order to increase the confidence in our findings, we now turn to an analysis of case studies of democratic reversion. Case Study Data Analysis A case study of the events surrounding each of the thirty instances of democratic reversal was undertaken.  This chapter explores the relationship between economic issues and democratic reversion in these cases.  For the purpose of that analysis, three main 86  variables are examined: GDP per capita, GDP per capita growth, and Inflation.  The intent was to also include the two social economy indicators but that massive amount of missing data (+95%) made this endeavor impossible.  In this section of the chapter, we examine the aggregate data relevant to these cases.  For every case, the data under consideration are drawn from a six year period ending with the year the democratic reversion took place and going back through the five year period leading up to the reversion.  As most analysts rely on a one year lag, it was felt that a five year lag would provide a more complete picture of the data and allow us to test the validity of such lag- related decisions.  As well, it is our belief that the impact of the variables in question throughout the dissertation loses salience beyond five years.  In all cases, we will first examine the impact of variables assuming first that no lag should occur (simply look at the data during the reversion year) and then we will take the more common approach of examining the data given a one year time lag from the date of the reversion.29 GDP per capita      In order to test the relationship between GDP per capita (also referred to as economic development) and democratic reversion, we will begin with some basic descriptions of the data (see Table 5).  At the most basic level, the mean GDP per capita for the twenty nine cases for which we have data is 639.0 for the year prior to the reversion (R1) and 621.7 for the reversion year (R).  The median score for R1 is 461.4 while R is 480.0.  Right away, even at this most basic level of analysis, a problem  29 Please note that in the case of the Comoros 1977 reversion, almost all the data is missing across the various indicators. 87  becomes clear.  When we examine the mean for the entire sample, it declines from R1 to R (-17.4) while the median increases from R1 to R (+18.6).  Taking the full six year view, we see a steady decline in mean score for all twenty nine cases from a high of 691 at R5 down each year to a low of 622 at R.  The median score presents quite a different picture over the same six year period.  Beginning at 488 in R5, we see (basically) a decline over the next three years and then a recovery in R1 and a return to 480 by R. While this, in and of itself, is not damning, it is worth noting as it will stress a theme that will recur throughout this portion of the analysis: examination of the particulars of the case level data paint a much murkier picture than the now-standard binary logit analysis of 3,500 to 6,500 country-year data. As there is a lot of debate about the ability of democratic regimes to survive at different levels of development, it seems useful to look at the groupings of the level of development.  For the Reversion year, there are 13 cases at less than $1 a day (0-365), 8 cases at less than $2 a day (366-730), 5 cases at less than $3 a day (731-1095) and 4 cases above $3 per day.  The numbers for R1 are basically the same (12, 8, 5, and 4 respectively).  While this may appear to provide credence to the argument that poorer countries are more likely to experience democratic reversion, it should be noted that eight of the twenty nine cases took place about the $2 per day level and that four of those cases were between the 1500 and 2000 level.  As we are only looking at cases where the dependent variable was scored 1 (there are no non-reversion cases) we are unable to draw comparisons between reversion and non-reversion cases.  Rather, the strength of this portion of the analysis is found in its ability to debunk commonly held assertions regarding reversion through an examination of the universe of such cases during this time 88  period (the MDS approach).  There is little reason to examine the manner in which GDP per capita changes across the time period under consideration as GDP growth captures exactly this movement. GDP per capita Growth  Again, at the most basic level, the mean GDP growth across our twenty nine cases at R1 was -0.9 while at R it was -2.6.30  In this case, the median is almost identical (R1=- 0.7, R=-2.4).  These data fit the general notion that reversions take place when economic growth is in decline.  As we begin to disaggregate the data, problems with this conclusion become apparent (see Table 6). Beginning with an examination of the Reversion year, 17 cases experienced negative growth while 12 experience positive growth.  As the data are further disaggregated, it becomes apparent how magnitude comes into play.  Of the 17 negative growth cases, 8 experienced growth rates between 0 and -5, while 9 experienced rates in excess of -5.  Of the 12 positive growth cases, 8 were between 0 and 5.  In four cases, there was strong positive growth: Albania (10.1), Armenia (7.5), Sudan (6.7) and Thailand (6.7).  If, for the sake of argument, we assume that growth ranging between -5 and +5 is the norm for countries in this development range, 16 of our 29 cases would fit the criteria while only 8 of 29 fall outside this norm on the negative side.  While we recognize that this is a big spread and that any level of negative growth may be viewed as  30 As with GDP, the data is missing for Comoros 1977.  As well, the first three years of data are missing for both Azerbaijan and Uganda. 89  a bad thing, given the substantial fluctuation in growth rates across the cases under consideration, this range seems well suited to identify what falls outside the norm in such countries.  The picture of things at R1 is basically the same: 16 negative cases and 13 positive cases.  However, only 3 of the 16 negative cases fall outside the 0 to -5 range. Again, there are four cases above 5: Albania (10.3), Armenia (9.1), Fiji (6.8) and Iran (5.7).  From the perspective of a MDS approach, the wide variance across this explanatory variable appears to cast doubt on its ability to explain change in the dependent variable. Let us now turn our attention to examining the evolution of economic growth over time.  Beyond a simple analysis of the events of a particular year, it seems most people are capable of considering some kind of over–time perspective.  So, for example, if we were to hypothesize that the GDP growth rate was the main deciding factor for a given individual actor and we examined the reversion year and saw growth at -2, we could view this as confirmation of our hypothesis.  However, what if we looked at the previous five years and saw something like this: R5 = -15, R4 = -10, R3 = -7, R2 = -5, R1 = -4.  How is it possible, based on this example, to argue GDP growth was the deciding factor?  Sure you might argue there is some sort of negative cumulative effect and R was the straw that broke the camel’s back but it at least should counsel some caution in looking only at the R (or R1) year.  To take a more concrete example consider Peru (R5 = 5.6, R4 = -10.7, R3 = -13.6, R2 = -7.1, R1 = 0.2, R = -2.3).  Knowing nothing else about the case, it would seem prudent to consider that actors may not be pleased with a -2.3% GDP growth rate but that they put up with the regime though years of worse GDP growth, so why withdraw their support now?  Also, this is a really troublesome case for those who lag the 90  independent variable a year as the lagged year presents a completely different picture of the situation than does a longer analysis.  In addition, we need to remember that positive economic growth is not a panacea.  If a country used to experiencing double digit growth over a number of years suddenly slows to low single digit growth it could have the same effect as negative growth has in other situation. So, given the importance of looking at the evolution of growth over time, the question becomes, over what period of time?  The most obvious comparison would be the immediate past.  So, the first step is to compare the growth rate during the reversion year with the rate from one previous year.  We will begin by comparing growth rates at R1 with R.  On average, across our sample of cases, the mean growth was 1.8 points lower at R however the overall median difference was only -0.1%.  The difference in the rate of growth was negative in 18 cases and positive in 11.  Of the 18 negative cases, the decline in 14 was between 0 and 5 points.  And of the positive cases, 9 of 11 were between 0 and 5.  The notable outliers include Sierra Leone (-22.7), Fiji (-13.8), Guinea Bissau (+7.4) and Sudan (+9).  Again, if we look at our normal range, only 4 of 29 cases fall outside on the negative side.  It is also worth noting that in four cases, there was a decline in the growth rate from R1 to R yet growth was positive in R, while in three cases, there was an increase in the growth rate from R1 to R yet growth was negative in R.  Given the difficulties identified with the expected relationship, this analysis fails to confirm the relationship between GDP growth and democratic reversion. There is not a tremendous difference in findings if we lag the independent variable and compare R2 with R1. On average, across the sample of cases, growth was 91  0.9 (as compared to 1.8 for the non-lagged analysis) points lower at R1 however; the overall mean difference was only -0.1% (the same difference identified in the non-lagged analysis).  The difference in the rate of growth was negative in 20 (as compared to 18 for the non-lagged analysis) cases and positive in 9 (as compared to 11).  Of the 20 negative cases, the decline in 16 (as compared to 14 for the non-lagged analysis) was between 0 and 5 points.  And of the positive cases, 5 of 9 (as compared to 9 of 11 for the non-lagged analysis) were between 0 and 5.  It is worth noting that in 6 cases (as compared to 4), there was a decline in the growth rate from R2 to R1 yet growth was positive in R1, while in 2 cases (as compared to 3 for the non-lagged analysis), there was an increase in the growth rate from R2 to R1 yet growth was negative in R1.  These data are not substantially different than the data in the non-lagged analysis, thus supporting the conclusion that this analysis fails to confirm the relationship between GDP growth and democratic reversion. To take things one step further, it seems reasonable to provide at least some actors with a longer time horizon that twenty four months.  Thus we also compare the Reversion year with the average of the previous five years.  In 17 cases, there is a decline in the rate of growth, while there is an increase in 12 cases.  If we disaggregate the negative cases, we find that 8 declined from 0 to 5 points while 9 declined more than 5 points.  Of the positive cases, 3 of 12 increased more than 5 points.  So, were we to look again at the spread of -5 to +5 comparing the Reversion year to R1, only 9 of 29 cases fall outside of the range on the negative side.  This again seems to indicate there is no substantial relationship. 92  For the sake of consistency, we also compare R1 with the average of the previous four years.  In 16 cases, there is a relative decline in the rate of growth, while there is an increase in 13 cases.  If we disaggregate the negative cases, we find that 11 declined from 0 to 5 points while 5 declined more than 5 points.  Of the positive cases, 5 of 13 increased more than 5 points.  So, were we to look again at the norm of -5 to +5 comparing R1 to the Reversion year, only 5 of 29 cases fall outside of the range on the negative side. These results are essentially the same as the non-lagged analysis.  The only notable difference is 5 of 29 cases falling outside the norm as opposed to 9 of 29 in the non- lagged analysis. The results drawn from a variety of different approaches to the analysis all point to the same conclusion.  The difference between lagging and not lagging the independent variable did not produce substantially different results.  As well, when looking at the general positive or negative trend in analyzing growth during the reversion year, the reversion versus the previous year, and the reversion versus the average of the previous five years, the results were basically identical.  While little difference between these approaches can be seen, that should also serve to increase the confidence in our findings. Despite the pervasive notion of a connection between economic growth and democratization, when we specifically examine the relationship in light of all of the existing cases of democratic reversion during the third wave, evidence of such a relationship fails to materialize.  93  Inflation  Next we turn to an examination of the relationship between inflation and democratic reversion (see Table 7).  It is worth first noting there is more missing data for this variable.31  Again, at the most basic level, the mean inflation rate at R1 is 152.5 while during the reversal year it is 97.7.  The heavy influence of a few cases (particularly Azerbaijan and Belarus) can be seen in the median scores, which at R1 is 10.6 and at R is 12.6.  As with the GDP data, one interesting result that becomes immediately apparent is the mean score for the entire sample declines 54.8 points from R1 to R while the median score increases 2 points from R1 to R.  Looking across the full six years of data we see the mean rapidly increases from R5 to R2 and then rapidly declines from that point through R.  Except for the slight uptick from R1 to R, the median moves in a similar fashion to the mean, but at a much lower magnitude.  The discrepancies in this most basic examination of the data do a good job of illustrating the point that we should be cautious when accepting results based on larger, cross-national, time series efforts to examine the data. These approaches simply assume away the importance of context, arguing that a one year lag captures everything that is important in the case as well as the data.  The very simple problems just illustrated point to the problematic nature of such an assumption.  31  No data exists for any of the six years for Comoros 1977, Comoros 1999 or Guyana. Azerbaijan is missing four years of data while Armenia and Cambodia are missing two. The Republic of Congo is missing data for the reversion year only.  So this section will analyze the results, in most iterations, for twenty five or twenty six cases. 94  To begin a more in-depth analysis of the relationship between inflation and democratic reversion, we first look at the state of inflation during only the year in question (the reversion year or the year prior to the reversion).  During the reversion year, 16 cases had inflation rates greater than 10 and one had negative inflation.  The remaining 9 cases had inflation rates under 10.  Of the 16 cases of higher inflation, half experienced inflation rates above 20.  Turing to the year prior to the reversion (R1), the results are quite similar: 17 cases have higher rates and 10 have lower rates.  Of the higher cases, 10 of 17 are above 20.  These results lend some credence to the possibility that inflation rates may play a role in democratic reversion but from the perspective of a MDS approach, the wide variance across this explanatory variable appears to cast doubt on its ability to explain change in the dependent variable. Turning now to the evolution of inflation over time, we again assert that people do not consider the influence of issues such as inflation in isolation; rather they are capable of employing a perspective considering change over time.  To begin this analysis, we compare a two year time horizon by comparing inflation rates at R1 and R.  Across the sample, the mean decline in inflation from R1 to R was 60.2.  However, the median difference was only 1.8.  Inflation got worse (it increased or turned negative- deflation) in 12 cases while it got better in 14 cases.  Of the 12 cases where inflation worsened, in only 5 cases was the increase greater than 10 points.  On the other hand, there are 3 cases where inflation declined by at least 10 points.  Using R1 as the base year instead of R does not really change the results.  Inflation got worse in 10 cases and remained the same or got better in 16 cases.  The 10 cases where inflation was worse in R1 are evenly split; in 5 cases inflation grew by at least 10 points.  On the other hand, in 8 of the 15 cases of 95  improvement, saw inflation decline by at least 10 points.  When we expand the time horizon in the analysis to compare the Reversion year to the average of the previous five years, the results were basically identical.  Inflation improved in 15 cases, it was worse in 10 and turned deflationary in 1.  Again, there is little difference in the findings when we start with R1 and compare it to R2 in either of these analytical approaches.  The bottom line is again it appears that the analysis fails to confirm the relation between inflation and democratic reversion. Discussion  If we rely on the benefits of the MDS design, the case study data analysis would leave us unable to reject the null hypothesis that economic issues and democratic reversion are related.  As discussed in the data and methods section of the Chapter 2, the basic idea behind the MDS approach is that we analyze a set of cases that are selected by holding the dependent variable constant and then looking for similarities.  Here we examined thirty cases of democratic reversion.  Based on the data employed to select these cases, this represents the universe of democratic reversion cases between 1975 and 2004.  In this section of the chapter, we then analyzed the potential influence of economic issues, in particular the level of economic development (GDP per capita), economic growth (GDP per capita growth) and inflation.  What becomes clear in examining the thirty cases across six years for each of the three independent variables is that no clear effect can be identified.  In the case of all three measures, employing a variety of techniques to examine the data, the effects are variable.  There are cases where the independent variable moves in the expected direction.  But for all three variables there 96  are also a substantial number of cases where the variables move in the opposite direction necessary to support a relation between economic issues and democratic reversion. These results should give us pause when considering the macroeconomic analyses in the extent literature.  While such analyses are important, if they are unable to stand up to a study based on the universe of democratic reversion cases during the third wave then it seems reasonable to be highly skeptical of such conclusions. Case Level Analysis Generally, the case study narratives bear out the conclusions drawn by examining the aggregate case study data.  What does become apparent however is that even in cases where the economic indicators are moving the expected direction (if we assume a relationship exists), the particular circumstances of such cases often caution against drawing the conclusion that such relationships are essential to explaining the reversion. In addition to this general conclusion, the case narratives point to several additional important lessons.  First, legislative gridlock plays an important role.  Not only does gridlock hinder a regime’s response to an economic crisis, it has a related effect on the perception of democratic uncertainty.  If the legislature is, for whatever reasons, unable to respond to crisis due to gridlock, actors seeking to add to their support and translate that support into representation should have less confidence that such representation will translate into their preferred policies (and outcomes).  As such, their tolerance for lower economic payoffs should also diminish.  Second, quite a few cases illustrate the conclusion that the data can sometimes be deceiving.  While there are certainly cases where the data mask what is actually happening in the country (for example, the pyramid lottery scheme in Albania), the bigger issue is with the assignment of responsibility for 97  performance, particularly as it relates to the common practice of lagging variables one year.  For example, the regime Haiti (1991) was not in power for a large part of 1990, so it difficult to understand why people would assign blame for economic performance in 1990.  While the data indicate a -2% economic growth rate which is certainly consistent with the importance of poor economic performance lagged a year, it would be appear more relevant to examine the +2% growth rate during the reversion year.  Even more realistically, we should consider the idea that economic growth may not have been the issue that was resonating with actors.  Along similar lines, we should consider the Nigerian case, which experienced a reversion late in the evening on December 31.  The point is that before we accept the notion that actors simply respond to economic stimuli from the previous year, an examination of the results of cross-national, time-series analysis in light of the actual cases of reversion is warranted. Albania:  The economic data in this case are fairly misleading.  The GDP growth was 10-11% in the four years up to and including the reversion year.  Inflation declined from 226% to 7% in the four years leading up to the reversion and increased to 12.7% in the reversion year.  These data indicate an economy that is substantially improving in the years leading to the reversion.  Such data were taken as an indication that President Berisha’s economic reforms including widespread privatization, lifting trade restrictions and reforming the Bank of Albania was a success.  However, it appears that much of the success pointed to by the economic data were in fact due to a nation-wide pyramid scheme supported by Berisha’s Democratic Party and promoted on state television.  At its height, the scheme was said to include someone from every family in the country and involve more than $2 billion in savings.  The year after the reversion, the scheme 98  collapsed and Berisha was forced to resign (Poggioli 1997, 47; Vickers 2008; Nicholson 1999; Sunley 1998). Armenia:  The Armenian economy was experiencing renewed growth at the time of the democratic reversion.  The Armenia war with neighboring Azerbaijan had disastrous consequences for the economy.  Following the end of the war, President Ter- Petrossian introduced a liberalization program that included steps to “liberalize prices, stabilize the currency, reduce the budget deficit, accelerate privitazation and abolish subsidies” (Keesings November 1994).  As a result of these actions, the country achieved economic stability and growth rates between 7.5% and 9.1% in the three years leading up to the reversion.  Inflation was brought down from over 5000% to just under 19% during this period.  There are indications that much of the wealth generated by this economic boom was concentrated in the hands of the President’s supporters.  However, it is difficult to make the case that economic problems contributed to the democratic reversion in Armenia (Bremmer and Welt 1997, 5; Herzig 2008; Specter 1997; Freedom House: Armenia 2006). Azerbaijan: While the economy played a role in the reversion, its importance appears to be secondary to President Elchibey’s failure to end the war with Armenia.  The country declared independence in August of 1991 and the first president was forced from office in March 1992.  Elchibey took office in June of 1992 and was ousted from power by a coup twelve months later.  Given these short time periods, there is a limited amount of data available.  What data are available shows a 24% decline in economic growth in the two years leading to the reversion and inflation running between 900 and 1100%. 99  However, the primary cause of these economic problems was the war with Armenia. While Elchibey came to power “promising to win the war quickly, instead, the country suffered a string of spectacular defeats and lost about 10% of its territory that year.” (Economist June 19, 1993, 1)  So while it is true that the president failed to reignite the economy, this seems more of a symptom of a failed war effort than a cause of the subsequent democratic reversion (Yorke and Fumagalli 2008; Cornell 2001; Freedom House: Azerbaijan 2002; OSCE Report Azerbaijan 2006; Kamrava 2001). Belarus:  The economy played a role in this country’s democratic reversal but seemingly the opposite role one would expect.  The common assumption in the literature is that when the economy declines, the incumbent, and hence the regime, should be in an increasingly perilous position. In Belarus, the incumbent president used the declining economy as an excuse to seize power.  The data indicate declining economic growth in the four years leading to the democratic reversion.  While inflation rates were extremely high during the reversion year (709%), they were substantially better than the year before (2221%).  Prior to the initial democratic election following independence, Alyaksandr Lukashenko was appointed to chair a parliamentary committee investigating corruption. (Potock 2002, 1)  His efforts were directly responsible for forcing the transitional leader from office, propelling him to the Presidency in July 1994. (Keesings January 1994) Following his election, Lukashenko continued to push the message that the country’s corrupt elite were responsible for the severe economic decline and collapsing living standards in the country. (Vitali 2005, 2; Sannikov 2005, 1)  Poor economic performance thus served as the foundational argument for Lukashenko’s efforts to undermine 100  democracy and ultimately seize control of the state (Ryder 2008; Vera 1997; Potock 2002; Vitali 2005; Sannikov 2005). Burkina Faso:  The economy played, at best, an indirect role in the democratic reversion.  The data indicate an economic situation that would be considered pretty good for the region at that time.  Economic growth had increased the two years prior to the reversion year (2.5%, 1.6%) and slightly declined during the reversion year (-1.3%). Inflation improved from 30% four years prior to the reversion down to 12.2% during the reversion year.  The data however do not paint a complete picture of the economic situation.  Economic decline due to drought was a major contributing factor to the democratic reforms during 1977.  A teacher’s strike in 1980 caused considerable unrest and major divisions in the legislature.  In an attempt to reduce criticism of his handling of this situation, President Lamizana imposed restrictions on the media.  These restrictions triggered a failed attempt to pass a no confidence motion in the legislature.  This was shortly followed by a successful coup, bringing an end to this period of democracy in the country (Englebert and Murison 2008; Freedom House 1983; Keesing’s February 1980; Keesing’s June 1981). Burundi:  The economy played no important role in the democratic reversion. This was a case related to ethnic conflict and ultimately genocide.  The data indicate an economic situation that was fairly flat.  In the five years prior to the reversion year, economic growth fluctuated between -1.4% and +2.7%.  It was lower during the reversion year (-7.8%) but the decline is a reflection of the economic effects felt after the reversion due to the genocide that took place.  The story is largely the same with 101  inflation.  So, prior to the reversion, the economy was actually in pretty decent shape.  It was political events related to ethnic strife that triggered this reversion not economic situation (Mthembu-Salter 2008; Lemarchand 1989; Boyer 1992; Reyntjens 1993; Watson 1993). Cambodia:  Economic problems did not play a role in the democratic reversion. In Cambodia, the economic situation was actually improving right up to the democratic reversion.  The economy was growing at about 3% per year for the three years leading to the reversion.  Likewise, inflation was fluctuating a bit but was relatively low (3.2% during the reversion year).  The case information indicates this was really a situation where a number of actors, each controlling a substantial military force, were struggling for control of the state.  Once Hun Sen’s party (CPP) got the upper hand in the legislature, by bribing four members of the opposition (FUNCINPEC) to cross the floor, he sent in his forces to attack the FUNCINPEC’s headquarters in Phnom Penh and took control of the country in a coup.  If anything, this case runs counter to the expectation of the development and democratization literature (Peou and Summers 2008; Chad 1996; Lapidus 1998; Maley and Sanderson 1998; Peang-Meth 1997). Central African Republic: Economic problems played an important role in the democratic reversion.  The economy was basically stagnant in the years leading up to the reversion.  The year before the reversion, growth declined 2.1% and during the reversion years it declined 8.8%.  On the other hand, during the four years leading to the reversion, inflation was fairly low (2-4%) and steady.  What the data do not reveal is that in the mid-1990s, France stopped proving funds to pay the public sector in the CAF.  In late 102  2002 and early 2003, the civil service went on strike.  The protests resulting from the strikes turned violent and this violence triggered a failed coup attempt in May 2001.  In response to this attempt, President Patasse unleashed a wave of violence against the opposition to solidify his position.  In response to this violence, the ongoing civil service strike and declining public support for the president, former General Francois Bozize seized control of the capital in March 2003 (Englebert et al. 2008; New African 2001; Freedom in the World: Central African Republic 2006; World News Digest July 7, 2001; Keesing’s “Government Changes” December 2000; Economist December 8, 2003). Comoros (1976):  There is no evidence that the economy played a role in the democratic reversion.  The data on Comoros at this point are very limited.  There is some indication that the Comoros economy was in fact experiencing positive economic growth leading up to the reversion.  This case, however, appears to hinge on a politically premature call for independence that was not supported by the former colonial ruler, France (Recent History: The Comoros 2008; Bakar 1988; Merrill 1993; The Globe and Mail May 15, 1978). Comoros (1999): There is some evidence that the economy played a role in the democratic reversion.  The data on the Comoros are limited but what is available indicates a stagnant economy.  Growth rates decline a bit the year prior to the reversion (- 0.5%) and increased slightly during the reversion (0.7%).  There are indications that the ability of President Mohammed Taki Abdoulkarim to govern effectively was limited by failure to pay public servant salaries.  Members of the army may have been among the groups receiving irregular pay.  In November 1998, following the death of President 103  Taki, Tadjidine Ben Said Massoundi, in accordance with the constitution was designated as the acting president.  In April 1999, Army Colonel Assoumani Azali seized control of the country in a coup justified on the grounds of President Massoundi’s inept handing of the country’s security situation.  As such, the economy seems, at best, to be a contributing factor to the democratic reversion (Recent History: The Comoros 2008; Bakar 1988; Merrill 1993; The Globe and Mail May 15, 1978; Bratton 2007; National Post 1999; Freedom in the World: Comoros 2002). Republic of Congo: The economy does not appear to have played a role in the democratic reversion.  The available data indicate that was slightly positive leading up to the reversion and turned down (-3.9%) during the reversion year.  While inflation data are not available for the reversion year, inflation during the previous two years (9-10%) was not that high compared to neighboring countries.  Democratic elections began the transition in the Republic of Congo in 1992.  The elections were followed by a period of violent clashes between militias controlled by the ruling party and the opposition. Ultimately, a pact requiring the militias to disarm and integrate into the military ended the conflict and ushered in a relatively peaceful period led by President Pascal Lissouba. In 1997, Sassou Nguesso, the pre-transition leader, returned to the country to contest the upcoming election.  When government forces surrounded his house in an attempt to arrest two of his associates, fighting broke out.  Nguesso, who with the support of Angola and France had built a private army in northern Congo, forcibly retook the presidency in October 1997.  While economic growth issues did not trigger these events, there are good indications that a desire to gain access to oil revenues was an important motivation for Nguesso as well as his benefactors (Englebert and Murison 2008; Bazenguissa-Ganga 104  1998; Clark 1997; Clark 2002; Eaton 2007; Roberts 1998; StarPhoenix October 1997; Economist November 1997). Fiji: The data and research indicate the economy did not play a role in the democratic reversion.  The data indicate the economy was fluctuating in the years leading up to the coup.  During the five years leading up to the reversion, growth ranged from - 6% to +6% to -6% to +7% to -7%.  Likewise, inflation ranged between 4% and 7%. Neither set of data meet the expectation of the economic development argument.  The economic growth figures point out the folly of simply lagging economic variables one year and leaving out any real context.  The trigger seems to have occurred in April 1987, when the Fijian Indians elected a majority coalition to the country’s parliament.  This was unacceptable to the Fijian military which was largely composed of Melanesians who were intent on preventing Indian political dominance.  In addition, Prime Minister Timoci Bavadra’s pronouncement that the country would seek to move out from under the US defense umbrella undercut any opposition the US may have had to the coup (Recent History: Fiji 2008; Bedford 1987; Digirolamo 1987; Keesings December 1987; Economist December 12, 1987). The Gambia: The economy played at least an indirect role in the democratic reversion.  The data indicate that economic growth was stagnant and slightly negative (between -0.2% and -0.6%) in the four years leading up to the reversion and then it declined to -3.2% during the reversion year.  Inflation, on the other hand, declined from 12% to 1.7% in the five years leading to the reversion.  The case narrative indicates the important, yet indirect, role of the economy.  After almost thirty years of democratic rule, 105  in July 1994, members of the military led riots that broke out in the capital, Banjul.  After several days, during which the military gained control of critical points in the capital, President Kairaba Jawara fled to a US warship and was later granted amnesty in Senegal. Prior to the outbreak of the riots, there are indications of discontent within the military over the issue of back pay owed by the government (Wiseman and Murison 2008; US State Department Background Note 2007; Keesing’s Failed Coup Attempt January 1995; Keesing’s Military Coup July 1994). Ghana: The economy played a critical role in the democratic reversion.  In the three years leading up to the reversion, economic growth of -4.4%, -2% and -6% illustrate a case where the economy was in serious difficulty.  This is confirmed by the inflation data which show the rate jumping from 54% to 50% to 116% during this time. The case information confirms the data.  The Ghanaian economy was heavily dependent on cocoa exports and foreign capital.  Following his election in June 1979, President Hilla Limann made substantial progress fighting corruption but was unable to overcome a severe economic decline driven by falling global cocoa prices.  These problems were exacerbated by the government’s efforts to impose price controls which eventually led to an overvaluation of the country’s currency hurting the country’s exports even more.  By September of 1981, the country was experiencing shortages of vital commodities.  As well, strikes and riots, led by unpaid civil servants, were rampant.  On December 31, 1981, Lt. Jerry Rawlings seized power in a coup (Synge and McCaskie 2008; Keesings May 1992; Petchenkine 1993; Freedom House: Ghana 1983; CIA World Factbook: Ghana 2008). 106  Guinea-Bissau: The economy played an important role in the democratic reversion.  The data point to an economy that was in serious trouble leading to the reversion.  In the three years leading to the reversion, the growth rate was -3%, -10% and -2.5%.  Inflation, which had been very low (3.3% and 3.3%), turned negative (-3.5%) during the reversion year.  The case information confirms this picture.  President Kumba Yala, who had come to power during the initial democratic election in late 1999, became increasingly intransigent following a string of coup attempts.  He refused to take action on the 2001 Constitution and repeatedly delayed elections following his dismissal of the legislature citing the government’s financial constraints.  Civil servants, who had not been paid in more than a year, led widespread strikes.  Following the military’s seizure of power, not only did the public express widespread support but the parliament sitting at the time passed a motion supporting the coup.  Elections were held a little more than a year later, returning democracy to the country (Peitte et al. 2008; Freedom House: Guinea-Bissau 2004; Gazette September 152003; Sonko 2002). Guyana: The economy played a critical role in the reversion, although in this case, the declining economy operated as a catalyst for the executive to seize dictatorial powers. Prime Minister Forbes Burnham won elections held in 1968 and 1973 despite serious allegations of fraudulent activity.  In 1974, Burnham attempted to implement a socialist agenda, nationalizing key sectors of the economy.  In response to the economic downturn that followed, the police cracked down on striking workers and suppressed dissent.  In 1978, a national referendum that Burnham seriously rigged provided him with almost unlimited power to govern (Smith 2008; Singh 1997; Felix 1998; Griffith 1991; Chandisingh 1983; Rodney 1981). 107  Haiti (1991): It is difficult to make the case that economic performance played a role in the reversion as the regime lasted less than one year.  This is an excellent example of the difficulty with simply lagging the independent variables one year and ignoring all context.  In Haiti, if we lag the variables one year, we are assuming that the reversion had something to do with performance that took place during the previous regime.  That said, economic growth during the reversion year was positive for the first time in at least the previous six years.  While inflation was 15.4% during the reversion year, that rate is lower than the previous year when it was 21.3%.  The economic data therefore run the opposite of the expectation of the development literature.  The case narrative does however indicate an indirect economic role.  Upon assuming office, President Jean- Bertrand Aristide laid off more than 10,000 civil servants in an attempt to reduce the size of the bureaucracy and strengthen the economy.  Needless to say, this undercut his support among this group, likely making it easier for the military to seize power (Aurthur 2008; Pierre-Pierre 1991; Danner 1987; Hull 1997; French 1992; Crosette 1992). Haiti (1999): Although not the key element, there were economic performance aspects to the democratic reversion.  The data indicate positive economic growth during the three years leading to the reversion.  Although small (1.2% in the reversion year), given Haiti’s history, a level of positive growth casts doubt on the development hypothesis.  Likewise, while inflation was 8.7% during the reversion year that represents a steady decline over at least the previous five years.  The case data point to several economic factors in the reversion.  The initial split between President Preval and former President Aristide was triggered by Preval’s privatization plan.  As well, the legislative paralysis following the 1997 Senate elections blocked Preval’s economic reform package 108  which resulted in the suspension of a significant amount of foreign aid.  When in response to the ongoing paralysis, legislators attempted to extend their terms beyond four years, Preval declared the law unconstitutional, in effect dissolving the legislature. Shortly thereafter Preval announced he would rule by decree (Aurthur 2008; Shultz 1997; Rohter 1996; Fatton 1999; Economist December 4, 1999; Lundahl and Silie 1998). Iran:  The economy did not play a role in the democratic reversion.  The data indicate just the opposite.  Economic growth was steadily positive during the years leading up to the reversion.  While inflation ranged between 11% and 16% during the years prior to the reversion, there were no inflationary shocks during this period.  The case narrative indicates the reversion was a product of a backlash by hard-line clerics in response to the reform movement’s electoral gains as well as President Bush’s rhetoric and military actions following 9/11 (Cronin 2008; Momayesi 2000; Takeyh 2003; Sanam 2007; Mason 2002; Rajaee 2004; Gheissari and Nasr 2004). Niger: The economy played a crucial role in the democratic reversion.  The data point to an economy with stagnant economic growth (0% during the reversion year).  On the other hand, inflation declined from 36% to 11% to 5% in the years leading to the reversion.  The case narrative points out that Niger, one of the poorest countries on the planet, was in the midst of an economic crisis when the 1995 elections resulted in a government in which President Mahamane Ousmane and Prime Minister Hama Amadou represented opposing political parties.  The resulting legislative gridlock left the government unable to deal with the economic crisis.  Following failed efforts to pass the country’s 1996 budget, a group of senior military advisors, led by Chief of Staff Colonel 109  Ibrahim Barre Mainassara seized control of the government and imprisoned both the President and the Prime Minister (Englebert and Murison 2008; Keesing’s: Reaction to the Coup January 1996; Keesings: Legislative Elections November 1996; Amnesty International 1996; CIA World Factbook: Niger 2008). Nigeria: The economy played a role in the democratic reversion.  The data point to a difficult economic situation.  Economic growth declined during each of the three years leading to the reversion, including a -7.8% growth rate during the reversion year. Inflation also increased from 7.7% to 23.2% heading into the reversion.  The case narrative confirms these problems.  President Alhaji Shehu Shagari assumed power following the transitional election in 1979.  During his first term in office, the country experienced regional power struggles, widespread rioting, class-based unrest and widespread corruption issues (particularly as regards the oil revenues).  In 1983, Shagari was reelected in elections widely considered fraudulent.  At the same time, a decline in global oil prices seriously impacted the government’s revenues.  The country experienced widespread rioting following the election and the military took power on December 31st of that year (Synge 2008; Freedom House: Nigeria 1985; Keesing’s: General Elections January 1984; Keesing’s: Overthrow of Civilian Government May 1984; Joseph 1984), Pakistan (1977): The economy played a tangential role in the democratic reversion.  The data point to an economy with flat, yet positive growth during the four years leading to the reversion.  Growth declined from 1.9% to 0.8% during the reversion year.  As for inflation, after three years of inflation ranging from 2% to 27%, in the two years leading up to the reversion it was 7% and 10%.  These figures indicate mixed 110  support for the economic development approach.  There were widespread allegations of vote rigging following victories by Prime Minister Bhutto and the PPP in the 1977 provincial and national election.  Following the elections, rioting and civil unrest broke out across the country.  In order to combat this unrest, the government cut spending, further adding to its unpopularity.  In July 1977, General Muhammad Zia ul-Haq ordered the arrest of Bhutto and the leading ministers in his cabinet bringing an end to the democratic regime (McPherson 2008; Keesing’s: Riots 1977; Keesing’s: Arrest of Bhutto 1978; Freedom House: Pakistan 1979; CIA World Factbook: Pakistan 2008). Pakistan (1999):  The economic situation played a contributory role in the democratic reversion. The data indicate an economy with flat growth and declining inflation.  During the three years leading to the reversion, Pakistan experienced economic growth rates of -1.4%, 0.1% and 1.2%.  While growth actually increased during the reversion year, it was nonetheless fairly marginal.  Inflation on the other hand declined during those same three years form 11.4% to 6.2% to 4.1%.  The case narrative points a growing economic crisis exacerbating an already unstable political situation leading to the reversion.  It appears however, that Prime Minister Sharif’s repeated attempts to curtail the military’s power were the triggers for coup, led by General Musharraf, which ultimately ended the regime (McPherson 2008; Keesing’s: May 2000; CIA World Factbook: Pakistan 2008; Shah 2002; Freedom House: Pakistan 2002; Islam 2001; Ameen 1999). Peru: The economic situation played an important role in the democratic reversion.  The data point to a country with declining economic growth rates and 111  extremely high inflation rates.  In the five years prior to the reversion, there was positive economic growth (0.2%) only in the year prior to the reversion.  Inflation had declined from 7481% to 409% to 73% leading up to the reversion but remained a serious problem. The case narrative points to deteriorating economic and security conditions as laying the foundation for the reversion.  President Alberto Fujimori came to power promising to end inflation, reduce unemployment and lower the public debt without drastic reforms or extensive privatization.  However Fujimori confronted economic conditions so severe that he implemented a shock therapy program that curbed inflation but hurt economic growth and was largely unpopular.  Growing impatient with the opposition-dominated legislature’s failure to pass his budget, Fujimori dissolved the legislature, suspended the constitution and placed the country under emergency rule (Markwick 2008; Cameron 1998; Cameron 1994; Cameron and Mauceri 1997; Cameron 1988; Friedman April 11, 1992; Friedman April 14, 1992;  Holmes February 25, 1993; Nash May 18, 1992; Nash April 23, 1992; Crosette April 7, 1992; Freedom House: Peru 1995; Keesings April 1992; The Economist April 11, 1992). Sierra Leone: The economic situation, while chaotic, was not directly related to the democratic reversion.  The economic growth data point to an economy that was in a shambles.  Despite this, the data indicate that growth was a respectable 4.2% the year before the reversion.  During the reversion year, the economy reverted to a -18.5% rate. This again points to the problems with simply lagging the independent variable one year and not considering any context.  The inflation rate declined from the mid 20% range to 15% during the reversion year.  Despite this grim economic picture, the case narrative points to the idea that the economy did not trigger the reversion but rather was a 112  byproduct of the civil war which was the trigger (Synge and Clapham 2008; Bell 2000; Keesing’s: Return to Civilian Rule March 1996; Keesing’s: Chaotic Aftermath June 1997; Freedom House: Sierra Leone 1999). Sudan: The economy contributed to but did not directly cause the democratic reversion.  The data indicate there was actually positive economic growth in three of the four years prior to the reversion.  The inflation rate was also relatively low, particularly as compared to the rest of the region at that time.  The structural adjustment programs undertaken in the early 1980s caused substantial economic difficulty for the Sudanese population.  One year after a 1985 coup, the military turned power over to a civilian government led by Sadiq al Mahdi.  He brought members of the southern opposition into the government and the country saw violence decrease, particularly in the south.  The integration of the opposition ultimately made governing impossible due to constantly shifting coalitions.  There were also allegations of embezzlement of government funds as well as hording of foreign currency.  Three years after coming to power, the Mahdi government ended when General Omar Bashir seized power (Synge and Clapham 2008; Morrison 2005; Keesing’s: Fighting South War January 1990; Freedom House: Sudan 1992). Thailand:  The economic situation in the country did not contribute to the democratic reversion.  The data indicate that the Thai economy actually performed quite well in the period leading up to the reversion.  Economic growth was stable and increased to 6.7% during the reversion year.  Inflation was reduced to 5.3% and 4.1% in the last two years.  The case narrative confirms the positive economic situation as well as the lack 113  of a connection between the economy and the reversion.  Instead, the reversion seems to be a product of the inability of Prime Minister Seni Pramoj to resolve the existing legislative deadlock or to effectively combat the northern communist insurgency.  A coup, led by a twenty four member junta, ended the regime in October 1976 (McVey and Jory 2008; Keesing’s: General Election July 1976; Keesing’s: Military Coup in Bangkok December 1976; Keesing’s: National Administrative Reform December 1976; Freedom House: Thailand 1978). Turkey: The economic situation contributed to the democratic reversion.  The data point to an economy in extreme distress.  Economic growth declined for three years prior to the reversion, including a 4.6% decline during the reversion year.  The inflation rate rose for five straight years, ending up at 110% during the reversion year.  While the case narrative confirms the poor state of the economy, it also indicates that the economy was in bad shape due to the political instability and rampant political violence the country was experiencing at the time.  So rather than cause the reversion, the economic situation was itself a product of the causes of the breakdown (Day and Hale 2008; Keesing’s: Developments Following the Coup May 1981; Keesing’s: Assumption of Power October 1980; Birand 1987; Amnesty International 1988; Karasapan 1989). Uganda: The economy was not directly related to the democratic reversion.  The data point to an economy in distress.  Economic growth (-3.6%, -6.6%) declined in both years leading to the reversion.  Inflation increased from 43% to 158% in the two years prior to the reversion.  The case narrative however again points to a situation where poor economic performance is a byproduct of the events that led to the democratic demise in 114  the country.  In the case of Uganda, the “Bush War” during President Obote’s second term in office, which resulted in more than one hundred thousand deaths, brought put the country under a substantial amount of international pressure.  The military ultimately broke with Obote and seized power (Rake and Jennings 2008; Freedom House: Uganda 1987; Keesing’s: Internal Security Situation April 1985; Keesing’s: Military Coup December 1985; Keesing’s: Overthrow January 1986). Zimbabwe: The economy was not related to the democratic reversion.  The economic data point to a country experiencing negative economic growth and moderate inflation prior to the democratic reversion.  Economic growth was negative in five of the six years prior to the reversion.  Inflation declined during this period but remained at 12.5% during the reversion year.  The case narrative points to an early case of executive seizure.  So President Mugabe was able to ignore the negative implications of a declining economy in seizing power.  The impetus for the reversion in this case appears to be a problematic security situation (Brown and Saunders 2008; Freedom House: Zimbabwe 1989; Keesing’s: Creation of Executive President January 1988; Keesing’s: Progress Toward Party Merger February 1987; Economist: Zimbabwe November 7, 1987; Economist: Zimbabwe December 5, 1987). Discussion  An important theme running through a good portion of the sample is the role that legislative gridlock plays on the economy and, often independently, on democratic reversion.  For example, in Burkina Faso a nation-wide teacher’s strike divided the legislature resulting in gridlock.  The inability to reach a settlement on how to pay the 115  teachers triggered a massive wave of protest and unrest which led the military to take power via a coup.  The data in this case show a slight downturn in economic development and improvement in inflation.  The narrative demonstrates that the economy was much more relevant than the data would lead us to believe.  In the case of Niger, the data paint a picture of improvement on both economic growth and inflation.  The narrative, on the other hand, points to a legislature gridlocked over ratification of the new constitution. One result of this gridlock was the inability to pass a budget to address the troubled economy.  In response to the governmental standstill, the military instigates a coup. Similar situations can be identified in Central African Republic, Comoros, Gambia, Ghana, Haiti, Peru and Sudan. Another important theme is the relevance of war.  Azerbaijan, Sierra Leone, Sudan and Uganda are all cases where the narrative indicates that the effects of war were the primary causes of democratic reversion.  There are indirect effects on the economy but the narratives indicate that economic factors were not crucial to the reversions.  In Sierra Leone, for example, the ongoing civil war had decimated the country.  A civilian government took power well after the effects of economic devastation were understood by all actors in the country.  Here it was the collapse of a ceasefire agreement that ultimately triggered the reversion.  Despite indications in the data of economic problems, the civilian government was not in power long enough to begin to address (or get blamed for) economic difficulties. In addition, it is important to recognize that even in cases where the economic data appear to paint a clear picture, looks can be deceiving.  For example, in Armenia, the 116  data indicate the economic situation is good and improving.  The case narrative however points to a problem of corruption at the highest levels of government leaving many groups feeling like they did not receive the benefits of the economic situation.  In Haiti (1991), the data point to improvement with inflation and the first year of positive economic growth in at least the previous five years.  The narrative on the other hand argues the economic situation was extremely poor.  Furthermore, in this case, the threat of firing a substantial portion of civil servants would not have shown up in the data but represents a real economic concern that at least partially explains the reversion.  In addition, the civilian government was in power for such a short period of time, it is difficult to establish any realistic connection between economic improvements and the democratic government. In a similar vein, the data for Albania paint a picture of a booming economy yet the narrative points out that such figures masked a tremendous economic problem revolving around a nation-wide lottery scheme that was more of a pyramid scam.  The collapse of this lottery does not manifest itself in the data until the year after the reversion yet the case narrative demonstrates it plays a key role in the events leading up to the reversion. Finally, it is important to remember that often the economy simply had nothing to do with the reversion.  In both Armenia and Thailand, the economy was in good shape yet the regime collapsed due to corruption issues.  In Cambodia and Iran, the economy was performing well at the time of the reversion.  On the other hand, in Burundi, the economy was in trouble but ethnic strife is what triggered the reversion.  Likewise, in Turkey, the economy was performing poorly yet it was political violence that led the military to intervene.  The bottom line in the case narrative analysis is that we have to be 117  very cautions when interpreting these course annual measures of economic performance. While there is a place for analysis of such data across time and space, it is essential to check conclusions based on such data with conditions on the ground. Conclusions In the end, the analysis in this chapter recognizes that economic issues do have a role to play in understanding democratic reversion.  However, it is important to not overstate the importance of this role.  The cross-national, time-series analysis examined at the beginning of the chapter points out the importance of actually measuring the dependent variable about which we are attempting to make claims.  Once we move from an analysis of regime type to reversion to democratic reversion, the analysis indicates the significant and important role economic development and economic growth play in democratic reversion.  Had we stopped there however, we would have ended up with an incomplete picture of reality.  In undertaking the analysis of thirty cases of democratic reversion, we are able to see that it is necessary to temper expectation as to the exact extent of the role of these issues.  Given the variety of approaches to analyzing the case study data and the lack of consistent finding across the cases, we are left unable to reject the null hypotheses.  When we turn to an exploration of the case study narratives, the difficulties in attempting to reject the null hypothesis become even more apparent.  There are cases where the data indicate the economy should be relevant yet an analysis of the events points fails to confirm this supposition.  There are cases where the data indicate the economy is not relevant yet the case analysis indicates that it is relevant.  In the end, it seems fair to conclude that economic factors have some role to play in explaining democratic reversion but we should be extremely cautious in overstating that role. 118   These findings are important for understanding the relevance of the main theoretical position advanced in the dissertation.  Research examining the interaction between economics and regime usually has a structural orientation.  As such, it assumes actors respond mainly to economic stimuli and do not explicitly address the elements more commonly associated with process/choice approaches.  The basic findings of this chapter are at odds with the position that the economic situation of a case predicts reversion.  Given this, the position taken by Mainwaring and Pérez-Liñán (2005, 30) that the limited performance of structural models in Latin America suggests that scholars were right to emphasize process and choice seems relavant.  While the findings in this chapter indicate a role for economic inputs into the decision-making process, given the ambiguity of the results, a consideration of the situation that moves beyond the one year time horizon that almost all cross-national, time-series research relies upon (see, for example, Przeworski and Limongi 1997, 169) seems warranted.  We believe that when issues such as economic performance are viewed through the lens of democratic uncertainty, it becomes much easier to understand why actors may tolerate poor economic performance for quite a few years or why they may be intolerant of slightly positive economic performance, as well as why they may remove support for the regime after a single year of crisis.  119  Table 3.0.1: Economic Development - Regime Type Rated 2-14  Source: World Bank, World Development Indicators, 1972-2003 Method: Logistic Regression, n = 6016 Table 3.0.2: Economic Development - Regime Type Rated Free, Partly Free, Not Free Source: World Bank, World Development Indicators, 1972-2003 Method: Logistic Regression, n = 6016    Significance Pseudo R2 GDP per capita 0.000 0.109 GDP per capita Growth 0.871 0.001 Consumer Price Index 0.000 0.026 Inflation 0.996 0.001 Infant Mortality 0.140 0.024 Literacy (adult) 0.000 0.167 Literacy (youth) 0.000 0.211  Significance Pseudo R2 GDP per capita 0.000 0.047 GDP per capita Growth 0.000 0.009 Consumer Price Index 0.000 0.020 Inflation 0.029 0.002 Infant Mortality 0.000 0.390 Literacy (adult) 0.000 0.206 Literacy (youth) 0.000 0.164 120  Table 3.0.3: Economic Development - Democratic Reversions versus All Remaining Country Years (both Democratic and Non-Democratic) Source: World Bank, World Development Indicators, 1972-2003 Method: Logistic Regression, n = 6016 Table 3.0.4: Economic Development - Democratic Reversions versus Democratic Country Years Source: World Bank, World Development Indicators, 1972-2003 Method: Logistic Regression, n = 342  Significance Pseudo R2 GDP per capita 0.005 0.011 GDP per capita Growth 0.009 0.010 Consumer Price Index 0.705 0.000 Inflation 0.317 0.001 Infant Mortality 0.260 0.046 Literacy (adult) 0.363 0.001 Literacy (youth) 0.585 0.001  Significance Pseudo R2 Odds Ratio GDP per capita 0.000 0.071 0.998 GDP per capita Growth 0.000 0.038 0.908 Consumer Price Index 0.036 0.008 1.041 Inflation 0.162 0.003 121  Table 3.0.5: GDP per capita (30 Reversion Cases, World Bank: World Development Indicators, 1972-1993)  GDP per capita Rev-5 Rev-4 Rev-3 Rev-2 Rev-1 Reversal R-v-1L Rv1L% 5Yr Av R-v-5Av Rv5Av% 1L-v-2L 1Lv2L% 4YrAv 1L-v-4Av 1Lv4Av% Albania 708.9 663.1 736.7 818 902.1 993.5 91.4 10% 765.8 227.7 30% 84.1 10% 731.7 170.4 23% Arm enia 708.8 419.9 391.9 423 461.4 496 34.6 7% 481.0 15.0 3% 38.4 9% 485.9 -24.5 -5% Azerbaijan .. .. 1250.7 1222.9 932.3 706.1 -226.2 -24% 1135.3 -429.2 -38% -290.6 -24% 1236.8 -304.5 -25% Belarus 1409.8 1392.2 1255.8 1157.8 1023.5 920 -103.5 -10% 1247.8 -327.8 -26% -134.3 -12% 1303.9 -280.4 -22% Burkina Faso 178.3 189.5 186.4 191.1 194.2 191.7 -2.5 -1% 187.9 3.8 2% 3.1 2% 186.3 7.9 4% Burundi 153.3 151.1 152.6 156.8 155.3 143.3 -12.0 -8% 153.8 -10.5 -7% -1.5 -1% 153.5 1.9 1% Cam bodia .. 205.3 217.7 225.6 231.6 239 7.4 3% 220.1 19.0 9% 6 3% 216.2 15.4 7% Central African Republic 246.6 250.8 252.2 252.2 246.8 225.2 -21.6 -9% 249.7 -24.5 -10% -5.4 -2% 250.5 -3.6 -1% Com oros 77 .. .. .. .. .. .. NA NA NA NA NA NA NA NA NA NA Com oros 99 380.8 386.3 373.3 380.2 375.5 378.3 2.8 1% 379.2 -0.9 0% -4.7 -1% 380.2 -4.7 -1% Congo, Rep. 1097.4 1052.4 963 978.7 987.6 949.4 -38.2 -4% 1015.8 -66.4 -7% 8.9 1% 1022.9 -35.3 -3% Fiji 1715.7 1606.9 1705.1 1601.3 1709.6 1590.2 -119.4 -7% 1667.7 -77.5 -5% 108.3 7% 1657.3 52.3 3% Gam bia, The 326.3 325.5 323.6 322.9 321.2 310.8 -10.4 -3% 323.9 -13.1 -4% -1.7 -1% 324.6 -3.4 -1% Ghana 232.3 233.6 249.1 238.1 233.3 218.4 -14.9 -6% 237.3 -18.9 -8% -4.8 -2% 238.3 -5.0 -2% Guinea-Bissau 143.9 151 157.8 153.5 138.4 135.1 -3.3 -2% 148.9 -13.8 -9% -15.1 -10% 151.6 -13.2 -9% Guyana 758.6 812.8 876.4 883.1 852.5 830.1 -22.4 -3% 836.7 -6.6 -1% -30.6 -3% 832.7 19.8 2%  Haiti 91 678.5 657.3 646.9 639.6 626.6 641.9 15.3 2% 649.8 -7.9 -1% -13 -2% 655.6 -29.0 -4% Haiti 99 477.9 452.8 464.9 470.7 474.1 480 5.9 1% 468.1 11.9 3% 3.4 1% 466.6 7.5 2% Iran, Islam ic Rep. 1535.5 1591 1627.3 1722.5 1821.4 1887.8 66.4 4% 1659.5 228.3 14% 98.9 6% 1619.1 202.3 12% Niger 176.7 160.1 157.4 158.6 157.5 157.5 0.0 0% 162.1 -4.6 -3% -1.1 -1% 163.2 -5.7 -3% Nigeria 438.3 454 459.5 388.2 376.9 347.6 -29.3 -8% 423.4 -75.8 -18% -11.3 -3% 435.0 -58.1 -13% Pakistan 77 268.9 278.9 279.7 282.3 287.7 289.9 2.2 1% 279.5 10.4 4% 5.4 2% 277.5 10.3 4% Pakistan 99 498.4 510.4 522.2 515 515.6 521.7 6.1 1% 512.3 9.4 2% 0.6 0% 511.5 4.1 1% Peru 2311.5 2064.7 1784.9 1659.1 1662.4 1624.7 -37.7 -2% 1896.5 -271.8 -14% 3.3 0% 1955.1 -292.7 -15% Sierra Leone 205.1 208.1 204.1 187.1 195.1 159 -36.1 -19% 199.9 -40.9 -20% 8 4% 201.1 -6.0 -3% Sudan 269.5 245.7 252.8 282.5 275.9 294.4 18.5 7% 265.3 29.1 11% -6.6 -2% 262.6 13.3 5% Thailand 540.5 548.9 589.7 600.6 615 656 41.0 7% 578.9 77.1 13% 14.4 2% 569.9 45.1 8% Turkey 1875.5 2026.8 2053.2 2042.7 1988.1 1896.7 -91.4 -5% 1997.3 -100.6 -5% -54.6 -3% 1999.6 -11.5 -1% Uganda .. .. 176.3 180.7 174.2 162.8 -11.4 -7% 177.1 -14.3 -8% -6.5 -4% 178.5 -4.3 -2% Zim babwe 640.3 625 589.2 606 595.7 580.8 -14.9 -3% 611.2 -30.4 -5% -10.3 -2% 615.1 -19.4 -3% M ean 691.4 654.2 651.7 646.2 639.0 621.7 -17.4 -3% -31.2 -3% -7.2 -1% -19.0 -1% M edian 488.2 452.8 459.5 423.0 461.4 480.0 -10.4 -2% -10.5 -4% -1.5 -1% -4.3 -1% 122  Table 3.0.6: GDP per capita Growth (30 Reversion Cases, World Bank: World Development Indicators, 1972-1993)  GDP per capita Growth Rev-5 Rev-4 Rev-3 Rev-2 Rev-1 Reversal R-1L R-1L % 5YrAv R-v-5Av Rv5Av% 1L-2L 1L-2L % 4YrAv 1L-v-4Av 1L-v-4Av % Albania -27.5 -6.4 11.1 11 10.3 10.1 -0.2 -2% -0.3 10.4 3467% -0.7 -6% -3.0 13.3 -449% Armenia -10.9 -40.8 -6.7 7.9 9.1 7.5 -1.6 -18% -8.3 15.8 191% 1.2 15% -12.6 21.7 -172% Azerbaijan .. .. .. -2.2 -23.8 -24.3 -0.5 2% -13.0 -11.3 -87% -21.6 982% -2.2 -21.6 982% Belarus .. -1.2 -9.8 -7.8 -11.6 -10.1 1.5 -13% -7.6 -2.5 -33% -3.8 49% -6.3 -5.3 85% Burkina Faso 0.8 6.3 -1.6 2.5 1.6 -1.3 -2.9 -181% 1.9 -3.2 168% -0.9 -36% 2.0 -0.4 -20% Burundi 1.9 -1.4 1 2.7 -0.9 -7.8 -6.9 767% 0.7 -8.5 1282% -3.6 -133% 1.1 -2.0 -186% Cambodia .. .. 6 3.6 2.7 3.2 0.5 19% 4.1 -0.9 22% -0.9 -25% 4.8 -2.1 -44% Central African Republic 2.6 1.7 0.6 0 -2.1 -8.8 -6.7 319% 0.6 -9.4 1671% -2.1 1.2 -3.3 -271% Comoros 77 .. .. .. .. .. .. NA NA NA NA NA NA NA NA NA NA Comoros 99 -7.3 1.4 -3.4 1.9 -1.2 0.7 1.9 -158% -1.7 2.4 141% -3.1 -163% -1.9 0.7 -35% Congo, Rep. -0.6 -4.1 -8.5 1.6 0.9 -3.9 -4.8 -533% -2.1 -1.8 -82% -0.7 -44% -2.9 3.8 -131% Fiji -8.4 -6.3 6.1 -6.1 6.8 -7 -13.8 -203% -1.6 -5.4 -343% 12.9 -211% -3.7 10.5 -285% Gambia, The 1.9 -0.3 -0.6 -0.2 -0.5 -3.2 -2.7 540% 0.1 -3.3 5433% -0.3 150% 0.2 -0.7 -350% Ghana -5.4 0.5 6.7 -4.4 -2 -6.4 -4.4 220% -0.9 -5.5 -596% 2.4 -55% -0.7 -1.4 208% Guinea-Bissau -30 4.9 4.5 -2.7 -9.8 -2.4 7.4 -76% -6.6 4.2 64% -7.1 263% -5.8 -4.0 68% Guyana 0.7 7.2 7.8 0.8 -3.5 -2.6 0.9 -26% 2.6 -5.2 200% -4.3 -538% 4.1 -7.6 -185%  Haiti 91 -2.5 -3.1 -1.6 -1.1 -2 2.5 4.5 -225% -2.1 4.6 221% -0.9 82% -2.1 0.1 -4% Haiti 99 -9.5 -5.3 2.7 1.2 0.7 1.2 0.5 71% -2.0 3.2 159% -0.5 -42% -2.7 3.4 -126% Iran, Islamic Rep. 0.5 3.6 2.3 5.9 5.7 3.6 -2.1 -37% 3.6 0.0 0% -0.2 -3% 3.1 2.6 85% Niger -0.6 -9.4 -1.7 0.7 -0.7 0 0.7 -100% -2.3 2.3 100% -1.4 -200% -2.8 2.1 -75% Nigeria -8.6 3.6 1.2 -15.5 -2.9 -7.8 -4.9 169% -4.4 -3.4 -76% 12.6 -81% -4.8 1.9 -40% Pakistan 77 -2.3 3.7 0.3 1 1.9 0.8 -1.1 -58% 0.9 -0.1 13% 0.9 90% 0.7 1.2 181% Pakistan 99 1.2 2.4 2.3 -1.4 0.1 1.2 1.1 1100% 0.9 0.3 -30% 1.5 -107% 1.1 -1.0 -91% Peru 5.6 -10.7 -13.6 -7.1 0.2 -2.3 -2.5 -1250% -5.1 2.8 55% 7.3 -103% -6.5 6.7 -103% Sierra Leone -19.2 1.5 -1.9 -8.3 4.2 -18.5 -22.7 -540% -4.7 -13.8 -290% 12.5 -151% -7.0 11.2 -160% Sudan -7.9 -8.8 2.9 11.8 -2.3 6.7 9 -391% -0.9 7.6 879% -14.1 -119% -0.5 -1.8 360% Thailand 2.1 1.5 7.4 1.9 2.4 6.7 4.3 179% 3.1 3.6 -119% 0.5 26% 3.2 -0.8 -26% Turkey 4.7 8.1 1.3 -0.5 -2.7 -4.6 -1.9 70% 2.2 -6.8 311% -2.2 440% 3.4 -6.1 -179% Uganda .. .. .. 2.5 -3.6 -6.6 -3 83% -0.6 -6.1 -1100% -6.1 -244% 2.5 -6.1 -244% Zimbabwe -1.3 -2.4 -5.7 2.9 -1.7 -2.5 -0.8 47% -1.6 -0.9 -52% -4.6 -159% -1.6 -0.1 5% Mean -4.8 -2.1 0.3 0.1 -0.9 -2.6 -1.8 -0.1 -1.6 -1.1 4.0 -0.9 -0.1 -1.4 0.5 -0.4 Median -1.3 0.1 1.0 0.8 -0.7 -2.4 -1.1 -0.1 -0.9 -0.9 0.6 -0.9 -0.4 -1.6 -0.4 -0.7 123  Table 3.0.7: Inflation (30 Reversion Cases, World Bank: World Development Indicators, 1972-1993)  Inflation, Consumer Prices Rev-5 Rev-4 Rev-3 Rev-2 Rev-1 Reversal R-1L R-1L % 5YrAv R-v-5Av Rv5Av% 1L-2L 1L-2L % 4YrAv 1L-v-4Av 1Lv4Av % Albania .. 226 85 22.6 7.8 12.7 4.9 63% 85.35 -72.7 -85% -14.8 -65% 111.2 -103.4 -93% Armenia .. .. .. 5244.2 176 18.7 -157.3 -89% 2710.10 -2691.4 -99% -5068.2 -97% 5244.2 -5068.2 -97% Azerbaijan .. .. .. .. 912.3 1129 216.7 24% 912.30 216.7 24% Belarus .. .. .. 1190.2 2221 709.3 -1511.7 -68% 1705.60 -996.3 -58% 1030.8 87% 1190.2 1030.8 87% Burkina Faso 18.8 -8.4 30 8.3 15 12.2 -2.8 -19% 12.74 -0.5 -4% 6.7 81% 12.2 2.8 23% Burundi 4.5 11.7 7 9 1.8 9.7 7.9 439% 6.80 2.9 43% -7.2 -80% 8.1 -6.3 -78% Cambodia .. .. .. 1.1 10.1 3.2 -6.9 -68% 5.60 -2.4 -43% 9.0 818% 1.1 9.0 818% Central African Republic -1.9 -1.4 3.2 3.8 2.3 4.1 1.8 78% 1.20 2.9 242% -1.5 -39% 0.9 1.4 149% Comoros 77 .. .. .. .. .. .. Comoros 99 .. .. .. .. .. .. Congo, Rep. -3.9 4.9 42.4 9.4 10 .. 12.56 0.6 6% 13.2 -3.2 -24% Fiji 7 6.7 5.3 4.4 1.8 5.7 3.9 217% 5.04 0.7 13% -2.6 -59% 5.9 -4.1 -69% Gambia, The 8.3 12.2 8.6 9.5 6.5 1.7 -4.8 -74% 9.02 -7.3 -81% -3.0 -32% 9.7 -3.2 -33% Ghana 56.1 116.5 73.1 54.4 50.1 116.5 66.4 133% 70.04 46.5 66% -4.3 -8% 75.0 -24.9 -33% Guinea-Bissau 6.5 -0.7 8.6 3.3 3.3 -3.5 -6.8 -206% 4.20 -7.7 -183% 0.0 0% 4.4 -1.1 -25% Guyana .. .. .. .. .. ..  Haiti 91 3.3 -11.4 4.1 6.9 21.3 15.4 -5.9 -28% 4.84 10.6 218% 14.4 209% 0.7 20.6 2838% Haiti 99 39.3 27.6 20.6 20.6 10.6 8.7 -1.9 -18% 23.74 -15.0 -63% -10.0 -49% 27.0 -16.4 -61% Iran, Islamic Rep. 20.1 14.5 11.3 14.3 16.5 14.8 -1.7 -10% 15.34 -0.5 -4% 2.2 15% 15.1 1.5 10% Niger -7.8 -4.5 -1.2 36 10.6 5.3 -5.3 -50% 6.62 -1.3 -20% -25.4 -71% 5.6 5.0 88% Nigeria 21.7 11.7 10 20.8 7.7 23.2 15.5 201% 14.38 8.8 61% -13.1 -63% 16.1 -8.4 -52% Pakistan 77 5.2 23.1 26.7 20.9 7.2 10.1 2.9 40% 16.62 -6.5 -39% -13.7 -66% 19.0 -11.8 -62% Pakistan 99 12.4 12.3 10.4 11.4 6.2 4.1 -2.1 -34% 10.54 -6.4 -61% -5.2 -46% 11.6 -5.4 -47% Peru 85.8 667 3398.7 7481.7 409.5 73.5 -336 -82% 2408.54 -2335.0 -97% -7072.2 -95% 2908.3 -2498.8 -86% Sierra Leone 65.5 22.2 24.2 26 23.1 14.9 -8.2 -35% 32.20 -17.3 -54% -2.9 -11% 34.5 -11.4 -33% Sudan 34.1 45.4 24.5 20.6 64.7 66.7 2 3% 37.86 28.8 76% 44.1 214% 31.2 33.6 108% Thailand 0.5 4.8 15.5 24.3 5.3 4.1 -1.2 -23% 10.08 -6.0 -59% -19.0 -78% 11.3 -6.0 -53% Turkey 19.2 17.4 27.1 45.3 58.7 110.2 51.5 88% 33.54 76.7 229% 13.4 30% 27.3 31.5 115% Uganda .. 108.7 49.3 24.1 42.7 157.7 115 269% 56.20 101.5 181% 18.6 77% 60.7 -18.0 -30% Zimbabwe 10.6 23.1 20.2 8.5 14.3 12.5 -1.8 -13% 15.34 -2.8 -19% 5.8 68% 15.6 -1.3 -8% Mean 19.3 57.8 169.8 550.8 152.5 97.7 -60.2 0.3 304.7 -218.2 0.1 -427.6 0.3 379.2 -256.0 1.3 Median 10.6 12.3 20.2 20.6 10.6 12.6 -1.8 -0.2 15.3 -1.9 -0.2 -2.8 -0.2 15.3 -3.6 -0.3 124  Chapter 4: Political Institutions and Democratic Reversion 125  Introduction  It is assumed that political institutions matter because they set the rules of the game and thus provide the structure for all subsequent interactions within a regime.  The objective of this chapter is to test the notion that the structure of the political system influences the likelihood of a reversion from democracy.  In the context of the dissertation writ large, this chapter serves as an illustration of the structural conditions that triggers actors to enter a reversion game.  This chapter examines three issues related to the structure of the political system that are brought together into a, broadly defined, political institutions approach.  The first issue addresses the structure of executive power in the context of the debate between presidential and parliamentary systems.32  A large portion of the literature over the last two decades points to the idea that presidential systems are more prone to democratic reversion.  For example, Linz (1990, 1994) argued that zero sum elections and dual legitimacy undermined the ability of a president and the legislature to compromise.  Others, for example Shugart and Carey (1992), defend presidential systems and argue that the difficulties often associated with presidential systems are related more to the fragmentation of the legislature or the veto power the executive or the legislature holds.  So, the chapter tests the hypothesis that presidential systems are more likely to experience democratic reversion than parliamentary systems.  32 See, for example, Linz 1990a; Linz 1990b; Linz and Valenzuela 1994; Shugart and Carey 1992; Mainwaring 1993; Stepan and Skach 1993; Mainwaring 1998; Power and Gasiorowski 1997; Hadenius 1994; Przeworski et al. 1996) 126  The second component of the political institutions model addresses electoral rules in the context of the debate between plurality and proportional representation systems.33 While some argue that proportional systems may increase the fragility of the party system, others argue they increase representation.  Likewise, plurality systems are viewed as creating zero sum contests and reducing representation (Diskin et al, 2005).  The chapter tests the hypothesis that proportional representation systems are more likely to experience democratic reversion than plurality systems. The third component of the political institutions model addresses legislative fractionalization.34  While some early scholars (Dahl, 1971) took the position that a two party system was essential to democratic stability, others have advocated more moderate levels of fragmentation (Dishkin et al, 2005).  Few however take the position that high fractionalization is a positive force for democratic stability.  So, the chapter tests the hypothesis that regimes with higher levels of legislative fractionalization are more likely to experience democratic reversion. To accomplish this undertaking, the analysis relies on three basic methodological approaches.  First, a cross-national, time-series analysis is undertaken.  This analysis examines the dependent variable from four different perspectives: regime type rated 2-14, regime type rated 1-3, democratic reversions as compared to all non-reversion country years, and democratic reversions as compared to only democratic country years.  Second,  33 See, for example, Blais 1991; Blais and Dion 1990; Hadenius 1994; Lardeyret 1991; Lijphart 1994. 34 See, for example, Duverger 1954; Lipset 1960; Dahl 1971; Lijphart 1977; Midlarsky 1984. 127  an aggregate analysis of thirty cases of democratic reversion during the third wave will be explored.  Finally, a case level analysis of each of the thirty cases is undertaken. The results of these analyses indicate that while the structure of the political system seems to matter, it becomes more difficult to discern the role of political systems as we focus the analysis specifically on democratic reversions.  This might be expected as many of the cases under examination had low system tenure scores and thus may simply not have had enough time for such institutions to take root in the country.  The analysis that follows demonstrates a significant and influential role for political institutions when the dependent variable is focused on differences in regime type.  As the focus of the dependent variable is narrowed down more specifically on democratic reversions, we see a decline in the importance of the role played by these institutions.  The findings regarding the debates surrounding presidential versus parliamentary systems as well as those surrounding plurality versus proportional electoral rules are mixed at best.  The findings regarding legislative fractionalization are more consistent; higher fractionalization seems problematic for democracy. Cross-National, Time-Series Analysis  This section of the chapter examines the relationship between political institutions and democratic reversion during the third wave of democratization.  So the time period under consideration is 1972 – 2003.  The data for the dependent variable are drawn from Freedom House’s Freedom in the World (Freedom House).  Data for the independent variables are drawn from the World Bank’s Database of Political Institutions (Keefer). 128  In order to examine the relationship between political institutions (broadly defined) and democratic reversion, we examine four groupings of variables.  First, we examine the structure of executive power by comparing presidential and parliamentary systems.  Second, the legislature is examined by looking at the fractionalization of political parties.  Third, the electoral rules are examined in light of the difference between plurality and proportional representation systems.  Finally, the stability of the regime is examined in light of the tenure of the current system or executive holding power. Dependent Variable 1  We begin with an examination of the regime type variable utilizing the Freedom House indicator that provides a 2 to 14 ranking for each country year (with 2 being the most democratic and 14 being the most undemocratic).  As is illustrated in Table 1, the political institutions model is significant (0.000) and it explains 41% of the variance in the dependent variable.  When we turn to the analysis of the coefficients, the importance of the individual measure becomes clear.  In terms of the chief executive, for every one unit increase in this measure, the Freedom House rankings are lowered 1.3 units (in the Freedom House rating, lower scores are more democratic).  So if we move the two units from a Presidential (0), to a Strong President Elected by the Legislature (1), to a Prime Minister (2), then the Freedom House score is lowered 2.6 units.  The relationship is significant (0.000).  In terms of electoral rules, a regime moving to a plurality voting system (a one unit change) reduced the Freedom House score 0.005.  However, this relationship was not found to be significant.  On the other hand, for systems that employ 129  a proportional representation system (a one unit change), the Freedom House score is reduced 1.1 units.  This relationship is significant (0.000). Next we consider the situation in the legislature.  While the Herfindahl Index and the Total Fractionalization should move in opposite directions because they both are measures of legislative fractionalization, the model indicates they both have a negative influence.  Both of these indexes range from 0 to 1, so when looking at their impact, it is important to remember that the changes observed in these scores are usually, at most, a few tenth of a percent.  The analysis of the Herfindahl Index variable indicates that a one unit change in the Herfindahl Index reduces the Freedom House score only 0.76 and also that this relationship is not significant.  On the other hand, while the Total Fractionalization variable is not significant at the 0.05 level, it only falls marginally outside of that range (0.08) so it may be worthwhile to consider its impact.  A one unit change in Fractionalization reduces the Freedom House score 3.7 units.  So this later result indicates that as the legislature becomes more fractionalized, the regimes become more democratic.  Finally, the variable that simply considers the tenure of the current regime is significant (0.000).  However, for each additional year the regime survives, the Freedom House score is only reduced 0.04 units.  The conclusion we can draw from these data is that regimes receive more democratic scores are parliamentary systems, that have survived longer and have proportional representation electoral rules. Dependent Variable 2  Unlike the analysis in the previous chapters, the Political Institutions data are not plagued by an extensive amount of missing data.  So while it was practically impossible 130  to analyze the disaggregated Freedom House scores given other structural influences, here that approach to the dependent variable was able to withstand an analysis of this more finely divided variable.  However, analyzing the second approach to the dependent variable, which divides regime type into the Freedom House categories of Free, Partly Free and Not Free, remains worthwhile.  As we are interested in comparing democratic to non-democratic regimes, we first compare Free and Partly Free regimes and then compare Free and Not Free regimes.  In this way we avoid consideration of the influence of the difference between Partly Free and Not Free regimes.  Thus, we can discern what differentiates these two categories of non-democracy from democracy and we can explore whether political institutions influence this more course division in regime (1-3) in a different manner than an approach relying on smaller gradations (2-14) in the dependent variable. Beginning with the chief executive variable, we can see it is significant in the case when we examine the difference between regimes rated as Free and regimes rated as Partly Free (0.000) as well as when we compare regimes rated as Free and regimes rated as Not Free (0.000).  In the case of the former, a one unit increase in the chief executive variable (moving from presidential to strong president elected by the legislature to parliamentary) decreases the odds of moving from Free to Partly Free 5.2%. Interestingly, the odds of moving from Free to Not Free are decreased only 2.5%.  The findings are similar to those found in the analysis of Dependent Variable 1 where we also found a positive relationship between parliamentary systems and democracy.  Turing to the issue of the influence of electoral rule, the results point to Plurality systems having a significant influence (0.011) when comparing Free and Partly Free however there is not a 131  significant relationship (0.075) when examining Free and Not Free regimes (although it is not far outside the range of significance).  For the former, as we move to a plurality system, the odds of a Partly Free regime decline 1.4%.  While not quite significant, the odds of a Not Free regime decline 1.0%.  Again these findings are consistent, at least in terms of the direction of the relationship, with the findings from the analysis of Dependent Variable 1.  Turing to the analysis of the influence of Proportional Representation systems, the data indicate there is a significant relation between this form of electoral system and regime type (0.000 for both analyses).  Here, a one unit change (or moving to a PR system) results in a 3.8% increase in the odds of a Partly Free regime and a 5.2% increase in the odds of a Not Free regime.  The direction of change in this portion of the analysis is the opposite of what was found in the results from Dependent Variable 1 (where both systems had a negative influence on non-democracy, although not a significant one in the case of Plurality).  The results here make much more sense in that the different systems should exert a different influence.  For Dependent Variable 2, moving to a plurality system decreases the odds of a non-democratic system, while moving to a PR system increases the odds of a non-democratic system.  Thus plurality systems are associated with more democratic regimes while PR systems are associated with more non-democratic regimes.  Given the difficulties of having a sufficiently large number of cases in each cell when the dependent variable runs from 2 to 14, the results of the current analysis may be the more reliable of the two approaches. Turning to the distribution of political party representation in the legislature, we encounter more interesting findings.  The Herfindahl Index and the Total Fractionalization for a legislature should be mirror images of each other.  As the 132  Herfindahl Index moves from 0 to 1 it indicates the system is moving from high fractionalization to low fractionalization.  On the other hand, as the Fractionalization Index moves from 0 to 1 it indicates the system is moving from low fractionalization to high fractionalization.  However, the results of the analysis of Dependent Variable 1 indicated that they both exerted a negative influence on non-democracy (although neither relationship was significant).  The analysis of Dependent Variable 2 indicates that higher legislative fractionalization is associated with non-democracy.  The examination of the Herfindahl index variable indicates a one unit increase reduces the chances of a Partly Free regime 38.7% and reduces the chances of a Not Free regime 70.3%.  Both of these relationships are significant (0.00).  When considering these chances, it is important to remember that the index runs from 0 to 1.  So it may be more instructive to interpret the results as indicating that for each 10% jump in the scale, the odds of a non-democratic regime are reduced 3.87% (PF) and 7.03% (NF).  This means that as the legislature becomes less fractionalized, regimes becomes more democratic.  The results from the Fractionalization Index are similar and are both significant (0.000).  For each 10% increase in the Fractionalization Index (as the legislature becomes more fragmented), the odds of a Partly Free regime increase 6.26% and the odds of a Not Free regime increase 23.76%.  So again, as the legislature becomes more fractionalized, the odds of a non- democratic regime increase. Next, we turn to the analysis of the tenure of the regime.  Here we find that for each additional year the regime is in place, the odds of a Partly Free regime increase 4.4% and the odds of a Not Free regime increase 3.6%.  Both of these relationships are significant at the 0.000 level.  While this appears a bit odd, it seems to indicate that 133  countries tend to experience non-democratic regimes for longer period s of time.  Finally, if we look at the overall political institutions model, we find that the model examining Free and Partly Free regimes is significant (0.000) and explains 27.7% of the variance. The model examining Free and Not Free regimes is also significant (0.000) and explains 55.1% of the variance.  The conclusions we draw from both of these analyses is that regimes are more likely to be democratic if they are parliamentary systems with plurality, not PR, electoral systems, when legislative fractionalization is lower and the system tenure is lower. Dependent Variable 3  While the findings in the previous sections are instructive in terms of distinguishing regime types, they tell us little about the reversion from democratic to non- democratic regimes.  Such analyses can tell us what variables differentiate the various gradations of regime type but they remain incapable of providing information as to what distinguishes regimes that experience reversions from those that do not.  To address this shortcoming, the analysis of Dependent Variable 3 examines country years that experience such reversions in comparison to the remaining universe of country years.  To accomplish this, all cases of a country moving from a Free to a Partly Free or Not Free rating are coded as 1 during the first year that score changes and the remaining country years are scored as 0. The results of the analysis of the chief executive variable are puzzling.  Whereas the analyses of Dependent Variables 1 and 2 lead us to believe that we are more likely to encounter democratic regimes in parliamentary rather than presidential systems, when we 134  focus specifically on reversions from democracy we find that the odds of reversion increased as we move to parliamentary systems.  The relationship is significant (0.001) and the odds ratio indicates that for each one unit increase (move from President to Strong President elected by legislature to Parliamentary) the odds of a democratic reversion increase 65.9%.  These results appear to indicate that while parliamentary systems are more likely to be democratic, presidential systems are less likely to experience democratic reversions.  These findings are difficult to explain given the findings from the analyses of the first two approaches to the dependent variable as well as the findings in the literature as referenced in the introduction to the chapter. The results from the analysis of electoral systems are much more similar to those of the prior regime type analyses.  Regimes that have plurality voting are 2.9% less likely to experience a democratic reversion.  This relationship falls just outside the bounds of significance at 0.059.  On the other hand, regimes that have proportional representation voting are 3.7% more likely to experience a democratic reversion.  This relationship is significant (0.023).  Where the analysis of Dependent Variable 1 and 2 differed in terms of the direction of the relationship, these results are more in line with those we observed in the analysis of Dependent Variable 2, both in terms of direction and odds ratio magnitude.  This is also the case when we turn to the analysis of legislative fractionalization.  Here we see that for each 10% increase in the Herfindahl Index (which indicates declining fractionalization) we observe a 4.97% decrease in the odds of a democratic reversion.  Likewise, for each 10% increase in the Fractionalization Index (indicating increased fractionalization), we find a 10.06% increase in the odds of a reversion.  Both of these relationships are significant (the former at 0.022 and the later at 135  0.020).  Again, where the analysis of Dependent Variable 1 and 2 differed in terms of the direction of the relationship, these results are more in line with those we observed in the analysis of Dependent Variable 2, both in terms of direction and odds ratio magnitude.  In turning to the analysis of the regime tenure variable, we find a different relation than observed in prior iterations of the dependent variable.  Here each additional year of regime tenure decreases the odds of a democratic reversion 3.7%.  The relationship is significant (0.000). These findings may best be understood as a cautionary tale for drawing conclusions based on regime type differences without also examining reversion events. While the results of the current analysis examining electoral rules and legislative fractionalization are similar to those found in Dependent Variable 2 (the 1-3 scale), we need to remember that these results are different from those based on Dependent Variable 1 (2-14 scale). In a similar vein, when we examine the influence of system tenure, we see that the results of the current analysis based on democratic reversions are similar to those based on Dependent Variable 1 but the opposite of those based on Dependent Variable 2. And finally, the findings of the chief executive variable (comparing presidential and parliamentary systems) also should make us pause to consider the implications of examining the dependent variable from different perspectives.  The analysis of Dependent Variables 1 and 2 both conclude that the odds of a system being democratic increase under parliamentary systems whereas the analysis of Dependent Variable 3 concludes that parliamentary systems increase the odds of a democratic reversion.  The data simply don’t provide us with a means to reconcile this seeming paradox.  Finally, it is important to examine the results of the overall political institution models.  While the 136  models for all three approaches to the dependent variable are significant, there are substantial differences in the amount of variance they explain.  The models for Dependent Variable 1 and 2 both explain a good amount of the variance of regime type (27.7% and 55.1% respectively), the model for Dependent Variable 3 explains only 1.7% of the variance of democratic reversion.  In light of these differences in the findings, it seems reasonable to argue that if we are interested in understanding democratic reversions, we should not only analyze the differences in regime type but should also consider the relevance of specifically examining democratic reversions. Dependent Variable 4  While the analysis relying of Dependent Variable 3 more tightly focuses our attention on reversions, there is an additional problem to consider.  This approach compares democratic reversion years to all non-reversion years.  The non-reversion years include regimes that are democratic as well as non-democratic.  And while is instructive to understand what makes reversion years unique; it also seems reasonable to question the inclusion of non-democratic regimes in the analysis.  After all, if we are interested in why reversions from democracy occur, it is not clear what we get out of comparing reversion years to non-reversion years in non-democratic regimes.  We should probably be more interested in comparing ongoing democratic years to reversion years and this is something that rarely occurs in the literature.  So, for Dependent Variable 4, any country year not experiencing a democratic reversion that is scored as Partly Free or Not Free is excluded from the analysis.  We are left comparing ongoing democratic country years and democratic reversion country years. 137  In examining the results of this analysis, it may be that the most interesting result is what was not found.  No significant relationship is found for the chief executive, electoral rules or legislative fractionalization variables.  Whereas the difference between presidential and parliamentary systems was significant for each of the three previous iterations of the dependent variable, when we focus specifically on democratic reversions, the difference between these systems is no longer significant.  Now it should be noted that at 0.068, it does not fall far outside the bounds of significance.  The results point to a 6.3% increase in the odds of a reversion for each one point move from presidential to strong president selected by the legislature to parliamentary systems.  The direction of this finding further confirms the difference observed between both regime type dependent variables (DV 1 and DV2) and the first approach to democratic reversion (DV3).  While neither of the electoral rules variables is significant, it is worth noting that the direction of the proportional representation variable flips again so that as with DV1, but opposed to DV 2 and DV3, the proportional representation variable moves in the direction of reducing the odds of a democratic reversion.  Neither of the legislative fractionalization variables is significant.  The one variable that remains significant (0.000) is system tenure.  Here for each additional year the system is in existence, the odds of a democratic reversion are decreased by 3.8%.  This offers further confirmation of the difference in findings between DV3 and DV2.  In the end though, it seems obvious.  The longer democratic systems endure, the less likely they are to experience a democratic reversion.  It seems the explanation for the differences in findings between dependent variables based on regime type as opposed to reversions may lie in the idea 138  that authoritarian regimes endure longer than democratic regimes but that the longer a democratic regime endures, the less likely it is to experience a reversion. Case Study Data Analysis  In this section of the chapter, the aggregate data from the thirty democratic reversion cases are examined.  In looking at the System variable (see Table 2), the case data indicate that during the reversion year, 19 countries were presidential systems and only 9 were parliamentary.  There were no cases of strong presidents that are elected by the legislature.  The two cases that are missing data, due to timing issues, would not alter these results as during the previous year, Iran was coded as a presidential system while Pakistan (1977) was coded as a parliamentary system.  Over the six years under consideration, only three cases experienced a change on this variable.  Cambodia changed from strong president elected by the legislature to a parliamentary system four years prior to reversion.  The Congo changed from strong president elected by the legislature to a presidential system five years prior to the reversion.  And Niger changed from strong president elected by the legislature to a presidential system three years prior to reversion. These findings are more in line with those found in the existing literature referenced in the introduction to this chapter. The data in Table 3 indicate that of the 16 cases that are coded on electoral systems, 12 are plurality systems.  The data in Table 4 indicate that of the 16 cases that are coded on this variable, 7 are proportional representation systems.  When we combine the data found the in the two tables, we find that there are 10 cases of plurality systems, 5 139  of proportional systems and 3 that combine both systems.  The extent of the missing data make it difficult to draw conclusions based on this portion of the analysis. The case data on legislative fractionalization are more consistent with the analyses in the previous section.  The data for the Herfindahl Index (Table 5) and the Fractionalization Index (Chapter 6) are, as expected, very close to mirror images of one and other.  The missing data cases are identical for both variables and the basic direction of the score is also the same.  As such, we will examine only the Fractionalization Index data.  Here, using 0.50 as a rough division point, we find that of the 20 cases with coded data, 14 are above this point (higher fractionalization) and 6 are below this point (lower fractionalization).  While noting that we are missing data for one third of the cases, these findings are consistent with the findings in the previous data analysis section of the chapter.  We should also point out that almost half of the cases (9) range between 0.45 and 0.61 so some of this conclusion clearly could be the result of the arbitrary nature of the 0.5 cutoff point. Finally, the case data for the tenure of the system variable are consistent with the findings in the previous section specific to democratic reversion.  Here, at the point of reversion, 19 cases had a system tenure ranging from one to five years, 4 cases had a system tenure ranging from six to ten years and 7 cases had a system tenure above ten years.  These data would appear to confirm the finding that younger democratic systems are more prone to reversion.  As previously discussed, it is likely that all young systems (both democratic and non-democratic) are vulnerable to some form of reversion (as may 140  be indicated by the DV2 results), so any conclusion based on these data must be taken with a grain of salt. It is difficult to know exactly what conclusions should be drawn from this portion of the analysis.  One problem is that we are examining static variables (with the exception of the tenure variable) to explain dynamic events.  Unlike the previous chapters where we could employ the aggregate case data to examine change over the five year period preceding the reversion, that is not possible in this chapter.  That said, we did find more presidential systems in the case set under consideration.  There were also more plurality voting systems, although differences in the number of cases are fairly small.  The findings on legislative fractionalization are consistent with the findings of the analyses found in previous sections of this chapter.  Systems with higher fractionalization appear to be more vulnerable to reversion.  As well, the system tenure measure among the thirty cases seems to confirm the prior findings that younger regimes are more prone to reversion. Case Level Analysis  We now briefly consider the political institutional situation in each of the thirty cases.  Again as there is no real dynamic element to this portion of the analysis, the case description information will be kept brief. Albania:  This country has a parliamentary system with a mix of both plurality and proportional representation electoral rules.  The legislative fractionalization was just under 0.5 and the system tenure was five years at the reversion year. 141  Armenia: This country has a presidential system with a mix of both plurality and proportional representation electoral rules.  The legislative fractionalization was 0.6, having recently declined from 0.7.  The system tenure was five years at the reversion year. Azerbaijan:  This country has a presidential system.  There are no data on electoral rules or party fragmentation.   The system tenure was one year at the reversion year. Belarus:  This country has a presidential system with plurality electoral rules.  The legislative fractionalization had declined from 0.71 to 0.6 the year prior to the reversion. The system tenure was one year at the reversion year. Burkina Faso:   This country has a presidential system.  There are no data on electoral rules.  The legislative fractionalization was 0.73 and the system tenure was two years at the reversion year. Burundi:  This country has a presidential system with plurality electoral rules. The data on legislative fractionalization are missing.  The system tenure was six years at the reversion year. Cambodia:  This country had a parliamentary system with proportional representation electoral rules.  The legislative fractionalization data were missing and the system tenure was four years at the reversion year. Central African Republic:  This country had a presidential system with plurality electoral rules.  The legislative fractionalization was 0.73 and the system tenure was ten years at the reversion year. 142  Comoros (1976):  This country had a presidential system.  The data are missing on electoral rules and legislative fractionalization.  The system tenure was one year at the reversion year. Comoros (1999):  This country had a presidential system with plurality electoral rules.  The legislative fractionalization was 0.30 (down from 0.50 four years earlier) and the system tenure was nine years at the reversion year. Congo-Brazzaville:  This country had a presidential system.  The data are missing for the electoral rules.  The legislative fractionalization was 0.79 and the system tenure was four years at the reversion year. Fiji:  This country had a parliamentary system with plurality electoral rules.  The legislative fractionalization was 0.54 and the system tenure was twelve years at the reversion year. The Gambia:  This country had a presidential system with plurality electoral rules. The legislative fractionalization was 0.50 and the system tenure was twenty years at the reversion year. Ghana:  This country had a presidential system with plurality electoral rules.  The legislative fractionalization was 0.64 and the system tenure was two years at the reversion year. Guinea-Bissau:  This country had a presidential system.  The data on electoral rules are missing.  The legislative fractionalization was 0.75 and the system tenure was three years at the reversion year. 143  Guyana:  This country had a parliamentary system with proportional representation electoral rules.  The legislative fractionalization was 0.45 and the system tenure was three years at the reversion year. Haiti (1991):  This country had a presidential system with plurality electoral rules. The legislative fractionalization was 0.83 and the system tenure was one years at the reversion year. Haiti (1999):  This country had a presidential system with plurality electoral rules. The legislative fractionalization was 0.33 (a substantial decline from 0.83 in 1995) and the system tenure was five years at the reversion year. Iran:  This country had a presidential system with plurality electoral rules.  The legislative fractionalization was 0.0 and the system tenure was fourteen years at the reversion year. Niger:  This country had a presidential system with mix of plurality and proportional representation electoral rules.  The legislative fractionalization was 0.76 and the system tenure was three years at the reversion year. Nigeria:  This country had a presidential system with proportional representation electoral rules.  The legislative fractionalization was 0.75 and the system tenure was four years at the reversion year. Pakistan (1977):  This country had a parliamentary system.  The data are missing electoral rules as well as legislative fractionalization.  The system tenure was thirty one years at the reversion year. 144  Pakistan (1999):  This country had a parliamentary system with plurality electoral rules.  The legislative fractionalization was 0.55 and the system tenure was eleven years at the reversion year. Peru:  This country had a presidential system with proportional representation electoral rules.  The legislative fractionalization was 0.78 and the system tenure was twelve years at the reversion year. Sierra Leone:  This country had a presidential system with proportional representation electoral rules.  The legislative fractionalization was 0.75 and the system tenure was one year at the reversion year. Sudan:  This country had a presidential system.  The data are missing electoral rules.  The legislative fractionalization was 0.76 and the system tenure was three years at the reversion year. Thailand:  This country had a parliamentary system.  The data are missing on electoral rules as well as legislative fractionalization.  The system tenure was one years at the reversion year. Turkey:  This country had a parliamentary system with proportional representation electoral rules.  The legislative fractionalization was 0.61 and the system tenure was twenty years at the reversion year. Uganda:  This country had a parliamentary system.  The data are missing on electoral rules.  The legislative fractionalization was 0.50 and the system tenure was five years at the reversion year. 145  Zimbabwe:  This country had a parliamentary system with plurality electoral rules.  The legislative fractionalization was 0.55 and the system tenure was seven years at the reversion year. Conclusions  The findings in this chapter indicate that while the structure of the political system is clearly relevant to democratic reversion, it is difficult to support the position that it is a determinant factor.  At a sort of meta-level, what we have called the political institutions model which includes variables addressing the presidential versus parliamentary debates, the electoral systems debates, the legislative fractionalization debates and an indication of system tenure, is demonstrated to be relevant but the question of how relevant is left open to interpretation.  While the models that test dependent variables examining differences in regime type are significant and explain a good deal of the variance, the models that test dependent variables by specifically examining democratic reversions are much less clear. The model that examines reversions in comparison with all remaining cases is significant but explains little of the variance in the dependent variable and the model that examines democratic reversion in comparison only to ongoing democracies is not significant and would also explain little of the variance.  Likewise both the aggregate as well as the individual level analyses of the thirty democratic reversion cases paint more of a mixed picture of the influence of structure of the political system. In terms of the four elements of the political system examined in this chapter, the findings on the first, the presidential versus parliamentary system debate, are mixed. When regime type is employed as the dependent variable (Dependent Variable 1 and 2), 146  the findings indicate parliamentary systems are more likely to be democratic.  When democratic reversion is employed as the dependent variable (Dependent Variable 3 and 4), the findings indicate that parliamentary systems are more likely to experience a reversion.  The case data indicate that of the thirty reversion cases, there are twice as many presidential as parliamentary systems. The findings regarding the plurality versus proportional representation systems debate are also mixed.  When regime type is employed as the dependent variable, one approach (DV1) indicates proportional representation systems are more likely to be democratic while the other approach (DV2) indicates that plurality systems are more likely to be democratic.  When democratic reversion is employed as the dependent variable, the findings indicate that proportional representation systems are more likely to experience a reversion (although neither variable is significant in the DV4 analysis).  The case data indicate that of the thirty reversion cases, ten are plurality while five are proportional. The findings regarding the importance of legislative fractionalization are more consistent.  Higher levels of legislative fractionalization appear to be problematic for democracy.  When considering regime type as indicated by Dependent Variable 1, the relationship is not significant and runs the opposite of the expected direction.  But the analysis of Dependent Variable 2 indicates that legislative fractionalization is significant and has a fairly substantial influence.  The analysis of the more specific democratic reversion variables points to the significant and substantial role increasing fractionalization has on the risk of reversion (although neither fractionalization measure 147  is significant in the DV4 model).  As well, the case analysis indicates that fourteen reversion cases had higher fractionalization while only six had lower levels. Finally, the evidence on system tenure is somewhat mixed.  It is significant in all of the models but it is difficult to ascertain its importance.  This is most likely a Yogi Berra type situation; what we really learned is that the longer the system is around, the longer the system is around.  The evidence based on regime type is mixed.  The reversion analyses both indicated that lower tenure is problematic.  As well, the case analysis demonstrates that most of the reversion cases were very young regimes. So while the structure of the political system seems to matter, it is not as clear that it plays a determinant role on the issue of democratic reversion.  While the models based on dependent variables examining regime type indicate a significant and influential role for political institutions, the models specifically focused on democratic reversions indicate a less substantial role for institutions.  The findings regarding the debates surrounding presidential versus parliamentary systems as well as those surrounding plurality versus proportional electoral rules are mixed, at best.  The findings regarding legislative fractionalization are more consistent; higher fractionalization seems problematic for democracy.  Finally, the analysis indicates that younger regimes appear more vulnerable to reversion.  148  Table 4.0.1: Political Institutions – Four Dependent Variables     DV1: Regime Type Rated 2-14 (n= 6016) DV2 (1v2): Regime Type - Free versus Partly Free (n= 6016) DV2 (1v3): Regime Type - Free versus Not Free (n= 6016) DV3: Democratic Reversions Democratic Reversions versus All Remaining Country Years (both Democratic and Non-Democratic) (n= 6016) DV4: Democratic Reversions versus Democratic Country Years (n= 3424) Source: World Bank, Database of Political Institutions, 1972-2003 DV 1 DV2 (1v2) DV2 (1v3) DV3 DV4 System System Signif 0.000 Signif 0.000 0.000 0.001 0.068 B -1.273 Exp (B) 0.948 0.975 1.659 1.063 Plurality Plurality Signif 0.969 Signif 0.011 0.075 0.059 0.840 B -0.005 Exp (B) 0.986 0.990 0.971 0.994 PR PR Signif 0.000 Signif 0.000 0.000 0.023 0.898 B -1.089 Exp (B) 1.038 1.052 1.037 0.996 HERFTOT HERFTOT Signif -0.076 Signif 0.000 0.000 0.022 0.387 B 0.655 Exp (B) 0.613 0.297 0.503 0.680 FRAC FRAC Signif 0.081 Signif 0.000 0.000 0.020 0.385 B -3.711 Exp (B) 1.626 3.376 2.006 1.474 TENSYS TENSYS Signif 0.000 Signif 0.000 0.000 0.000 0.000 B -0.044 Exp (B) 1.044 1.036 0.963 0.962 Model Model Sig 0.000 Sig 0.000 0.000 0.030 0.302 R Sq 0.410 R Sq 0.277 0.551 0.017 0.019 149  Table 4.0.2: Presidential versus Parliamentary System (30 Reversion Cases, World Bank: Database of Political Institutions, 1972-2003)    R-5 R-4 R-3 R-2 R-1 R Albania 1 2 2 2 2 2 Armenia 0 0 0 0 0 0 Azerbaijan 0 0 Belarus 0 0 0 0 Burkina Faso 0 0 0 0 0 0 Burundi 0 0 0 0 0 0 Cambodia 1 1 2 2 2 2 Central African Republic 0 0 0 0 0 0 Comoros 1976 0 Comoros 1999 0 0 0 0 0 0 Congo, Rep. 1 0 0 0 0 0 Fiji 2 2 2 2 2 2 Gambia, The 0 0 0 0 0 0 Ghana 0 0 0 0 0 0 Guinea-Bissau 0 0 0 0 0 0 Guyana 2 2 2 2 Haiti 1991 0 0 0 0 0 0 Haiti 1999 0 0 0 0 0 0 Iran, Islamic Rep. 0 0 0 0 0 0 Niger 1 1 1 0 0 0 Nigeria 0 0 0 0 0 0 Pakistan 1977 2 2 2 2 Pakistan 1999 2 2 2 2 2 2 Peru 0 0 0 0 0 0 Sierra Leone 0 0 0 0 0 0 Sudan 1 1 0 0 0 0 Thailand 2 2 Turkey 2 2 2 2 2 2 Uganda 0 2 2 2 2 2 Zimbabwe 2 2 2 2 2 2 0=Direct Presidential 1=Strong President Elected by the Legislature 2=Parliamentary 150  Table 4.0.3: Electoral Rules – Plurality? (30 Reversion Cases, World Bank: Database of Political Institutions, 1972-2003)    R-5 R-4 R-3 R-2 R-1 R Albania 1 1 1 1 1 Armenia 1 1 1 Azerbaijan Belarus 1 1 1 1 Burkina Faso Burundi 1 1 0 0 0 0 Cambodia 1 1 1 1 1 1 Central African Republic Comoros 1976 1 1 1 1 1 1 Comoros 1999 0 Congo, Rep. 1 1 1 1 1 1 Fiji 1 1 1 1 1 1 Gambia, The 1 1 Ghana 1 1 Guinea-Bissau 0 0 0 0 Guyana Haiti 1991 1 1 1 1 1 1 Haiti 1999 1 1 1 1 1 Iran, Islamic Rep. 1 1 1 Niger 0 0 0 0 Nigeria Pakistan 1977 1 1 1 1 1 1 Pakistan 1999 0 0 0 0 0 0 Peru 1 0 Sierra Leone Sudan Thailand Turkey Uganda 1 1 1 1 1 1 Zimbabwe 0=No, 1=Yes, 151  Table 4.0.4: Electoral Rules – Proportional Representation? (30 Reversion Cases, World Bank: Database of Political Institutions, 1972-2003)   R-5 R-4 R-3 R-2 R-1 R Albania 1 1 1 1 1 Armenia 1 1 Azerbaijan Belarus 0 0 0 0 Burkina Faso Burundi Cambodia 1 1 1 1 Central African Republic 0 0 0 0 0 0 Comoros 1976 Comoros 1999 0 0 0 0 0 0 Congo, Rep. 0 Fiji 0 0 0 0 0 0 Gambia, The 0 0 0 0 0 0 Ghana 0 0 Guinea-Bissau Guyana 1 1 1 1 Haiti 1991 0 Haiti 1999 0 0 0 0 0 0 Iran, Islamic Rep. Niger 1 1 1 Nigeria 1 1 1 1 Pakistan 1977 Pakistan 1999 0 0 0 0 0 0 Peru 1 1 1 1 1 1 Sierra Leone 0 1 Sudan Thailand Turkey Uganda Zimbabwe 0 0 0 0 0 0 0=No, 1=Yes, 152  Table 4.0.5: The Legislature – Herfindahl Index Total (30 Reversion Cases, World Bank: Database of Political Institutions, 1972-2003)     R-5 R-4 R-3 R-2 R-1 R        Albania 1.000 0.573 0.508 0.508 0.508 0.508 Armenia 1.000 0.293 0.293 0.293 0.293 0.402 Azerbaijan       Belarus  0.708 0.708 0.254  Burkina Faso  0.280 0.280 Burundi       Cambodia 1.000 1.000     Central African Republic 0.257 0.243 0.243 0.243 0.243 0.243 Comoros 1976       Comoros 1999 0.510 0.510 0.510 0.708 0.708 0.708 Congo, Rep. 1.000 1.000 0.217 0.217 0.217 0.217 Fiji 0.536 0.470 0.470 0.470 0.470 0.470 Gambia, The 0.761 0.761 0.761 0.761 0.515 0.515 Ghana  0.361 0.361 Guinea-Bissau 0.439 0.439 0.259 0.259 0.259 0.259 Guyana  0.559 0.559 0.559 0.559 Haiti 1991 1.000      Haiti 1999 0.182 0.182 0.674 0.674 0.674 0.674 Iran, Islamic Rep. 1.000 1.000 1.000 1.000 1.000  Niger 1.000 1.000 1.000 0.251 0.251 0.251 Nigeria NA 0.249 0.249 0.249 0.249  Pakistan 1977       Pakistan 1999 0.192 0.192 0.192 0.192 0.449 0.449 Peru 0.432 0.432 0.432 0.432 0.236 0.236 Sierra Leone 1.000  0.256 Sudan 1.000 1.000 1.000 0.243 0.243 0.243 Thailand       Turkey 0.301 0.301 0.301 0.409 0.405 0.392 Uganda 1.000 0.506 0.506 0.506 0.506 0.506  Zimbabwe  0.406 0.406 0.406 0.406 0.456 0.456   153  Table 4.0.6: The Legislature – Fractionalization Index Total (30 Reversion Cases, World Bank: Database of Political Institutions, 1972-2003)     R-5 R-4 R-3 R-2 R-1 R        Albania 0.000 0.429 0.495 0.495 0.495 0.495 Armenia 0.000 0.709 0.709 0.709 0.709 0.601 Azerbaijan       Belarus  0.292 0.292 0.749  Burkina Faso  0.732 0.732 Burundi       Cambodia 0.000 0.000     Central African Republic 0.753 0.729 0.729 0.729 0.729 0.729 Comoros 1976       Comoros 1999 0.502 0.502 0.502 0.299 0.299 0.299 Congo, Rep. 0.000 0.000 0.789 0.789 0.789 0.789 Fiji 0.473 0.540 0.540 0.540 0.540 0.540 Gambia, The 0.246 0.246 0.246 0.246 0.498 0.498 Ghana  0.643 0.643 Guinea-Bissau 0.567 0.567 0.749 0.749 0.749 0.749 Guyana  0.450 0.450 0.450 0.450 Haiti 1991 0.000      Haiti 1999 0.828 0.828 0.330 0.330 0.330 0.330 Iran, Islamic Rep. 0.000 0.000 0.000 0.000 0.000  Niger 0.000 0.000 0.000 0.758 0.758 0.758 Nigeria  0.753 0.753 0.753 0.753  Pakistan 1977       Pakistan 1999 0.812 0.812 0.812 0.812 0.554 0.554 Peru 0.571 0.571 0.571 0.571 0.768 0.768 Sierra Leone 0.000  0.754 Sudan 0.000 0.000 0.000 0.760 0.760 0.760 Thailand       Turkey 0.701 0.700 0.700 0.593 0.596 0.610 Uganda 0.000 0.498 0.498 0.498 0.498 0.498  Zimbabwe  0.600 0.600 0.600 0.600 0.549 0.549   154  Table 4.0.7: Tenure of System or Chief Executive (30 Reversion Cases, World Bank: Database of Political Institutions, 1972-2003)      R-5 R-4 R-3 R-2 R-1 R Albania 6 1 2 3 4 5 Armenia 1 1 2 3 4 5 Azerbaijan 2 1 Belarus 1 2 3 1 Burkina Faso 1 2 3 4 1 2 Burundi 1 2 3 4 5 6 Cambodia 14 15 1 2 3 4 Central African Republic 5 6 7 8 9 10 Comoros 1976 1 Comoros 1999 4 5 6 7 8 9 Congo, Rep. 13 1 1 2 3 4 Fiji 7 8 9 10 11 12 Gambia, The 15 16 17 18 19 20 Ghana 4 5 6 1 1 2 Guinea-Bissau 18 1 2 1 2 3 Guyana 0 1 2 3 Haiti 1991 15 1 2 3 1 1 Haiti 1999 3 1 2 3 4 5 Iran, Islamic Rep. 10 11 12 13 14 Niger 4 5 6 1 2 3 Nigeria 3 1 2 3 4 1 Pakistan 1977 29 30 31 Pakistan 1999 6 7 8 9 10 11 Peru 7 8 9 10 11 12 Sierra Leone 7 1 2 3 1 1 Sudan 15 16 1 1 2 3 Thailand 2 1 Turkey 15 16 17 18 19 20 Uganda 1 1 2 3 4 5 Zimbabwe 2 3 4 5 6 7 155  Chapter 5: International Influences and Democratic Reversion 156  Introduction  People fighting for democracy in non-democratic countries seem to have an expectation that western democracies will come to their aid.  The monks protesting in Burma last year clearly stated their expectations regarding international support.  They were disappointed to find that the west could not or would not do much to aid their cause (BBC, October 5, 2007).  The military junta in control of the country was comforted by the fact that their neighbors’ economic interests (Thailand) and political viewpoints (China) pretty much guaranteed they would not feel the pain of an international reaction (Hong 2007).  The framework advanced in this dissertation takes the position that actors considering whether to undertake a democratic reversion go through a very similar line of reasoning.  When evaluating the benefits of reversion, actors compare what they receive under the current system to what they expect to receive under a new system.  The framework advanced here takes the position that their evaluation of what they receive under present system is modified by their perceptions regarding democratic uncertainty. In addition, the framework argues their perceptions of what they expect to receive under a new system are also modified by their evaluations of risk of failure and costs of the fight.  In this chapter, we take the position that actors’ consideration of the potential roles of international actors influences their evaluation of these risks and costs.35  35 In this chapter, the notion of international actors is taken to include any actor outside the border of the country in question.  So international actors may include, for example, other countries as well as regional and  international organizations. 157  The democratization field did not devote a great deal of research attention to the role of international actors for quite some time.  As the initial purview of comparative politics, democratization research largely concentrated on domestic influences when considering regime change.  What little attention that was devoted to international factors was largely contained in economic development issues.  As late as 1995, Karen Remmer (1995, 105-6) observed that,   disciplinary traditions have created major barriers to the development of (democratization) theory capable of comprehending new international realities.  Bemoaning the artificiality of disciplinary boundaries is standard comparativist fare.  Less often noted, and far more problematic at the present theoretical juncture, are boundaries separating the study of comparative and international politics.  Precisely because of these boundaries, comparativists continue to limit their search for causal patterns to the level of the nation-state and rarely develop comparisons or generalizations that incorporate international variables…In the contemporary world, however, attempting to understand national politics in isolation from international forces is likely to prove particularly futile, if not counterproductive. Over the last decade, the field has responded to this call.  Issues of democratization are no longer viewed from a purely domestic standpoint.  While Huntington (1991) pointed out the potential for a neighborhood effect, research has now moved beyond simple regional “counter” variables and begun to attempt to understand 158  what mechanisms underlie these regional contagion effects (see, for example, Brown 2000; Gleditsch 2002; Pevehouse 2002).  As well, a huge debate has sprung up regarding democracy promotion (see, for example, Carothers 2006; Knack 2004; Ethier 2003; Kopstein 2006; Gershman and Allen 2006; Fukuyama and McFaul 2007; Finkel et al. 2007).  The events surrounding the ongoing war in Iraq have also brought increase attention to the question of the imposition of democracy (see, for example, Diamond 2005; Russett 2005; Owen 2002; Enterline and Greig 2005).  The list of research topics involving the international elements of democratization research seems to grow on a weekly basis. The research in this chapter is specifically concerned with the vulnerability of democratic regimes to international pressure to avoid democratic reversion.  As such, an examination of a country’s susceptibility to such pressure as well as the inclination of outside actors to employ these tools is considered.  Democratic conditions on aid and trade are of particular concern (Simensen 1999, 400; Finkel et al. 2007, 410; Rich 2001, 28; Fukuyama and McFaul 2007, 43).  While countries and international organizations may attempt to enforce such conditions in an effort to preserve democracy, the lack of unanimity in supporting such actions often undermines their effectiveness (Halperin and Lomasney 1998, 141, 145). While the research in this chapter examines whether we can identify such vulnerabilities and actions to exploit them across cases, the root of our concern lies in the perceptions of such possibilities.  Because for all the Burundi’s or Nigeria’s or Turkey’s of the world, actors should surely consider the possibility that they might be the next 159  Haiti or Sierra Leone.  And likewise, they should consider not only the losses implied by the potential reactions to democratic reversion but as is illustrated by cases such as Azerbaijan or Belarus, the potential benefits (in these cases of closer ties with Moscow). Cross-National, Time-Series Analysis To begin, we explore the notion that a vulnerability to pressures based on aid and trade conditions is positively related to the survival of democratic regimes.  Starting with the analysis of the dependent variable ranging from 2 to 14 according to Freedom House, we find that neither Trade (% of GDP) nor Trade in Services (% of GDP) is significant. Likewise, none of the Aid variables are found to be significant.  The only international variable that is close to being significant (0.082) is Fuel Exports (% of merchandise exports).  However, even if we were to try and count this as a significant relationship, the Pseudo R2 is so small (0.007), that it provides us with no purchase for explaining the variance.  So, when we look at regime type in its most disaggregated form, we find that none of the international influence variables are significant. Moving to the analysis of the dependent variable ranging from Free to Partly Free to Not Free, we find a number of significant relationships.  On the trade side, the base trade variable (Trade as a percent of GDP) is significant (0.000) and explains at least some of the variance (0.038) although not very much.  That other trade variable again is not significant.  Turning to the aid indicators, we find several significant relationships. The base Aid indicator (Aid as a percent of GNI) is significant (0.003) but with a pseudo R2 of 0.003 it allows us to explain very little of the variance.  The Aid per capita variable is also significant (0.000) and it allows us to explain about five percent (0.053) of the 160  variance in regime type.  We found similar results for ODA (significance of 0.000 and pseudo R2 of 0.036).  The Net Financial Flows from the IMF was not significant. Turning to the portion of the analysis where we attempt to compare reversion years to all other years, we again find that neither of the trade variables is significant. Likewise, neither Aid nor Aid per capita is significant.  Only ODA (0.056) and Net Financial Flows from the IMF (0.046) are found to be significant.  Again, however, the pseudo R2 scores (0.004 and 0.005) leave us unable to explain basically any of the variance on the dependent variable.  In this analysis, fuel imports, rather than exports, is found to be significant.  Again however, the pseudo R2 (0.010) leaves us with the ability to explain only one percent of the variance on the dependent variable. Finally we turn to the analysis of the dependent variable where we examine ongoing democratic years versus reversion country years.  Here, we find quite a few interesting results.  On the trade side, Trade as a percent of GDP is significant (0.048) but the pseudo R2 of 0.005 leaves us with little explanatory power.  Trade in Services is also significant (0.017) and while the pseudo R2 is better (0.011) it still provides us with little explanatory power.  On the aid side of the equation the results are very similar.  Again, neither Aid nor Aid per capita are found to be significant.  ODA and Net Financial Flows from the IMF are both significant (0.018 and 0.034).  However, once again, the pseudo R2 (0.012 and 0.013) leave us with little explanatory powers.  Once again, similar results are found on the fuel side.  Both fuel exports (0.045) and fuel imports (0.019) are significant) but as with the rest of the data, the pseudo R2 (0.010 and 0.015) leave us with little purchase for explaining the variance on the dependent variable. 161  The end result of the time series analysis is that while there are a number of significant relationships none of them provide us with anything beyond the most minimal purchase for explaining variance on the dependent variable.  While these results may discourage, in reality they are not unexpected.  It appears that there are simply too many confounding variables for such simple relationships to manifest themselves. Case Study Data Analysis One way around this problem is to examine the cases of democratic reversion to determine if, given all of their differences, they have in common an unusually low degree of vulnerability to the tools most commonly employed by international actors to prevent democratic reversions.  To accomplish this, first we examine trade.  The idea here is that countries that are more heavily engaged in foreign trade should be more vulnerable to international pressures to maintain democracy.  So as countries become more engaged in the international system, they should be more susceptible to pressures to maintain the norm of democratic government. Beginning with Trade as a percentage of GDP, the aggregate case study results indicate that it is difficult to make a case for the importance of this variable.  At the most basic level, during the reversion year the average level of trade was 58.6%.  The median is slightly lower (45.3%) indicating the existence of outlier cases.  Overall, this would seem to refute the case that countries with high trade levels should be more resistant to democratic reversion.  This conclusion is even more apparent when we examine the case- level data.  Of the twenty seven cases for which we have data, sixteen had trade levels below 50% of GDP.  However, only two had trade levels below 25%.  On the other hand, 162  eleven countries had trade levels in excess of 50%, nine of which were above 75%.  Of those, five cases of reversion took place when trade levels were above 100% of GDP. From this we can see that countries that engage in a moderate level of trade experience reversions and that there are examples of cases (Azerbaijan, Belarus, Congo-Brazzeville, The Gambia and Guyana) with extremely high levels of trade that also experienced reversions.  The story is the same if we look at change over time.  When we compare the reversion year to the prior year, there are thirteen cases where trade levels increase and thirteen cases where trade levels decrease.  Understanding that such relationships may take time, hence we further compared the reversion year to the average of the prior five years.  Again, the findings were mixed with twelve cases showing an increase in trade and fourteen cases showing a decline.  The results are basically the same when we use the year prior to the reversion as our base year. We find similar results when we examine trade in services (as a percent of GDP). When we compare trade during the reversion year with the previous year, it is increasing in 14 cases and decreasing in 11.  When we move the baseline year to the year prior to the reversion and compare it with the previous year, we find that trade is increasing in 9 cases and decreasing in 16.  These results appear to indicate that a fluctuation in trade in services does not make a country more or less susceptible to international influences to maintain democracy. There are a number of obvious difficulties with this position, not the least of which is it depends on with whom you are trading.  While countries that trade extensively with the United States my feel more pressure to uphold a democratic regime, countries 163  that trade with China would not.  In addition, the United States may have been more insistent on democracy during the Clinton presidency than it was following the events of 9/11.  An additional issue is that there is a relationship across time that may obscure some results.  The world in general is becoming more globalized over time, so varying levels of trade at the reversion point could be a function of time as much as an indication of vulnerability to outside pressures. Given this, we now turn to aid as an indicator that countries may be susceptible to international influences in support of democracy.  When we examine the reversion year and aid (as a percentage of GNI), we find the average across the sample is 10.34%.  The median of 5.74% indicates the influence of a few outliers.  Taking 10% as a cutoff point, there are twenty one countries below the 10% threshold during the reversion year and 9 above that threshold.  This might be taken as a positive indication of the international influence hypothesis.  It seems that very few of these countries are truly reliant on aid and thus they are able to withstand pro-democracy pressures from the international community.  There are however, nine cases that seem to cast doubt on this conclusion. These cases include several that are highly reliant on foreign aid including: Armenia (17.80%), Burundi (23.22%), Comoros 1976 (48.84%), The Gambia (19.25%) and Ghana (64.58%).  All of these cases should have been vulnerable to international pressures according to the working hypothesis yet they experienced reversions from democracy. When we look at change over time, the same story appears to be true.  When comparing the reversion year to the previous year, aid levels decreased in 17 cases and increased in 12.  So while the seventeen cases should feel they have more of a free hand 164  to change the regime, twelve cases became more susceptible to such pressures.  This problem is even more apparent when we take a longer time frame into consideration. When comparing the reversion year with the average of the previous five years, we find that there was an increase in aid in 15 cases and a decrease in aid in 14 cases.  And of those cases, only six experiences an increase or decrease of more than 10%.  This indicates that aid levels remain fairly stable and thus it is difficult to conclude that changes in aid levels influence decisions regarding regime support.  When we use the year prior to the reversion as out base year, we find very similar results.  When comparing the one year change, there are sixteen cases of increased aid and thirteen cases of decreasing aid.  As well, when we look at the five year average as the comparator, we see twelve cases of increasing aid and seventeen of declining aid. These sets of results should give us some pause for concern regarding the selection of the base year for the analysis.  In this case, the two different base years yield results pointing in different directions.  Such results should be instructive for research that relies on one or the other base year.  Different base years produce different results, so the best research will examine both for consistency. Finally, we find much the same thing when looking at Aid per capita and official development assistance and official aid.  In both cases, the overall levels do not appear to move much from year to year indicating we can find little support for something that triggers a shock (such as a regime change).  In terms of aid per capita, change in the year leading up to the reversion year is positive in fourteen cases and negative in fifteen. 165  Similarly in the ODA/OA, there is a positive change in fifteen cases and a negative change in fifteen cases. Case Level Analysis Given these findings, it is important to move our analysis “down” to the final level of the cases themselves.  Generally the case study narratives bear out the conclusions in the previous section; the impact of international influences is mixed, at best.  There appear to be four basic roles outside actors play: none, offering an ineffective response that does not restore democracy, offering an effective response that restores democracy, and actually supporting the reversion.  In this group of cases, there are only two cases where an international response resulted in return to democracy shortly after the reversion: Haiti (1991) and Sierra Leone.  On the other hand, there are quite a few cases where reversions received support (tacit or explicit) from outside actors.  Leaders seizing power in Azerbaijan and Belarus had Russian backing.  The coup in Fiji apparently got the green light from the United States (which was concerned with the loss of naval access).  There is also evidence of US support of the coup in Turkey.  The French played a role in reversions in Comoros (1976) as well as in Congo.  French forces put down a coup in Comoros in 1995 but not in 1999.  And reversions in Sudan and Thailand were generally greeted with support for their presumed stabilizing influence. So the cases analyses to follow do not paint a picture of a consistent, effective international response to democratic reversion.  This should not be surprising. In terms of the general framework advanced in the dissertation, the case evidence does point to the necessity for actors to consider the implications of the potential 166  international response.  The aggregate case data in the previous section pointed to the potential vulnerabilities of countries to the response of outside actors.  And the case information to be presented in this section points to some form of a response by outside/international actors.  While the basic structural approach either ignores or greatly under-specifies the consideration actors give to potential responses by outside actors to a reversion move, the general framework advanced in chapter two explicitly includes such a calculation.  Not only do actors consider problems with the payoffs they receive in the current system but they compare those payoffs to what they expect to receive under an alternative regime.  And the consideration of those payoffs is tempered by potential international ramifications of a democratic reversion.  In some cases, such as Azerbaijan or Belarus, this may be an attractive element (as there was an additional benefit from expected closer ties with Moscow).  In other cases, such as Pakistan (1999) it was clear that the implications would be more negative.  So the case findings in this section demonstrate the relevance of the consideration actors give to costs involved in reversion moves when comparing their current benefits to the benefits expected given a reversion move.  Albania:  There was little international involvement in the democratic reversion. The data point to a country that should be vulnerable to international pressures.  The importance of trade had declined over time but remained important (48%).  Likewise, the importance of aid had declined over time however at 7.3% of GNI, it was also important to the country.  The case narrative reveals little international involvement.  OSCE observers were present during the 1996 election and the organization issued a report outlining the violations of Albanian election law and condemning the manipulation of the 167  polls. Beyond that, the international community had no effect on the reversion (Poggioli 1997, 47; Vickers 2008; Nicholson 1999; Sunley 1998). Armenia: There was little international involvement in the democratic reversion. The data point to a country highly vulnerable to international pressure.  While trade levels had declined, they remained at 79% of GDP.  Meanwhile, aid constituted 18% of the GNI.  The case narrative reveals little international involvement.  The OSCE observed the 1996 election and issued a report pointing to serious breaches of election law and problems with ballot counts.  The United States, a heavy contributor of aid to Armenia, declined to offer routine congratulations to Ter-Petrossian.  None of these actions prevented the president from seizing power.  However, two years later, Ter- Petrossian was forced to resign under international pressure (Bremmer and Welt 1997, 5; Herzig 2008; Specter 1997; Freedom House: Armenia 2006). Azerbaijan:  International influence appears to have played a role in supporting (not opposing) the democratic reversion.  The data point to one of the most heavily trade- reliant countries in the sample (133%).  The aid data are very limited but indicates a low level of importance (2%).  The case narrative reveals international support for the 1993 coup.  Russia welcomed the coup leader’s appointment of Geidar Aliyev as President indicating the events would likely lead to an improvement in bilateral relations.  Within a year of the coup, agreements were reached with a consortium of Western-based oil companies over the development of Caspian Sea oil.  As well, economic reform policies pursued by Aliyev were supported by the World Bank and the IMF (Yorke and Fumagalli 168  2008; Cornell 2001; Freedom House: Azerbaijan 2002; OSCE Report Azerbaijan 2006; Kamrava 2001). Belarus: There are international efforts both to oppose and support the democratic reversion.  The data point to a country that was heavily engaged in trade (104%) but not overly reliant on aid (1.6%).  The case narrative reveals widespread opposition to President Lukashenko’s efforts to seize power.  The United States issued a statement saying the constitutional referendum was “deprived of legitimacy” (Facts on File). As well, most western nations refused to recognize the 1996 Constitution of the new parliament (Freedom House 2002).  The effect of such condemnation was likely offset by support Lukashenko (a former KGB agent) received from Moscow due to his pro-Russian policy positions (Sannikov) (Ryder 2008; Vera 1997; Potock 2002; Vitali 2005; Sannikov 2005). Burkina Faso:  There is no indication of international factors influencing the democratic reversion.  The data point to a country not particularly reliant on trade (40%) or aid (11%).  The case narrative does not point to any substantial international attention to the coup in 1980 (Englebert and Murison 2008; Freedom House 1983; Keesing’s February 1980; Keesing’s June 1981). Burundi:  There is no indication of international factors influencing the democratic reversion.  The data point to a country that was not particularly reliant on trade (37%); however the country was heavily dependent on aid (23%).  Despite the potential influence of aid, there is no indication of any substantial international involvement in the democratic reversion.  As well, the international community was slow 169  to respond to the subsequent genocidal events that followed the reversion (Mthembu- Salter 2008; Lemarchand 1989; Boyer 1992; Reyntjens 1993; Watson 1993). Cambodia:  There was very substantial international stake in the transition to democracy in Cambodia.  The data point to a country engaged in trade (79%) but not overly reliant on aid (10%).  The case narrative reveals a substantial international effort to establish democracy in Cambodia.  The United Nations Transitional Authority in Cambodia (UNTAC) supervised the 1993 elections and the World Bank and the IMF subsequently became involved in efforts to reform the Cambodian economy.  Following the 1997 coup, both of these organizations cut off assistance (Peou and Summers 2008; Chad 1996; Lapidus 1998; Maley and Sanderson 1998; Peang-Meth 1997). Central African Republic:  There was a substantial international stake in the transition to democracy as well as heavy international involvement in the reversion.  The data point to a country that was not heavily engaged in trade (29%) or reliant on aid (4%). The case narrative indicates that UN peacekeepers and international election observers were on the ground during the 1999 transitional election.  Efforts to seize the capital in 2001 were supported by troops from Rwanda and rebuffed with the support of troops from Libya and then Democratic Republic of the Congo.  President Patasse subsequently accused France of having substantial involvement in the attempt to take control of the CAF.  The following year Patasse employed a Congolese rebel group to suppress opposition to his government.  While there was clearly international involvement in the reversion, there is no indication of a concerted effort to restore democracy (Englebert et al. 2008; New African 2001; Freedom in the World: Central African Republic 2006; 170  World News Digest July 7, 2001; Keesing’s “Government Changes” December 2000; Economist December 8, 2003). Comoros (1976): The French were heavily involved in the democratic reversion. The trade data are missing.  The data point to a very high reliance (49%) on aid.  The case narrative indicates a very substantial French role in the events surrounding the reversion.  A referendum on independence from France received a 96% yes vote in December 1974.  France subsequently attempted to persuade the government to draft a constitution and hold another election to ratify it prior to independence.  President Ahmed Abdallah rejected these demands and declared unilateral independence in July 1975.  A mere twenty eight days later, the opposition parties supported by the French Army and French mercenaries overthrew the government.  Three years later, fifty former French mercenaries seized control of the country in a coup that received widespread popular support (Recent History: The Comoros 2008; Bakar 1988; Merrill 1993; The Globe and Mail May 15, 1978). Comoros (1999):  There was substantial international involvement in the democratic reversion.  The data point to a country substantially engaged in trade (47%) and fairly reliant (16%) on aid.  The case narrative points to the potential role international actors can play in preventing a reversion.  In September 1995, three hundred members of the Comoros military and thirty French mercenaries led by Bob Denard (a French national) ousted the democratic government in a bloody coup.  The French government suspended economic aid and subsequently sent in 1,000 troops to put down the coup.  However, when the Comoros Army seized power in April 1999, there is no 171  indication of any international effort to restore democracy (Recent History: The Comoros 2008; Bakar 1988; Merrill 1993; The Globe and Mail May 15, 1978; Bratton 2007; National Post 1999; Freedom in the World: Comoros 2002). Congo-Brazzaville: There is reliable evidence of heavy French support for the democratic reversion.  The Country was highly reliant on both trade (135%) and aid (16%).  The case narrative indicates a substantial role of international actors in support of the reversion and no international effort to restore democracy.  Sassou-Nguesso seized control of the country in October 1997 with Angolan military support and the political backing of France.  Following the coup, Sassou-Nguesso was given a lengthy audience with French President Chirac at a Francophone summit in Hanoi (The Economist November 1997).  Following the meeting, the French Secretary of State, Charles Josselin, remarked that Sassou-Nguesso was the man “to put in place a new democratic process” (Englebert and Murison 2008; Bazenguissa-Ganga 1998; Clark 1997; Clark 2002; Eaton 2007; Roberts 1998; StarPhoenix October 1997; Economist November 1997). Fiji: There was substantial United States support for the democratic reversion. The data show that trade accounted for 87% of the GDP and aid constituted 3% of the GNI.  The case data note that in his first press conference following the 1987 election, Prime Minister Bavarda stated that the Soviet Union did not represent a threat to the region and that Fiji could move out from under the US defense umbrella (Bedford).  He subsequently vowed to close Fijian ports to US nuclear warships.  There are indications that US officials visiting the island, including the US Ambassador to the United Nations Vernon Walters and Secretary of state George Shultz, gave the military a green light to 172  overthrow Bavarda (Recent History: Fiji 2008; Bedford 1987; Digirolamo 1987; Keesings December 1987; Economist December 12, 1987). The Gambia: There was a highly ineffective international response to the democratic reversion.  The data show that trade accounted for 103% of the GDP and aid constituted 19% of the GNI.  The case narrative points out that following the 1994 coup most of the country’s military and economic aid was suspended.  In response to ongoing international pressure, a new constitution was drafted and approved in a 1997 referendum.  Subsequently, the coup leader, Yahya Jammeh, gained the presidency in elections that were judged by international observers as not free and fair (Wiseman and Murison 2008; US State Department Background Note 2007; Keesing’s Failed Coup Attempt January 1995; Keesing’s Military Coup July 1994). Ghana: There are no indications of an international effort to prevent the democratic reversion.  The data show that trade accounted for 10% of the GDP and aid constituted 3% of the GNI.  The case narrative does not indicate any substantial international response to the 1981 coup (Synge and McCaskie 2008; Keesings May 1992; Petchenkine 1993; Freedom House: Ghana 1983; CIA World Factbook: Ghana 2008). Guinea-Bissau: There is evidence of a sustained international effort that resulted in a successful return to democracy following the reversion.  The data show that trade accounted for 77% of the GDP and aid constituted 65% of the GNI.  The case narrative indicates that in response to an executive seizure of power, the military staged a coup in September 2003.  The coup was received support from the public as well as the democratically elected legislature.  Supported by the international community, elections 173  returned democracy to Guinea-Bissau in 2005 (Peitte et al. 2008; Freedom House: Guinea-Bissau 2004; Gazette September 152003; Sonko 2002). Guyana:  There are no indications of an international effort to prevent the democratic reversion.  The data show that trade accounted for 126% of the GDP and aid constituted 6% of the GNI.  The case narrative points out that by the end of 1976, worsening terms of trade as well as donor fatigue on the part of the United States worsened an already poor economic situation in Guyana.  In response, Prime Minister Forbes Burnham turned to the IMF for assistance.  Despite Burnham’s seizure of power following the fraudulent 1978 referendum, the IMF continued to grant loans to the country (Smith 2008; Singh 1997; Felix 1998; Griffith 1991; Chandisingh 1983; Rodney 1981). Haiti:  The international community played a substantial role in the democratic transition and following the democratic reversion international pressure, led by the Organization of American States and the United States, eventually returned the President to power.  The data show that trade accounted for 58% of the GDP and aid constituted 6% of the GNI.  The United States was extensively involved in one way or another at every step of the democratization process following Duvalier’s exit in 1986.  While there are assertions that the U.S. backed the coup that ousted President Aristide in 1991, in the end, the Clinton administration was responsible for negotiating his return to power in 1993.  Over the course of time following the coup, the US and the UN imposed oil and trade embargoes and the Organization of American States took the unprecedented step of establishing a trade embargo.  The effectiveness of such actions was to some extent 174  undercut by countries in Europe, South American, and Africa who routinely ignored the trade embargo.  Facing an imminent US invasion, in September 1994 the Haitian military gave up power.  American troops took control of the country, and Aristide was reinstated to power (Aurthur 2008; Shultz 1997; Rohter 1996; Fatton 1999; Economist December 4, 1999; Lundahl and Silie 1998). Iran (2004):  International factors appear to have contributed to the democratic reversion and not to have played a role in attempting to restore democracy.  The data show that trade accounted for 55% of the GDP and aid constituted 0.12% of the GNI. While there were several factors that caused the demise of the reform movement in Iran, the case narrative points to the important role that rhetoric and actions by the United States played in the democratic reversion.  The reform movement favored a rapprochement with the US.  There is recent evidence of an unexpectedly toned-down Iranian response following the events of 9/11 as well as efforts on behalf of the Iranian government to assist the initial US-led invasion of Afghanistan.  These efforts, as well as the reform movement, were significantly undercut when Iran was included as part of the “Axis of Evil” in George W. Bush’s 2002 State of the Union address.  The reform movement suffered a further blow when the US invasion of Iraq was able to capture Baghdad so quickly.  Following the 2004 Iranian election, the international community has been concerned with the country’s nuclear program, not its democratic movement (Cronin 2008; Momayesi 2000; Takeyh 2003; Sanam 2007; Mason 2002; Rajaee 2004; Gheissari and Nasr 2004). 175  Niger:  There are no indications of an international effort to prevent the democratic reversion.  The data show that trade accounted for 40% of the GDP and aid constituted 13% of the GNI.  The case narrative points out that this was a domestic event driven mainly by the country’s economic situation.  There are no indications of a significant international reaction to the 1996 coup.  In fact, aid levels increased from 12.8% (of GNI) to 18.3% the year following the reversion (Englebert and Murison 2008; Keesing’s: Reaction to the Coup January 1996; Keesings: Legislative Elections November 1996; Amnesty International 1996; CIA World Factbook: Niger 2008). Nigeria:  There are no indications of an international effort to prevent the democratic reversion.  The data show that trade accounted for 31% of the GDP and aid constituted 0.14% of the GNI.  The case narrative demonstrates that the reversion was a domestic event and there is no indication of any substantial international reaction to the 1983 New Year’s Eve coup.  In fact, within a year aid levels had doubled.  So if anything was driving the international response it was access to Nigerian oil, not concerns regarding Nigerian democracy (Synge 2008; Freedom House: Nigeria 1985; Keesing’s: General Elections January 1984; Keesing’s: Overthrow of Civilian Government May 1984; Joseph 1984). Pakistan (1977):  There are indications of a largely ineffective international response to the democratic reversion.  The data show that trade accounted for 28% of the GDP and aid constituted 3.7% of the GNI.  When General Muhammad Zia-ul-Haq took control of the country in a coup on July 5, 1977, he announced that the Constitution had not been abrogated, only suspended and he promised elections within 90 days.  Shortly 176  thereafter, he reversed this policy in favor of seeking retribution against members of the Bhutto government.  In September 1978, following a favorable Supreme Court ruling on the necessity of the coup, Zia assumed the role of the President.  Bhutto was hung for his alleged involvement in the death of opposition members the following year.  The international response, particularly to the hanging of Bhutto, was one of widespread condemnation.  Aid levels were cut in half following the coup.  None of these actions resulted in enough pressure to prevent Zia’s actions following the coup or to restore democracy in Pakistan (McPherson 2008; Keesing’s: Riots 1977; Keesing’s: Arrest of Bhutto 1978; Freedom House: Pakistan 1979; CIA World Factbook: Pakistan 2008). Pakistan (1999):  There are indications of an ineffective international response to the democratic reversion.  The data show that trade accounted for 32% of the GDP and aid constituted 1.17% of the GNI.  The case narrative tells a slightly different story than the data with regards to aid during the lead-up to the reversion.  The narrative points out that the costs incurred due to the country’s nuclear efforts substantially increase Pakistan’s debt and placed it in violation of its agreements with the IMF.  As a result of these problems with the IMF, as well as in response to the nuclear program, foreign aid significantly decreased prior to the reversion.  As well, there a strong indications that the United States warned the Pakistani military not to attempt to take power.  The coup was widely condemned by the international community and was seen as a serious blow to American credibility in the region.  While aid levels slightly decreased following the coup, they massively increased following 9/11 (McPherson 2008; Keesing’s: May 2000; CIA World Factbook: Pakistan 2008; Shah 2002; Freedom House: Pakistan 2002; Islam 2001; Ameen 1999). 177  Peru (1992):  Despite strong, widespread international opposition to the democratic reversion, the international response did not result in the return to democracy immediately following the reversion.  The data show that trade accounted for 28% of the GDP and aid constituted 1% of the GNI.  The case narrative indicates the initial response to Fujimori’s autogolpe was more muted than the earlier response to the 1991 Haitian coup.  Initially, the international response centered on harsh public condemnation.  The Bush administration criticized Fujimori, as exemplified by Secretary of State James Baker’s statement that “you cannot destroy democracy in order to save it.”  The OAS passed a resolution that “greatly deplored” Fujimori’s action.  However, after failing to negotiate an immediate return to democracy with Fujimori, the international community turned to economic sanctions.  Washington halted its economic and military aid.  A $1.1 billion international aid package was blocked.  Loans and credits from the IADB and the World Bank were halted.  Japan did not however, suspend its aid.  Despite that, the effects of such international actions reverberated through the Peruvian economy. However these actions were not enough to force Fujimori to immediately restore democracy.  In the end, the US, for example, settled for promises rather than actual democratic changes.  Within a year, the Clinton administration had already begun to restore economic aid to the country (Markwick 2008; Cameron 1998; Cameron 1994; Cameron and Mauceri 1997; Cameron 1988; Friedman April 11, 1992; Friedman April 14, 1992;  Holmes February 25, 1993; Nash May 18, 1992; Nash April 23, 1992; Crosette April 7, 1992; Freedom House: Peru 1995; Keesings April 1992; The Economist April 11, 1992). 178  Sierra Leone:  The international response to the democratic reversion was instrumental in the eventual restoration of democracy.  The data show that trade accounted for 29% of the GDP and aid constituted 14% of the GNI.  The case narrative points out that following the coup in 1997, the Koroma government was completely isolated by the international community.  Sierra Leone was suspended from the Commonwealth.  The UN Security Council condemned the coup.  The Economic Community of West African States (ECOWAS) imposed an embargo on the country backed by Nigerian occupation of the Lungi airport (the country’s main airport) and an Economic Community of West African States Monitoring Group (ECOMOG) naval blockade.  The UN also subsequently imposed sanctions.  In February 1998, the ECOMOG forces retook control of the capital, Freetown.  President Kabbah returned to Freetown in March and was formally inaugurated (Synge and Clapham 2008; Bell 2000; Keesing’s: Return to Civilian Rule March 1996; Keesing’s: Chaotic Aftermath June 1997; Freedom House: Sierra Leone 1999). Sudan (1989):  The international community failed to restore democracy following the reversion.  The trade data are missing.  The data indicate that aid constituted 5% of the GNI.  Following the 1989 coup, General Omar Hassan Ahmad al- Bashir, formed Revolutionary Command Council for National Salvation (RCC) and declared that its primary objective was to resolve the conflict consuming the southern part of Sudan.  The RCC received diplomatic recognition from groups of neighboring countries and was generally welcomed by the international community as a potentially stabilizing influence in the region (Synge and Clapham 2008; Morrison 2005; Keesing’s: Fighting South War January 1990; Freedom House: Sudan 1992). 179  Thailand:  The international response to the democratic reversion was generally supportive and did not result in the return of democracy.  The data show that trade accounted for 43% of the GDP and aid constituted 1% of the GNI.  The case narrative indicates that Thailand was a major recipient of US aid during the period surrounding the reversion.  The major US objective at the time was stability and there was no substantial push to resort democracy (McVey and Jory 2008; Keesing’s: General Election July 1976; Keesing’s: Military Coup in Bangkok December 1976; Keesing’s: National Administrative Reform December 1976; Freedom House: Thailand 1978). Turkey:  The international community played no real role in the restoration of democracy following the reversion.  The data show that trade accounted for 17% of the GDP and aid constituted 1% of the GNI.  The case narrative indicates that upon taking control of the country in 1980, the Turkish military announced that it would draft a new constitution and hold elections returning civilians to power.  At the time of the coup, the US had more than 3,000 troops in the country.  There are indications that the US did nothing to discourage (and maybe encouraged) the military to seize power.  The military’s brutal repression of the instigators of political violence soured relations with Europe but the US remained steadfast and uncritical in its support of this vital geostrategic ally.  Civilian control was restored in 1983 following elections (Day and Hale 2008; Keesing’s: Developments Following the Coup May 1981; Keesing’s: Assumption of Power October 1980; Birand 1987; Amnesty International 1988; Karasapan 1989). 180  Uganda:  International pressure played more of a role in triggering the democratic reversion than restoring democracy.  The data show that trade accounted for 29% of the GDP and aid constituted 5% of the GNI.  The case narrative indicates that the international community attempted to pressure President Obote to reach a settlement with the rebel National Resistance Army (NRA) led by Yoweri Museveni.  This resulted in a rift between Obote and the military that eventually triggered the 1985 coup.  The coup leaders promised an election to return civilian power to Uganda and signed a peace accord with the NRA in December 1985.  The peace was short-lived and the NRA seized control of the country in January 1986.  There appears to have been little international opposition to the coup and no pressure to return democracy to the country (Rake and Jennings 2008; Freedom House: Uganda 1987; Keesing’s: Internal Security Situation April 1985; Keesing’s: Military Coup December 1985; Keesing’s: Overthrow January 1986). Zimbabwe:  The international community was unable to restore democracy following the reversion.  The data show that trade accounted for 45% of the GDP and aid constituted 4% of the GNI.  The case narrative indicates that Mugabe’s opposition to apartheid in South Africa as well as his Marxist tendencies did not invite support from the US or England.  He was however able to gather international support through the non- aligned movement. Prior to the reversion, South Africa attempted to destabilize the regime.  In 1987, they conducted raids into Zimbabwe in May and were behind a substantial set of bombings in October and in early 1988. The larger international community had no real response to his seizure of power in 1987 (Brown and Saunders 2008; Freedom House: Zimbabwe 1989; Keesing’s: Creation of Executive President 181  January 1988; Keesing’s: Progress Toward Party Merger February 1987; Economist: Zimbabwe November 7, 1987; Economist: Zimbabwe December 5, 1987). Conclusion The results found in this chapter demonstrate that international factors do not uniformly play an important role in decision-making regarding democratic reversion. While the democratization literature did not devote much attention to international influences for quite a long time, in recent years there has been an explosion of research addressing such issues.  The research in this chapter is specifically concerned with the vulnerability of democratic regimes to international pressure to avoid democratic reversion.  As such, an examination of a country’s susceptibility to such pressure as well as the inclination of outside actors to employ these tools is considered. While the results of the time series analyses point to a number of significant relationships, none of them provide us with anything beyond the most minimal purchase for explaining variance on the dependent variable.  The aggregate analysis of the thirty reversion cases similarly points to potential vulnerabilities to international pressure but concludes that they provide us with little purchase to explain reversion.  Generally the case study narratives bear out these conclusions; the impact of international influences is mixed, at best.  The case studies do however point to four basic roles outside actors play: none, offering an ineffective response that does not restore democracy, offering an effective response that restores democracy, and actually supporting the reversion. In terms of the general framework advanced in the dissertation, the case evidence does point to the necessity for actors to consider the implications of the potential 182  international response.  The aggregate case data in the previous section pointed to the potential vulnerabilities of countries to the response of outside actors.  And the case information to points to some form of a response by outside/international actors.  While the basic structural approach either ignores or greatly under-specifies the consideration actors give to potential responses by outside actors to a reversion move, the general framework advanced in chapter two explicitly includes such a calculation.  Not only do actors consider problems with the payoffs they receive in the current system but they compare those payoffs to what they expect to receive under an alternative regime.  And the consideration of those payoffs is tempered by potential international ramifications of a democratic reversion.  So the case findings in this chapter demonstrate the relevance of the consideration actors give to costs involved in reversion moves when comparing their current benefits to the benefits expected given a reversion move.  183  Table 5.0.1: International Influences - Regime Type Rated 2-14  Source: World Bank, World Development Indicators, 1972-2003 Method: Logistic Regression, n = 6016  Table 5.0.2: International Influences - Regime Type Rated Free, Partly Free, Not Free  Source: World Bank, World Development Indicators, 1972-2003 Method: Logistic Regression, n = 6016   Significance Pseudo R2 Trade (% GDP) 0.544 0.002 Net Trade in Goods & Services  0.491 0.003 Aid (% GNI) 0.885 0.002 Aid per capita  0.987 0.001 ODA & Official Aid  0.496 0.003 Net Financial Flows, IMF 0.822 0.002 Fuel Exports 0.082 0.007 Fuel Imports 0.272 0.004  Significance Pseudo R2 Trade (% GDP) 0.000* 0.038 Net Trade in Goods & Services  0.438 0.000 Aid (% GNI) 0.003* 0.003 Aid per capita  0.000* 0.053 ODA & Official Aid  0.000* 0.036 Net Financial Flows, IMF 0.172 0.001 Fuel Exports 0.642 0.002 Fuel Imports 0.836 0.000 184  Table 5.0.3: International Influences - Democratic Reversions versus All Remaining Country Years (both Democratic and Non-Democratic)  Source: World Bank, World Development Indicators, 1972-2003 Method: Logistic Regression, n = 6016  Table 5.0.4: International Influences - Democratic Reversions versus Democratic Country Years  Source: World Bank, World Development Indicators, 1972-2003 Method: Logistic Regression, n = 6016                  Significance Pseudo R2 Trade (% GDP) 0.161 0.003 Net Trade in Goods & Services  0.787 0.000 Aid (% GNI) 0.692 0.000 Aid per capita  0.648 0.000 ODA & Official Aid  0.056* 0.004 Net Financial Flows, IMF 0.046* 0.005 Fuel Exports 0.699 0.010 Fuel Imports 0.058* 0.010                Significance Pseudo R2 Trade (% GDP) 0.048* 0.005 Net Trade in Goods & Services  0.017* 0.011 Aid (% GNI) 0.395 0.002 Aid per capita  0.723 0.000 ODA & Official Aid  0.018* 0.012 Net Financial Flows, IMF 0.034* 0.013 Fuel Exports 0.045* 0.010 Fuel Imports 0.019* 0.015 185  Table 5.0.5: Trade, % GDP (30 Reversion Cases, World Bank: World Development Indicators, 1972-1993)   Trade (% of GDP) Rev-5 Rev-4 Rev-3 Rev-2 Rev-1 Reversal R-(R-1) Change 5YrAv R-5Av Change (R-1)-(R-2) Change 4YrAv (R-1)-4Av Change Albania 34.82 100.00 77.72 50.30 46.99 47.55 0.56 1.19% 61.96 -14.42 -23.27% -3.31 -6.59% 65.71 -18.72 -28.49% Armenia 100.87 101.08 107.97 112.43 86.11 79.23 -6.89 -8.00% 101.69 -22.46 -22.09% -26.31 -23.41% 105.59 -19.47 -18.44% Azerbaijan .. .. 83.10 86.86 140.80 133.42 -7.38 -5.24% 103.58 29.84 28.80% 53.94 62.11% 84.98 55.82 65.69% Belarus 89.61 70.26 117.12 151.00 155.37 103.72 -51.66 -33.25% 116.67 -12.96 -11.10% 4.37 2.89% 107.00 48.38 45.21% Burkina Faso 37.50 35.21 38.40 35.22 37.53 40.21 2.68 7.15% 36.77 3.44 9.36% 2.31 6.56% 36.58 0.94 2.58% Burundi 38.54 32.69 35.65 38.42 38.23 36.54 -1.68 -4.40% 36.71 -0.16 -0.44% -0.20 -0.51% 36.33 1.90 5.23% Cambodia 34.32 48.87 64.58 77.73 69.25 79.06 9.81 14.17% 58.95 20.11 34.12% -8.48 -10.91% 56.37 12.88 22.84% Central African Rep 42.26 25.71 29.44 27.91 28.67 .. 0.77 2.75% 31.33 -2.65 -8.47% Comoros76 .. .. .. .. .. .. Comoros99 67.32 64.33 59.70 61.17 46.53 47.06 0.53 1.13% 59.81 -12.75 -21.33% -14.64 -23.93% 63.13 -16.60 -26.30% Congo-B 82.95 94.36 132.73 128.31 128.68 135.76 7.08 5.50% 113.40 22.35 19.71% 0.37 0.29% 109.59 19.09 17.42% Fiji 92.87 92.65 86.72 89.10 81.14 87.36 6.22 7.67% 88.50 -1.14 -1.28% -7.96 -8.94% 90.33 -9.20 -10.18% Gambia 120.34 131.49 138.58 138.76 132.98 102.70 -30.27 -22.77% 132.43 -29.73 -22.45% -5.79 -4.17% 132.29 0.68 0.52% Ghana 31.76 22.04 18.05 22.39 17.62 10.08 -7.54 -42.80% 22.37 -12.29 -54.95% -4.77 -21.31% 23.56 -5.94 -25.21% Guinea-B 50.29 67.68 83.35 91.56 82.17 77.23 -4.94 -6.02% 75.01 2.22 2.96% -9.39 -10.25% 73.22 8.95 12.23% Guyana 121.08 134.76 149.79 157.08 141.98 126.18 -15.79 -11.12% 140.94 -14.76 -10.47% -15.11 -9.62% 140.68 1.30 0.92% Haiti91 38.07 40.78 39.51 37.36 37.53 57.92 20.39 54.35% 38.65 19.27 49.86% 0.16 0.44% 38.93 -1.41 -3.61% Haiti99 17.09 35.38 39.35 37.13 36.31 40.64 4.33 11.92% 33.05 7.58 22.95% -0.82 -2.20% 32.24 4.07 12.63% Iran 36.29 40.14 39.29 49.27 52.94 54.98 2.04 3.85% 43.58 11.39 26.14% 3.67 7.45% 41.25 11.69 28.35% Niger 32.78 34.59 33.74 43.41 41.47 40.41 -1.06 -2.55% 37.20 3.21 8.63% -1.95 -4.48% 36.13 5.34 14.77% Nigeria 43.31 43.88 48.57 49.11 38.65 31.14 -7.51 -19.44% 44.71 -13.57 -30.34% -10.46 -21.29% 46.22 -7.57 -16.37% Pakistan77 28.76 29.83 34.46 33.25 30.10 28.31 -1.79 -5.95% 31.28 -2.97 -9.50% -3.15 -9.48% 31.57 -1.48 -4.68% Pakistan99 35.33 36.13 38.33 36.85 34.01 32.32 -1.69 -4.97% 36.13 -3.81 -10.55% -2.84 -7.71% 36.66 -2.65 -7.23% Peru 23.69 41.45 24.64 29.60 27.14 28.37 1.24 4.55% 29.30 -0.93 -3.17% -2.46 -8.32% 29.84 -2.71 -9.07% Sierra Leone 45.83 43.32 55.16 45.10 53.32 29.11 -24.21 -45.40% 48.55 -19.44 -40.04% 8.21 18.21% 47.35 5.96 12.59% Sudan 24.93 17.63 12.96 16.66 .. .. Thailand 34.81 37.33 38.63 45.56 41.35 42.94 1.60 3.87% 39.54 3.41 8.62% -4.21 -9.25% 39.08 2.26 5.79% Turkey 14.97 14.92 13.99 11.08 8.87 17.09 8.22 92.64% 12.77 4.32 33.87% -2.21 -19.96% 13.74 -4.87 -35.43% Uganda 45.48 38.14 25.90 22.30 27.02 28.75 1.73 6.40% 31.77 -3.02 -9.51% 4.71 21.14% 32.95 -5.94 -18.01% Zimbabwe 39.15 35.92 41.37 44.21 45.57 45.29 -0.28 -0.61% 41.24 4.05 9.82% 1.36 3.07% 40.16 5.41 13.47% Average 50.18 53.95 58.92 61.00 61.01 58.64 -3.57 0.00 59.87 -1.23 -0.01 -1.58 -0.03 57.96 3.05 0.02 Median 38.31 40.46 39.51 45.10 43.52 45.29 -0.28 -0.01 43.58 -0.93 -0.01 -2.34 -0.06 40.70 0.81 0.01 186  Table 5.0.6: Trade in Services, % GDP (30 Reversion Cases, World Bank: World Development Indicators, 1972-1993)   Trade in services (% of GDP) Rev-5 Rev-4 Rev-3 Rev-2 Rev-1 Reversal R-(R-1) Change 5YrAv R-5Av Change (R-1)-(R-2) Change 4YrAv (R-1)-4Av Change Albania 3.74 15.42 19.50 10.66 10.53 10.57 0.04 0.4% 11.97 -1.40 -11.7% -0.13 -1.2% 12.33 -1.80 -14.6% Armenia .. .. 4.77 4.13 5.51 12.91 7.41 134.5% 4.80 8.11 168.8% 1.37 33.2% 4.45 1.05 23.7% Azerbaijan .. .. .. .. .. .. Belarus .. .. .. 1.98 3.02 5.37 2.35 77.8% 2.50 2.87 114.9% 1.04 52.8% 1.98 1.04 52.8% Burkina Faso 9.42 9.65 9.95 10.77 12.59 13.38 0.79 6.3% 10.48 2.91 27.7% 1.82 16.9% 9.95 2.65 26.6% Burundi 11.71 9.69 12.89 14.30 14.29 13.79 -0.50 -3.5% 12.58 1.22 9.7% -0.01 -0.1% 12.15 2.15 17.7% Cambodia 5.72 7.30 6.97 8.77 10.78 10.13 -0.65 -6.0% 7.91 2.23 28.2% 2.01 23.0% 7.19 3.59 50.0% Central African Rep .. .. .. .. .. .. Comoros76 .. .. .. .. .. .. Comoros99 40.07 36.38 .. .. .. .. Congo-B 27.40 46.98 60.07 36.82 40.46 36.52 -3.94 -9.7% 42.35 -5.83 -13.8% 3.64 9.9% 42.82 -2.36 -5.5% Fiji 32.73 33.98 32.04 35.38 32.61 31.83 -0.79 -2.4% 33.35 -1.52 -4.6% -2.77 -7.8% 33.53 -0.92 -2.7% Gambia 40.95 34.49 48.22 42.54 42.07 43.39 1.32 3.1% 41.65 1.73 4.2% -0.47 -1.1% 41.55 0.52 1.2% Ghana 13.48 12.21 10.81 8.90 8.47 9.81 1.34 15.8% 10.77 -0.96 -8.9% -0.43 -4.8% 11.35 -2.88 -25.3% Guinea-B .. .. .. 17.45 16.87 17.97 1.10 6.5% 17.16 0.81 4.7% -0.58 -3.3% 17.45 -0.58 -3.3% Guyana .. .. .. .. 17.09 15.43 -1.67 -9.7% 17.09 -1.67 -9.7% Haiti91 12.14 13.92 13.05 11.03 4.34 4.65 0.32 7.3% 10.90 -6.24 -57.3% -6.70 -60.7% 12.54 -8.20 -65.4% Haiti99 2.97 13.36 12.92 15.68 10.52 10.01 -0.50 -4.8% 11.09 -1.08 -9.7% -5.16 -32.9% 11.23 -0.72 -6.4% Iran 3.51 3.63 .. .. .. .. Niger 10.86 11.03 13.82 11.49 9.84 10.33 0.50 5.0% 11.41 -1.07 -9.4% -1.65 -14.3% 11.80 -1.96 -16.6% Nigeria 10.52 9.44 9.99 9.93 7.95 7.97 0.02 0.3% 9.57 -1.59 -16.7% -1.99 -20.0% 9.97 -2.02 -20.3% Pakistan77 .. .. .. .. 5.39 5.33 -0.07 -1.2% 5.39 -0.07 -1.2% Pakistan99 8.25 7.91 8.65 6.86 5.89 5.59 -0.31 -5.2% 7.51 -1.92 -25.6% -0.97 -14.1% 7.92 -2.02 -25.6% Peru 8.18 16.12 9.62 7.46 5.97 6.24 0.27 4.5% 9.47 -3.24 -34.2% -1.49 -20.0% 10.35 -4.38 -42.3% Sierra Leone 16.16 15.61 22.78 20.52 18.11 9.37 -8.74 -48.3% 18.64 -9.27 -49.7% -2.41 -11.7% 18.77 -0.66 -3.5% Sudan 5.41 5.78 2.56 2.00 2.71 4.12 1.42 52.5% 3.69 0.43 11.7% 0.70 35.0% 3.94 -1.23 -31.3% Thailand .. .. .. .. 8.27 7.39 -0.88 -10.7% 8.27 -0.88 -10.7% Turkey 2.26 2.06 2.00 1.17 1.18 1.81 0.62 52.8% 1.73 0.07 4.1% 0.01 1.2% 1.87 -0.69 -36.9% Uganda 10.71 10.82 .. .. 2.74 4.35 1.61 58.6% 8.09 -3.74 -46.3% 10.77 -8.02 -74.5% Zimbabwe 7.25 7.42 7.82 13.44 7.80 7.63 -0.17 -2.2% 8.75 -1.11 -12.7% -5.64 -42.0% 8.98 -1.18 -13.2% Average 13.50 15.39 16.23 13.87 12.20 12.24 0.04 0.13 13.08 -0.85 0.02 -0.94 -0.03 13.77 -1.30 -0.10 Median 10.52 11.03 10.81 10.77 8.47 9.81 187  Table 5.0.7: Aid, % GNI (30 Reversion Cases, World Bank: World Development Indicators, 1972-1993)    Aid (%GNI) Rev-5 Rev-4 Rev-3 Rev-2 Rev-1 Reversal R-(R-1) Change 5YrAv R-5Av Change (R-1)-(R-2) Change 4YrAv (R-1)-4Av Change Albania 29.71 59.26 24.28 8.19 7.26 7.34 0.08 1.1% 25.74 -18.40 -71.5% -0.93 -11.4% 30.36 -23.10 -76.1% Armenia 0.13 2.11 9.95 14.55 14.84 17.80 2.96 19.9% 8.32 9.48 114.0% 0.29 2.0% 6.69 8.15 121.9% Azerbaijan .. .. .. .. .. 2.09 Belarus .. 1.05 1.60 1.14 0.80 1.60 0.80 100.6% 1.15 0.45 39.3% -0.34 -30.2% 1.26 -0.47 -37.0% Burkina Faso 9.54 8.26 9.82 10.71 11.28 10.91 -0.37 -3.3% 9.92 0.99 10.0% 0.57 5.3% 9.58 1.70 17.7% Burundi 18.92 18.81 23.50 22.18 28.92 23.22 -5.70 -19.7% 22.47 0.75 3.4% 6.74 30.4% 20.85 8.07 38.7% Cambodia 10.13 11.97 11.44 16.27 12.19 9.88 -2.32 -19.0% 12.40 -2.52 -20.4% -4.08 -25.1% 12.45 -0.26 -2.1% Central African Rep 11.66 11.38 8.04 6.89 5.75 4.30 -1.45 -25.2% 8.74 -4.44 -50.8% -1.14 -16.5% 9.49 -3.74 -39.4% Comoros76 30.23 30.16 41.54 45.96 36.11 48.84 12.73 35.2% 36.80 12.04 32.7% -9.85 -21.4% 36.97 -0.86 -2.3% Comoros99 20.78 17.86 16.77 12.84 16.29 9.61 -6.68 -41.0% 16.91 -7.30 -43.2% 3.45 26.9% 17.06 -0.77 -4.5% Congo-B 4.42 7.41 23.89 10.17 29.50 16.24 -13.26 -45.0% 15.08 1.16 7.7% 19.34 190.2% 11.47 18.03 157.2% Fiji 3.03 2.98 2.72 2.87 3.36 3.11 -0.25 -7.4% 2.99 0.12 4.0% 0.49 16.9% 2.90 0.46 15.8% Gambia 36.92 33.37 32.85 31.80 23.17 19.25 -3.91 -16.9% 31.62 -12.37 -39.1% -8.64 -27.2% 33.74 -10.57 -31.3% Ghana 2.24 2.84 3.07 4.20 4.31 3.42 -0.89 -20.6% 3.33 0.09 2.8% 0.11 2.7% 3.09 1.22 39.6% Guinea-B 49.82 24.94 39.54 32.36 30.80 64.58 33.78 109.7% 35.49 29.09 82.0% -1.56 -4.8% 36.67 -5.86 -16.0% Guyana 2.77 2.95 2.10 3.87 2.74 5.76 3.01 109.8% 2.89 2.87 99.5% -1.12 -29.0% 2.92 -0.18 -6.1% Haiti 91 8.13 9.95 6.46 7.88 5.90 5.89 -0.01 -0.2% 7.66 -1.78 -23.2% -1.98 -25.1% 8.10 -2.20 -27.2% Haiti99 25.20 24.75 12.04 10.04 10.74 6.23 -4.51 -42.0% 16.55 -10.33 -62.4% 0.70 6.9% 18.01 -7.27 -40.4% Iran 0.15 0.13 0.10 0.10 0.10 0.12 0.02 19.0% 0.12 0.00 0.3% 0.00 -2.7% 0.12 -0.02 -18.8% Niger 16.15 15.41 21.38 24.33 14.81 12.85 -1.96 -13.2% 18.42 -5.57 -30.2% -9.52 -39.1% 19.32 -4.51 -23.3% Nigeria 0.11 0.05 0.06 0.07 0.07 0.14 0.06 90.0% 0.07 0.06 89.2% 0.00 6.7% 0.07 0.00 -0.6% Pakistan77 3.25 4.34 4.91 5.75 7.46 3.70 -3.75 -50.3% 5.14 -1.44 -28.0% 1.71 29.8% 4.56 2.89 63.4% Pakistan99 3.08 1.34 1.40 0.96 1.71 1.17 -0.53 -31.3% 1.70 -0.52 -30.9% 0.75 77.5% 1.70 0.01 0.7% Peru 1.28 2.48 1.74 1.56 1.89 1.16 -0.73 -38.6% 1.79 -0.63 -35.1% 0.33 21.4% 1.76 0.13 7.2% Sierra Leone 23.51 30.08 34.29 24.25 20.07 14.18 -5.90 -29.4% 26.44 -12.26 -46.4% -4.17 -17.2% 28.03 -7.96 -28.4% Sudan 6.42 9.42 6.06 4.44 6.48 4.93 -1.55 -24.0% 6.57 -1.64 -25.0% 2.04 46.0% 6.59 -0.10 -1.6% Thailand 0.85 0.66 0.57 0.52 0.57 0.99 0.43 74.8% 0.63 0.36 57.1% 0.05 9.9% 0.65 -0.08 -12.3% Turkey 0.12 0.23 0.15 0.26 0.66 1.33 0.67 102.0% 0.28 1.04 365.2% 0.39 150.8% 0.19 0.46 241.6% Uganda 9.16 10.24 6.11 6.23 4.54 5.18 0.64 14.0% 7.26 -2.08 -28.6% -1.69 -27.1% 7.93 -3.39 -42.8% Zimbabwe 2.59 2.77 4.80 4.28 3.70 4.45 0.75 20.3% 3.63 0.82 22.6% -0.58 -13.5% 3.61 0.09 2.5% Average 11.80 11.97 12.11 10.85 10.55 10.34 0.07 0.09 11.38 -0.76 0.14 -0.30 0.11 11.59 -1.04 0.10 Median 7.28 8.26 6.46 6.89 6.48 5.47 -0.37 -0.07 7.66 0.00 0.00 0.00 0.02 7.93 -0.10 -0.02 188  Chapter 6: The Military and Democratic Reversion 189  Introduction Civilian control over the military is a widely discussed area in the democratization research with almost all researchers agreeing that at any point in the democratization process, democracy is better served with increased civilian control of the military (Foster 2005, 91; see also, Needler 1975; das Gupta 1978; Pion-Berlin and Trinkunas 2005; Rice 1992; Foster 2005; Kohn, 1997).  There is little doubt that the military is one of the most important actors in the democratization process and one of the most common assumptions in the literature is that countries with less civilian control over the military are more likely to experience a democratic reversion than countries with more civilian control.  This is thought to be the case because the military is often the instrument, if not also the instigator, of the events surrounding a democratic reversion.  The argument is that civi