Empirical indicators of crisis phase in the Middle East, 1979-1995.

A number of studies of crisis behavior assume that political behavior goes through a series of clear phases characterized by distinct patterns of interactions. Correct identification of these phases is important in crisis forecasting and in the application of mediation techniques such as preventive diplomacy. To date, empirical work with these data sets has identified crisis phases contextually (by human coders) rather than through any systematic procedures. This study uses several statistical techniques to identify and analyze phases in an event data set measuring the political behavior between eight Middle Eastern actors for the period July 1979 to June 1995, and concludes with observations on how these analytic approaches might be applied to the problem of crisis early warning.

In recent years, the topic of early warning -- moribund for about a decade after substantial research funded by the U.S. Defense Advanced Research Projects Agency (DARPA) in the late 1970s (e.g., Singer and Wallace 1979; Choucri and Robinson 1979; Hopple, Andriole, and Freedy 1984) -- has received renewed attention in the international relations literature (Rupesinghe and Kuroda 1992; Gurr and Harff 1994, 1996). With the end of the perceived threat of communist exploitation of ethnic divisions, the liberal-democratic military powers -- the United States, Britain, and France -- are less inclined to intervene unilaterally in local or regional disputes. The international community has instead increasingly relied on multilateral responses, including the recycling of cold war organizations (NATO in the former Yugoslavia and the United Nations generally), ad hoc initiatives (Iraq-Kuwait, Rwanda, Bosnia), and the use of existing nonmilitary organizations in a peacekeeping role (ECOWAS in Liberia).

This dependence on multilateral responses enhances the attractiveness of early warning in two ways. First, there is general agreement (Cahill 1996; Crocker and Hampson 1996; Lund 1996; Schmeidl 1997) that a conflict in its early stages can often be contained by either limited force or diplomacy backed with the threat of force or other international sanctions. Second, multilateral actions require a substantially longer period of time to orchestrate than did the rapid responses of a superpower or a cold war alliance. This has led to significant interest by international organizations in early warning (e.g., Boutros-Ghali 1994; Dedring 1994; Alker, Gurr, and Rupesinghe 1995; Mizuno 1995).

In addition, improvements in communication and computer technologies have changed dramatically the quantity and timeliness of the information available for use in early warning. Material relevant to political early warning is available from the commercial efforts of Reuters, Agence France Press, and other news agencies, and from the equally vast, if more specialized, networks of intergovernmental and nongovernmental organization (IGO and NGO) fieldworkers such as the UN Department of Humanitarian Affairs Reliefweb site (King 1996; http: // www.reliefweb.int). The Internet and data providers such as NEXIS provide a quantity of real-time information far exceeding that available to the Central Intelligence Agency (CIA) and KGB during most of the cold war period.(1) Similarly, inexpensive desktop computers now surpass in speed and memory most of the computers available to national intelligence agencies until the mid-1980s. These machines can process text-based electronic communications of news organizations, IGOs, and NGOs directly without the delays caused by labor-intensive human event coding. Whether this massive quantity of information can be effectively analyzed is another issue -- this is the crux of the early warning challenge -- but a researcher working with public domain sources in the late 1990s has access to dramatically more real-time information and data-processing capability than would have been the case even a decade earlier.

In conjunction with this increased interest in early warning, researchers are also using a more sophisticated model of crisis than that employed earlier. Most of the quantitative DARPA studies worked with continuous indicators of crisis. For example, the DARPA-funded Early Warning and Monitoring System (EWAMS)

provides a comprehensive profile for the particular [dyadic] relationship in

terms of conflict and other probabilities (for monthly data) and indicator

readings (for total activity and other standard system indicators in their

raw and transformed or Z-scores versions, for daily, weekly, monthly,

quarterly or yearly time intervals selected by the user). (Hopple 1984, 52)

In contrast, a number of contemporary studies of crises assume that political behaviors go through a series of phases that are qualitatively delineated by an emphasis on different sets of behavior. In the statistical literature, crisis phase has been coded explicitly in the Butterworth international dispute resolution data set (Butterworth 1976), CASCON (Bloomfield and Moulton 1989, 1997) and SHERFACS (Sherman and Neack 1993).(2) Describing the early CASCON work, Sherman and Neack explain that

conflict is seen "as a sequence of phases." Movement from phase to phase in

a conflict occurs as "the factors interact in such a way as to push the

conflict ultimately across a series of thresholds toward or away from

violence" (Bloomfield and Leiss 1969). Characteristics of disputes can be

visualized as the timing and sequencing of movement between and among phases.

Processes of escalation of violence, resolution or amelioration of the

seriousness (threat of violence-hostilities) and settlement are identifiable

through the use of phase structures. (Sherman and Neack 1993, 90)

CASCON and SHERFACS, for instance, code six phases: dispute phase, conflict phase, hostilities phase, posthostilities conflict phase, posthostilities dispute phase, and settlement phase.

In the policy literature, crisis phase has emerged as a key aspect of the preventive diplomacy concept because of the assumption that diplomacy can be more effective in the early stages of a crisis (e.g., before the outbreak of military hostility) than in later periods (Rupesinghe and Kuroda 1992; Lund 1996; Bloomfield and Moulton 1997). To illustrate this argument, Lund (1996, 38-39) outlined a series of crisis phases ranging from durable peace to war and emphasized the importance of preventive diplomacy during the unstable peace phase. In situations where preventive diplomacy is not an option, crisis phase may still be of utility in providing early warning of, for instance, large-scale refugee movements. Depending on the crisis phase, a localized outbreak of military action may be contained without generating large numbers of refugees, or it might rapidly spread, requiring the need for an international response. Finally, much of the literature on ethnic conflict assumes that militarized ethnic disputes such as those found in the former Yugoslavia, the former Soviet Union, Rwanda, Sri Lanka, and elsewhere evolve through a series of relatively predictable phases (Alker et al. 1995; Leatherman and Vayrynen 1995).

The crisis phases identified in the Butterworth, CASCON, and SHERFACS data sets have all been assigned retrospectively by human coders. Although this type of coding is obviously necessary in the early stages of a new concept's development, it presents two problems. First, when the coding of a crisis phase is dependent on human judgment, the de facto definition of the phase is likely to drift over time. This can happen as a single coder becomes more familiar with the data and is also likely during attempts to transfer the definition of a crisis phase across projects. Consequently, the crisis phases coded in two different data sets may appear to have conflicting implications because the coders were, in fact, working with disparate definitions. In contrast, the statistical identification of phases -- combined with the machine coding of event data (Gerner et al. 1994) -- should make it possible to code a crisis phase consistently and efficiently within a variety of contexts and from an assortment of different news sources.

Second, the tendency of human analysts to impose order on political events means that in some instances human coders may identify phases that do not correspond to information reported in the data set. If the human coder correctly identifies the phase, but unconsciously makes that assessment based on exogenous knowledge rather than the variables in the data set, any model that attempts to predict the phase using those data will be subject to specification error. Conversely, if the human coder has incorrectly identified the phase, any statistical estimates made with the data will be biased. We suspect that human-coded phase identification contains both types of error.

This study identifies and analyzes the phase structure of political events involving Egypt, Israel, Jordan, Lebanon, the Palestinians, Syria, the United States, and the Soviet Union/Russia for the period from 1979 to 1995 using event data.(3) This region -- and the event data set describing it -- contains several interlinked disputes. The two dominant political themes have been the Israeli-Palestinian conflict and the Lebanese civil war, both of which have gone through phases of hostility and mediation. In addition, there were other key focal points, such as U.S. efforts to resolve the larger Arab-Israeli …