WEB APPENDIX. to accompany. Veto Players and Terror. Journal of Peace Research 47(1): Joseph K. Young 1. Southern Illinois University.

Transcription

1 WEB APPENDIX to accompany Veto Players and Terror Journal of Peace Research 47(1): 1-13 Joseph K. Young 1 Departments of Political Science and Criminology/Criminal Justice Southern Illinois University and Laura Dugan Department of Criminology and Criminal Justice University of Maryland 1 This material offers extensions, sensitivity analyses, and supplements to the Journal of Peace Research article. For questions about this material, please contact Joseph K. Young, jkyoung@siu.edu.

2 Indicators Concept Indicator Coding Source(s) Veto Players Veto Number of actors whose agreement is necessary to change policy. Beck et al. (2001) Gandhi & Przeworski (2006) Veto Players Polcon Index of political Henisz (2002) constraints. Contestation Press Freedom A dichotomous measure of whether a state clearly allows freedom of the press or not Democratic Participation Opportunities for Violence Military Capacity/ Development Participation Population Gross Domestic Product For democracies, the measure includes the % voter turnout (centered), and is coded as zero for autocracies The Natural log of a country s population in millions Real Gross Domestic Product per capita in thousands History of Attacks Past terror A lagged running average of terrorist attacks since the beginning of each country series Regional Effects Region As dichotomous measure for regions of the world Conflict for the War state A dichotomous measure indicating whether the state is involved in an internal or external conflict Cold War Effect Cold War A dichotomous measure of the time period before (0) and after the Cold War (1). Stability of the Regime Durable A counter variable that begins when a state s regime changes Van Belle (1997) Li (2005) Vanhanen (2000) Li(2005) Heston et al. (2009) Heston et al. (2009) Li (2005) LaFree & Dugan (2007) Bennett & Stam (2002) Sarkees & Wayman (2010) Bennett & Stam (2002) Marshall & Jaggers (2001)

3 Summary Statistics Variable Mean Std. Dev. Min Max Homegrown fatal Attacks Homegrown Ambiguous Fatal Attacks Foreign Fatal Attacks All Fatal Attacks Veto Free Press Participation Population Development War Cold war Durable Past Terror Europe Africa Asia America

4 The content below was cut from the original submission to Journal of Peace Research for space considerations. The following figures were used to graphically illustrate the expected relationships between veto players and policy change and veto players and expected terror attacks. Figure 1 Likelihood of Policy Change Given Number of Veto Players Present in a Political System Figure 2 Likelihood of Terror Attacks Given Number of Veto Players Present in a Political System

5 Percent Fatal Terror Attacks Figure 3 Histogram of Number of Terror Attacks in a Country-Year Source: Global Terrorism Database (2007)

6 Figure 4 Comparison of predictive ability of different count models using Long and Freese s (2005) SPOST command COUNTFIT. Global Terrorism Database ITERATE Number of Terror Attacks Frequency Percent Frequency Percent % % 1 or more % % Total Table 1 Data on Terror Attacks Source: GTD and ITERATE

7 This paragraph was cut from section NEW DATA, NEW METHODS, Paragraph 2 However, we do not expect our inferences would change dramatically if our data were expanded backwards to 1945 or forward to the present. In the post World War II era, according to Hoffman (2006, 43) [terrorism] became a pervasive global force. Anti- colonial struggles characterized the immediate post- war terrorism but subsequent terror campaigns have followed similar ethno- nationalist goals of major policy change or sovereign control of territory. This was cut from section DATA, Paragraph 2 Over twenty people were responsible for coding information over the years spanned by the data collection, but only two individuals were in charge of supervising data collection during the entire period. Cut from FOOTNOTE 9 The 1993 data was lost before the Pinkerton group transferred the data to the University of Maryland. We, therefore, estimated the models using the data without 1993 and with imputed values using the program Amelia II and find similar results. The following models exclude the imputations from Amelia II. Cut from FOOTNOTE 10 The 1993 value in Figure 3 was estimated by averaging the 1992 and 1994 values. Cut from FOOTNOTE 12 This data only includes attacks from the 166 countries for which we have veto player data. Thus the overall count of attacks in the GTD is lower than that reported in other sources (LaFree and Dugan 2007). The ITERATE data, on the other hand is from all countries, thus underreporting the actual difference between the two data sources. Cut from section DATA, Paragraph Also, the State Department definition is limited to politically motivated violence. By contrast, the GTD also includes economic, religious and social objectives. 2 At first it may appear as if this definition differs from the one adopted for this research (see footnote 2), which explicitly defines terrorism as force used for political purposes. However, while the State Department might not agree, we consider acts of violence for any of the above reasons as inherently political. Clearly acts such as the Oklahoma City bombing, the Tokyo subway attack, and the 1993 World Trade Tower bombing had diverse motivations, but all three events were perpetrated to 2 An economically motivated act of violence is considered terrorism when it is perpetrated by a known terrorist organization.

8 enact some change in the behavior of the state. Whether a group intends to express religious, economic, or social preferences, the goal is to change government policy. Another difference between our definition and that provided by the data collectors is that we require that the aim of the attack is to coerce a third party who is different from the target. The GTD data has no such restriction and likely contains cases where the aim of the group is to coerce the target. While it is impossible to filter out those cases, we examine the data with and without military targets, since military targets are more likely to be the subject of conventional coercion. Another reason for filtering military from non- military targets is that including military targets confounds terrorism with insurgency. 3 Using irregular means to fight against the state does not necessarily make a group a terrorist organization. Only the targeting of third parties, usually civilians, leads to the identification of this form of violence. CUT FROM TOP of PAGE 11, 1 st Paragraph Colombia also saw a relative decrease in number of fatal and nonfatal attacks after increasing executive constraints for a period during the 1990s. After removing the constraints, attacks went back to the levels previously experienced. Whether attacks led to a reduction in executive constraints or vice versa, what is clear is that reducing executive constraints did not decrease fatal and nonfatal terror attacks in our sample. CUT FROM PAGE 21, control variables section States with high capacity are better able to avert violent challenges from dissidents. 4 However, since terrorism is often a tool designed to shift the balance of power (Lake 2002, 17) between terror group and state, terror groups often challenge militarily capable states. To proxy this concept we use the correlates of war capabilities index (CINC) 3 See O Neill (2005) for a discussion related to these conceptual distinctions. In short, terrorism is a form of violence while insurgency is a type of resistance by a dissident group against a state. An insurgency may use terror as part of their campaign, but all insurgents are not terrorists. 4 These two measures proxy for different concepts but do correlate somewhat. In our sample they correlate at 0.44.

9 CUT FROM THE CONTROL VARIABLES SECTION We do not attempt to build a general model that includes all possible covariates that might explain why we see terror. Part of the logic for including these variables is that excluding them can bias our estimates. As Clarke (2005) demonstrates, however, including them may decrease or actually increase bias. Mathematically, we cannot say for sure. Achen (1999) argues for simply using three variables while Ray (2003) advocates simply controlling for confounding variables. We mostly follow their advice and include control variables that might correlate with the number of veto players and terror events as well as some well- established predictors of terror events.

10 References Achen, Christopher Toward a New Political Methodology: Microfoundations and ART. Annual Review of Political Science 5: Bennett, Scott and Alllan C. Stam Expected Utility Generation and Data Management Program, Version 3.1. Bloom, Mia Dying to Kill. New York: Columbia University Press. Clarke, Kevin The Phantom Menace: Omitted Variable Bias in Econometric Research. Conflict Management and Peace Science 22: Ray, James Lee Explaining Interstate Conflict and War: What Should be Controlled For? Conflict Management and Peace Science 20:1 31. Greene, William H Econometric Analysis Sixth Edition. Upper Saddle River, NJ: Pearson Education, Inc. Hegre, Håvard, Ellingsen, Tanja and Gates, Scott and Gleditsch, Nils Petter. Toward a Democratic Civil Peace? Democracy, Political Change, and Civil War, American Political Science Review 95(1): Heston, Alan, Robert Summers and Bettina Aten, Penn World Table Version 6.3, Center for International Comparisons of Production, Income and Prices at the University of Pennsylvania, August O Neill, Bard Insurgency and Terrorism. Washington, DC: Potomac Books. Pape, Robert Dying to Win: The Strategic Logic of Suicide Terrorism. New York: Random House Sarkees, Meredith Reid and Frank Wayman Resort to War: CQ Press.