email icon Email this citation

CIAO DATE: 11/00

The Global Trend of Civil Politics: The World Handbook IV Project

J. Craig Jenkins, Zeynep Benderlioglu and Charles Lewis Taylor

International Studies Association
41st Annual Convention
Los Angeles, CA
March 14-18, 2000


Contemporary discussions of global civil politics provide wildly different pictures of contemporary trends. Some see the end of the Cold War as unleashing a series of ethnic wars, geno/politicides, state breakdowns, rising criminality and a gradual descent into civil chaos (Rosenau 1990; Crevald 1991; Kaplan 2000). Others proclaim an "end of history" with the spread of democracy, capitalism, social peace and growing civility (Fukuyama 1991). Paralleling the latter, some have advanced a “social movement society” thesis (Meyer and Tarrow 1998) arguing that the globalization of civil society has internationalized social movements and protest, creating new transnational social movements and the creation of a truly global civil society. In contrast, others see a narrowing of the civil order with new mini-nationalisms, spreading ethnic intolerance and the rise of new civilizational struggles (Huntington 1996).

At the core of these concerns is mass political conflict , i.e. contests for political power that entail popular mobilization and state response to extra-institutional collective action. The major aim of the World Handbook IV Project is to construct a series of cross-national indicators of mass political conflict that can be used to gauge world-wide trends in social protest, civil contention and state response in terms of repression and facilitation to these trends. It has been almost two decades since that last edition of The World Handbook (Taylor and Jodice 1983; hereafter World Handbook III ), which has proven its value as one of the premier data sources about international political change (McGowan et al. 1988). This paper reviews the basic aims of the World Handbook IV Project by addressing: (1) the basic forms of contentious events and related summary measures of contentiousness, coerciveness and violence that are being used to map mass conflict and political change; (2) the refined agent code that is being used to map the rise of new civil actors and the internationalization of social protest; (3) recent experience with coding reliability using the automated coding of the Virtual Research Associates Knowledge Manager; and (4) a series of illustrative case studies of the new data drawing the social protests and democratizations in Eastern Europe during the 1980s and early 1990s.


Mapping Mass Conflict and Political Change

Mass political conflict has traditionally been studied in terms of its most dramatic outcome or byproduct, namely, violence. Most mass conflict, however, is non-violent and typically occurs in the context of institutionalized civil society. State officials are generally the main sources of violence, which means that violence is often a better index of state repression than of mass conflict. As outlined in our earlier work (Bond et al. 1997), we conceptualize mass conflict in terms of three major dimensions: (1) contentiousness (or extra-institutionality); (2) coerciveness (or the intensity of negative sanctions); and (3) violent outcomes. Drawing on the IDEA Project (Taylor and Bond 2000), which is currently using a Delphi technique to develop a comprehensive set of rankings of all political events in terms of their contentiousness, coerciveness and conflict/cooperation, we are developing a set of indicators for gauging these dimensions of civil politics for all countries in the world as well as a standardized set of mass conflict and state repression/relaxation events. Our plan is to make these available for secondary analysis and for early warning policy purposes on a real time basis using the automated coding capabilities of the Virtual Research Associates Knowledge-Manager system (Bond and Bond 1999).

Our concern here is to outline the general approach and provide an inventory of the event forms, which have been the focus of this work. The convention in the study of political events has been to focus on counts of specific designated forms, assuming that the key is accounting for, e.g. the frequency or participation level of protest demonstrations. While there is significant merit to this tradition, it is also critical to gauge overall system properties as these change over time. We therefore construct a series of indicators using the proportions of civil or intra-state events that are contentious, coercive or entail violence. This requires normalizing such events against the total range of civil or intra-state interactions. Thus we code the full range of civil events, including institutionalized and non-coercive and non-violent political interactions. We also embed these within a standard set of international or inter-state interactions, thus providing the basis for gauging the internationalization of social protest and the development of a transnational civil society.

Two rationales guide this approach. First and foremost, proportional measures are useful as early warning tools for gauging the stability of political systems. In earlier work (Jenkins and Bond 1999), we have shown how contentiousness and coerciveness can be used to construct an overall indicator of conflict carrying capacity, i.e. the ability of states to regulate and control intense mass conflicts. As polities move towards political breakdowns, their conflict carrying capacity drops below threshold values that forewarn (or postwarn) periods of intense violence and civil instability. Available on a real time basis, such indicators should be useful to policy-makers as well as scholars. Second, such system indicators have reliability advantages. Traditional event counts constructed from news sources have an inherent “tip of the iceberg” problem (see McCarthy et al. 1996) in that they display only political events deemed “newsworthy” and publishable by news editors and organizations. Thus there is always the problem of potential selection bias from the real world of political events, leading to potential distortion of the actual pattern of political events. One solution, infeasible at this point in the development of automation for large-scale coverage, is coding of multiple local and international news sources (cf. Rucht and Ohlemacher 1994; Olzak et al. 1998; Sommer and Scarritt 2000). Yet this does not rule out the possibility of systematic bias on the part of all local and international media, which is highly likely in repressive systems. A second solution is to normalize events of interest against the range of all political events, making the assumption that report bias is constant across event forms. Although we suspect that there is a report bias towards contentious and violent events, especially those that are unusual or dramatic in the context and therefore “newsworthy,” this seems a risk worth taking. It also facilitates the automated coding of one highly regarded international news source, which we draw upon here, namely, Reuters Business Briefs (or RBB), which contains over five times the story counts of Reuters North America and Europe (REUNA), which has been the international news source of choice for most recent international event studies.

In earlier work (Bond et al. 1997), we argued that there are four major political system indicators: (1) contentiousness (i.e. the ratio of contentious actions to total actions); (2) coerciveness (i.e. the ratio of coercive actions to total actions); (3) violence (i.e. the ratio of violent to total actions); and (4) conflict carrying capacity (or CCC). The last is a multiplicative product of three terms: civil contentiousness; state contentiousness (or repressiveness); and the proportion of violent actions in the system (regardless of the actor). The argument underlying the latter is that political systems can experience extensive contention on the civil and/or the state side without instability but, when both of these are high and combined with significant violence, the likelihood of system breakdown and instability is greater. In a comparative study of the U.S., China, Mexico, Sri Lanka, Algeria and Peru (Jenkins and Bond 1999), we show that institutionalized democratic and authoritarian regimes can weather significant upsurges of civil contention and state repression without system breakdown but when contentiousness on both civil and state sides is combined with significant violence, the likelihood of political breakdown increases. Conflict carrying capacity can also be used as an indicator of political restabilization, tracing increases in political stability generated by reductions in CCC.

Lying behind these dimensions is a comprehensive set of political events. Following conventional conceptions, we define an event as having: (1) an actor or agent; (2) a target or object of action; and (3) a specific event form. At the core is the event form. Drawing on the World Handbook III (Taylor and Jodice 1983) and other event data projects (Olzak et al. 1998; Davis and Moore 1995), we code 21 major forms of civil contention as outlined in Table 1. This captures the major contentious event forms traditionally studied by conflict researchers and, supplemented by the comprehensive mapping of all political events in the IDEA Project (Taylor and Bond 2000), will provide a basis for constructing a broad set of conflict and cooperation indicators.


Table 1: Forms of Contentious Politics


This protocol offers several advantages. First it taps several forms of civil contention previously neglected. Take for example “political flight”, which we define as attempts to cross international borders for political reasons. As Mueller (1999) argues, standard approaches have neglected this hybrid form of contention, which combines aspects of “exit” and “voice” (Hirschman 1979) by expressing opposition to the system combined with political withdrawl. Political flight was central to the anti-system protests in Eastern Europe during the 1980s and the government transitions and may be of increasing importance as the international systems becomes more respective of rights to dissent and protective of human rights, including the right of refuge from state repression. Second, we expand upon the broad category conventionally defined as “protest demonstrations” by distinguishing demonstrations from petitions and other informational protests, ceremonial protests, altruistic events (e.g. self-immolation), and protest speeches. Following Tarrow (1998: 95-105), we also distinguish three forms of disruptive protests: obstructive protests, such as building occupations and sit-ins; strikes and work slow-downs; and boycotts. Disruptive protests attempt to disrupt the on-going operation of institutions and thus create uncertainty for targets, thus pressuring them to modify practices. As Turner (1969) points out, disruptive protest combines an attempt to persuade the target, thus presupposing a common moral framework, with a high level of contentiousness, thus coercing targets to comply. On the state side, we distinguish political relaxations from repression and, within each, multiple forms of action. In our work on the East European democratizations (Jenkins and Benderlioglu 2000), the ratio of relaxations to repression provides an index of political opportunities which facilitated the development of protest and the breakup of these state socialist regimes. We also monitor attempts at reconciliation and peacekeeping, including truces, cease-fires, bargaining sessions, “truth commissions” and similar demobilization of conflict events.

Second, World Handbook IV will also tap important attributes of mass conflict events. Protests are often campaigned, occurring simultaneously in multiple locations. In addition to recording multiple actors (up to three) and multiple targets (up to three), we also identify events that are part of an interactive sequence or transformed from earlier event forms. In World Handbook III , “protest-into-riot” was the only such transformed event but, with refined coding methods, we think a significant portion of events will fit this category. Following World Handbook III , we treat violence as an outcome or an attribute of events. News sources often report alleged and planned events, which have not yet occurred. Analysts have varied as to whether to include such as “real world” events. We resolve this by including “alleged” and “planned” as attributes of events, leaving to the user the decision as to whether these should be incorporated into formal analyses. We are also developing automated methods for capturing the participation size of mass protest events, death estimates and the duration of protest events in conventional 24-hour clock days.

Third, we also code the location of actors and targets. If we are witnessing an internationalization of protest, an increasing portion should have an international component, either an actor located outside the territorial boundaries of its target (i.e. an international protest, such as the Cuban-American and Cuban national protests over the discussions over granting asylum status to Elian Gonzalez), a transnational social movement group (e.g. Amnesty International), or a national actor targeting a transnational institution, such as the World Court (i.e. a transnational protest). A significant share should also involve transnational social movement actors, who are no longer circumscribed by traditional state boundaries (Tarrow 1999).

A fourth area is the listing of actors and targets. Past work has typically used an overly simple treatment of actors and targets. Although our experience to date is that “lead” coding provides only limited information about actors and targets, we are developing an automated “look up” system to generate the formal names of political groups, new ethnicities and other political actors. With suitable refinements, we expect this system will generate greater detail on actors and targets than has traditionally been the case for large N projects. World Handbook III , for example, provided actor information on less than half of all events and, using an 8 actor scheme, lacked formal name and country-specific information. Table 2 (see appendix) provides an illustrative listing of the standardized agents we currently use.


Automated Coding and Reliability

The use of automated coding methods presents new issues for the evaluation of reliability. In the past, discussions have focused on inter-coder reliability (chiefly the problem of developing a coherent codebook and securing reasonable agreement among multiple independent coders). Recently, analysts have explored multiple sources, assessing the limitations of single sources and constructing integrative datasets that draw on multiple sources. Automation poses a different set of challenges that center on a variant of the traditional inter-coder reliability problem, namely, matching machine code against that generated by human coders.

Although ultimately the reliability of machine coding depends on precisely this match, our experience indicates that machine coding has advantages beyond the obvious one of scale, namely, training human coders to be more transparent and consistent about the rules that govern their coding. Machines are unforgiving, which makes them deadly accurate and highly reliable if not necessarily valid. This also means that in our programming, we must be fully explicit and cannot rely on the tacit inferences that humans routinely take for granted in their coding decisions. Thus, where a human coder might infer the existence of an event or an actor from partial or incomplete grammatical construction, the machine is unforgiving, requiring full and explicit instructions. All such inferences must be made explicit or else the machine will refuse to recognize an event or an attribute of an event. Our experience is that this forces human coding to be more explicit and consistent, taking greater care with grammatical constructions, as we discuss below. Hence our experience is that machine coding forcing human coding to be more precise and explicit, reducing the ambiguity that has traditionally characterized content coding of various types.

There are major advantages of automated coding. Quite obvious is the advantage of scale. Human coding projects are time-consuming, expensive and inflexible. Changes in coding rules that develop in the middle of a coding project rarely are programmed backward into standing data because of the immense costs involved. A human coded data set is generally fixed and inflexible while automation means that programming can be instituted and then, at relatively small cost, the data set recreated. In a nutshell, automation entails substituting high-skilled labor in computer programming and data evaluation for the traditional army of undergraduate and graduate student coders whose turnover and unreliability made large-scale coding unfeasible.

How do we evaluate reliability? Our basic method is an extension of traditional inter-coder reliability. First, we use human coding to establish a baseline, which is then treated as “base.” Through conventional inter-coder reliability tests, we test the validity of this “base.” Second we compare “base” against the output of the machine, focusing on both event recognition and the identification of various attributes. Through extensive iterations, we eventually train machine and human to converge. Third, we apply the machine to randomly selected leads that vary across time and space and then compare machine code against independent human coding of these texts.

Reliability testing using the VRA Knowledge-Manager is currently underway. Here we discuss briefly results to date. To date we have done only a first pass reliability check on 10 of the 21 contentious event forms using only main clause coding of RBB leads. By “main clause,” we mean coding of the noun-verb-object, which constitutes the core of the lead sentence. Our experience is that given the hierarchical structure of Reuters’ leads, coding of 2 nd and 3 rd sentences produces only redundant events already mentioned in the first complete sentence. Using basic event recognition as the criterion, our initial convergence ranged between 0 and 100 percent. The 10 event forms were: protest demonstrations; strikes; boycotts; petitions; political arrests; police surveillance and harassment; censorship; crowd control events; administrative and judicial sanctions; and bombings. These range from a low of 0% convergence for police surveillance and harassment to 100% “hits” for bombings and strikes and boycotts. The “hit” rate appears to be largely a function of the grammatical complexity of leads. As well as positive “hit” rates, we are also testing “false positives” to insure that the machine does not incorrectly identify events. With continued tweaking of the parser, we feel confident these can be raised to 80-90% reliability. VRA-Knowledge Manager uses frame parsing as its basic method (for a fuller discussion, see Bond and Bond 1999). This means that it infers the meaning of the main noun-verb pairing by reviewing logical grammatical combinations and selecting among these based on standardized rules. By iterative tweaking of the parser, we gradually train the machine to identify word patterns in a consistent and accurate fashion.

We are currently developing coding of secondary events from nouns, adjectives and adverbs. We have found that humans recognize as many as half of all contentious events from the secondary or non-main clause expressions. Thus “marching students were arrested by police” is seen as identifying two events: the main clause based “arrest by police”; and the “marching,” which constitutes a secondary event. Secondary events are recognizable through nouns (e.g. “students attended the rally ”), adjectives (e.g. “ marching students”) and adverbs (e.g. “the students blocked traffic while protesting”). We are currently implementing a comprehensive set of secondary terms that Reuters uses to refer to contentious events. With suitable tweaking and checks for duplicated events (e.g. “the marching students paraded through the village streets” constitutes a duplicate in that “ marching” and “ paraded” are both protest demonstrations), we anticipate reliabilities of 80 percent or better for basic event recognition. To date, we do not have experience with the reliability of event attributes but expect these to be comparable.

Ultimately the primary advantages of automated coding are scale, flexibility, transparency and near real-time availability. Large-scale human coding projects of the scale involved in World Handbook III can no longer be funded given the limited resources of conventional funding agencies. Using late 1970s prices, World Handbook III cost over $1.2 million. This is not feasible with NSF or other conventional funding routes. With automated coding, the major product is no longer a standing data set, which we consider “dinosaurs,” but rather the computer protocol which defines the rules linking word phrases to coded events. This is a flexible system, allowing errors to corrected and data recreated by relatively inexpensive reruns of computer routines. Automated coding is also more transparent than human coding. No matter how extensively documented the human coding project, analysts are always uncertain exactly what inferential rules the army of human coders actually used when doing their coding. Random reliability tests may reduce the range of discretion but does not eliminate the “black box” problem. Under automation, coding rules are fully transparent, i.e. logically defined by the machine which does only what is requested by the program, and highly consistent. Data can also be created on a near "real time" basis, which is strategic for policy purposes and makes scholarly work more timely and relevant. We think that with suitable refinement, coding reliabilities for basic event recognition will be in the 80-90 percent range and for event attributes in the 70 to 85 percent range.

A valid concern is the validity of relying on only a single international news source. There is not yet a clear method for integrating automated coding from multiple electronic sources. Human coding projects provide invaluable information about the comprehensiveness of various news sources, including REUNA (Sommer and Scarritt 2000). One of our collaborators, Doug Bond, is currently involved with a semi-automated field reporting project, which will provide an additional test of inter-source reliability by comparing field reports with coding of RBB. One route will be to compare automated coding of international and local sources. Our suspicion is that for civil contention there will be a bias towards reportage of large size and violent events with smaller and less contentious events received lower coverage rates (see McCarthy et al. 1996). There may also be other discrepancies that will guide us in developing the capacity for integrative data sets that combine international with local sources. All we can say is that future work will have to develop a suitable methodology for assessing multiple sources and the potential for integrating these together into combined data sets

An Illustration: Civil Contention and Democratization in East Europe

The payoff of this effort can be partially gauged by our work tracing the development of mass protest in Eastern Europe in the 1980s and its impact on democratization. Here we provide a snapshot overview of this work. Figures 1-5 trace the development of political protest and political opportunities between 1984 and 1994. These estimates are based on human coding of REUNA leads, which are currently being used to assess the reliability of the VRA Knowledge-Manager. They illustrate the basic ideas that can be tested with these data, specifically the argument that increases in political opportunities facilitate the development of mass protest. Political opportunities are gauged by the ratio of state relaxations to repression using counts of the event forms in Table 1. For Poland, Czechoslovakia and Hungary, a strong year-to-year correlation (ranging between .57 to .36) between 1984 and 1990 indicates that increased opportunities facilitated the mobilization of protest. These protests then played a central role in spurring the democratization of these regimes, creating a wave of interrelated “velvet glove” revolutions that transformed all of the former Soviet client states. Romania and Bulgaria did not display this close relationship between political opportunities and protest, reflecting the strong resistance of these neo-patrimonial Communist regimes to democratization. In these two countries, there was little political relaxation prior to 1989, meaning that the there was little political space for protest, minimal mass protest and, when the regime transition finally arrived, it was triggered more intra-elite maneuvering (a violent coup in Romania and a party purge in Bulgaria) and international pressures from outside. This created an important difference in the nature and stability of these post-Communist regimes with Romania and Bulgaria experiencing protracted protest, significant mass violence and an extended period of civil contention over the nature of the post-Communist regime.








This paper has reviewed recent developments in the development of the World Handbook IV Project, focusing on the development of automated coding using the VRA Knowledge-Manager and the new event forms that will be included in World Handbook IV. It will probably be two to three years before we have high quality, reliable event estimates and related indicators publicly available for secondary analysis. There is considerable work still to be done to test out the reliability of these data and their match with other sources. Nonetheless, the VRA Knowledge-Manager has demonstrated its value as a source of automated coding, improving on traditional sparse parsing methodology (Gerner et al. 1994; Schrodt and Gerner 1995). The advent of this and other systems for automated coding of textual information will eventually replace human coding of large amounts of event data. These systems are more transparent, flexible, and efficient (once developed). At some point, we think a technical solution can eventually be developed for multi-source coding, possibly assisting by human resolution. In any case, the days of the large fixed event data set generated solely by human coding are near a close. Automated coding also has advantages for improving human coding. By forcing analysts to be grammatically transparent about their inferential rules for assigning meanings, automated coding also reduces the “black box” problem that has long bedeviled human content coding.

World Handbook IV should provide us with the information for evaluating various arguments about contemporary trends in civil politics throughout the globe and the impact of civil contention on democratization, the internationalization of social protest, and the breakdown and reconstruction of political systems. We have sketched out various subprojects associated with the World Handbook IV effort, using these to illustrate the potential of these efforts. With further refinement, we should have the ability to gauge the growth of civil contention and its impact on political change and stability.



Bond, Doug, J. Craig Jenkins, Kurt Schock and Charles Taylor. 1997 "Contours of Mass Political Conflict: Issues and Prospects of the Automated Development of Conflict Event Data.” Journal of Conflict Resolution 41:553-579.

Bond, Doug and Joe Bond. 1999. VRA Knowledge-Manager: A Handbook. Weston, MA: VRA.

Creveld, Martin. 1991. The Transformation of War . NY: Free Press.

Davis, David R. and Will H. Moore 1995. International Political Interactions: Codebook . Dept. of Political Science, Emory University and Florida State University.

Fukuyama, Francis. 1991. The End of History and the Last Man . NY: Free Press.

Gerner, Deborah, Philip Schrodt, Ronald Francisco and Judith Weddle. 1994. "Machine Coding of Events Using Regional and International Sources." International Studies Quarterly 3 7:91-119.

Hirschmann, Albert. 1979. Exit, Voice and Loyalty . Cambridge, Ma.: Harvard University Press.

Howell, LD. and G. Barnes. 1994. "Event-Data for Region-Specific Interactions: A Research Note on Source Coverage." In R.L. Merritt, R.G. Muncaster and D.A. Zinnes (eds.) International Event-Data Developments . Ann Arbor: University of Michigan Press.

Huntington, Samuel. 1996. The Clash of Civilizations and the Remaking of the World Order . NY: Simon and Schuster.

Jenkins, J. Craig and Doug Bond. 1999. “Conflict Carrying Capacity, Political Crisis and Political Reconstruction.” Unpublished paper presented at International Studies Association meetings, February 1999, Washington DC.

Jenkins, J. Craig and Zeynep Benderlioglu. 2000. “Mass Protest and the Velvet Glove Revolutions in Eastern Europe, 1984-1994.” Unpublished paper, Dept of Sociology, Ohio State University, Columbus OH.

Jodice, David A., Charles Lewis Taylor and Karl W. Deutsch. 1980. Cumulation in Social Science Data Archiving . Berlin, W. Germany: International Institute for Comparative Social Research.

Kaplan, Robert. 1999. The Coming Anarchy . Boston, Ma.: Little and Brown.

McGowan, Pat, Harvey Starr, Gretchen Hower, Richard Merritt and Dina Zinnes. 1988. "International Data as a National Resource." International Interactions 14: 101-113.

McCarthy, John, Jackie Smith and xxx. 1996. “xxxx” American Sociological Review .

Meyer, David and Sidney Tarrow (eds.) 1999 The Social Movement Society . Lanham, Md.: Rowman and Littlefield.

Mueller, Carol. 1999. “Hybrid Exit Repertoires in a Disintegrating Leninist Regime” American Journal of Sociology 105: 697-735.

Olzak, Susan. 1989. "The Analysis of Events in the Study of Collective Action." Annual Review of Sociology 15:119-41.

Olzak, Susan, et al. 1998. “Handbook for Coding of U.S. Protest Events.” Dept. of Sociology, Stanford University.

Rosenau, James. 1990. Turbulence in World Politics . Princeton: Princeton University Press.

Rucht, Dieter and Thomas Ohlemacher. 1994. "PRODAT: Coding Handbook." Berlin, Germany: Wissenschafts Zentrum-Berlin.

Schrodt, Philip A. 1994. "Machine Coding of Event Data." Pp. 125-149 in Richard Merritt, R. Muncaster and D.A. Zinnes (eds.) International Event-Data Developments: DDIR Phase II. Ann Arbor: University of Michigan Press.

Schrodt, Philip A. 1995. Kansas Event Data System (KEDS) Lawrence, KS: Dept. of Political Science,University of Kansas.

Schrodt, Philip A. and Deborah Gerner. 1994. "Validity Assessment of Machine-Coded Event Data Set for the Middle East, 1982-1992." American Journal of Political Science 18: xxx.

Smith, Jackie. 1995 "Transnational Political Processes and the Human Rights Movement." Research in Social Movements, Conflicts and Change, Vol.18 . Greenwich, Ct.: JAI

Sommer, Henrik and James R. Scarritt. 2000 "The Utility of Reuters for Events Analysis in Area Studies: The Case of Zambia-Zimbabwe Interactions, 1982-1993." International Interactions 25:1: 1-31.

Tarrow, Sidney. 1998. Power in Movement (2 nd ed). NY: Cambridge University Press.

Tarrow, Sidney. 1999. “The Internationalization of Social Protest.” Unpublished paper, Dept. of Government, Cornell University. Ithaca, NY.

Taylor, Charles Lewis and David A. Jodice. 1983. The World Handbook of Political and Social Indicators . New Haven: Yale University Press.

Taylor, Charles Lewis and Joe Bond. 2000. “The IDEA Project.” Paper presented at International Studies Association meeting, Los Angeles CA, March 15, 2000.

Turner, Ralph. 1969. “xxxxx” Pp. xxx in Tamotsu Shibutani (ed.) Human Nature and Social Behavior. Indianapolis, In.: Bobbs-Merrill.

Wallensteen, Peter and Margareta Sollenberg. 1995. "After the Cold War: Emerging Patterns of Armed Conflict, 1989-1994." Journal of Peace Research 32:.

Zartman, I. William. 1995. Collapsed States . Boulder, Co.: Lnn Rienner.



Table 2: Actors in World Handbook IV

8, 10, 12, 14, 16, 18 Actors/Targets/Organizational Level.

02 Economic & Occupational Groups

03 Ethnic/Racial Groups

04 Gender Based Groups

05 Age Based Groups

06 Political and Other Oppositional Groups

07 Government

08 Residential Groups

09 Criminal Groups (Note: Refers only to criminal groups involved in political activities)

10 General Citizens

11 Cannot Determine/Missing

9, 11,13,15, 17, 19. Actor/Target Organizational Level

1 Individuals & Small Groups (under 10 persons)

2 Groups/Collectivities/Institutions