RESEARCH CONDUCTED THROUGH A MULTI-METHOD LENS
It is clear from the discussion that a variety of methodological approaches have been used to analyze conflict. Only a few of these studies have employed several methods. Researchers in this field have not, for the most part, taken advantage of the complementarities that exist among the various approaches to doing research.
This is evident as well in the chapters written for the Maoz et al. (2004) book on different methodological approaches to the study of mediation and conflict management. Each author illuminates the value of a particular approach used to address research questions in the field. The approaches include game theory, simulation, large-N statistical analyses, and historical case studies.3 In his concluding essay, Stoll (2004) describes the strengths and weaknesses of each approach but laments the state-of- the-art by noting that: “Research that relies on any one approach is usually inferior to research that makes use of several methods” (2004, 360). He adds “that it is a rare situation where one method is so superior that the others can safely be ignored” (2004, 360). Several multi-method studies are discussed in this section.A multi-method design trades the weaknesses of one approach for the strengths of another as these are described in this chapter and by Stoll (2004): for example, the breadth of a random-sample survey for the depth of a detailed case study. When the different approaches are used together in sequence, a more complete picture of the conflict or resolution process is achieved. This can be illustrated by two studies of peace agreements: our study of relationships between principles of justice and the durability of agreements negotiated during the late 1980s and early 1990s (Druckman and Albin, 2008) and Irmer and Druckman (2007) study of the relationship between the processes and outcomes of a variety of peace agreements.
The justice and durability study combined an aggregate analysis of 16 peace agreements with a focused comparison of selected cases. The former analysis produced a statistical model of the role played by justice in implementing the agreements: justice principles moderated the relationship between the conflict environment and implementation success. The latter analysis showed that many principles of justice increased the chances of success in difficult conflict environments. Few justice principles in an agreement decreased the chances of complete success in less difficult environments.
The qualitative results bolster the quantitative findings: both show that justice principles play a moderating role, first in the form of a statistical path model and then as a 2 ? 2 matrix of difficulty (high or low) by justice principles (many or few). Both analyses illuminate the value of control: partial correlations were used in the aggregate analysis, while case selection (the combinations of high-low, few-many of the two variables) created independent variables in the focused comparison. The complementarity is between the more robust analysis of a variety of cases and a more controlled analysis of cases selected on the theoretical variables of interest.
Irmer and Druckman's study is novel in the way it uses methods to build a logical argument that bolsters explanation. The analyses proceed in sequence from initial quantitative work to later qualitative inquiries. After constructing process and outcome indicators and assembling a data set of 26 cases, the investigator calculated a correlation among the process and outcome indices. An average correlation of.82 indicates a very strong relationship. Not considered in the analysis, however, were the effects of context: the strong association between process and outcome may be accounted for by correlated context variables. Thus, a second step consisted of controlling for context by calculating partial correlations. These correlations remained strong, indicating that process rather than context accounted for the outcomes.
The next question asked was whether this relationship was causal. By tracing the process in each of four cases (not part of the aggregate data set), they showed that outcomes - as comprehensive, partial, or deadlock - were indeed caused by the coded processes.Afinal step in the sequence entailed an attempt to discover a mechanism that could explain the causal relationship. Referred to as a plausibility probe, the analysis showed that the development of trust through phases of the talks appeared to provide a plausible explanation. The mode of inquiry used in this study - designated as A (association), C (causation), E (explanation) - may be considered a model for research on negotiation and peace processes.This study provides an example of moving beyond the use of single methods. It addresses the internal validity of statistical relationships. Earlier multi-method analyses of negotiation addressed issues of external validity. These include Hopmann and Walcott's (1977) pioneering effort to combine simulation and content analysis of transcripts in seeking to understand the behavior of arms-control negotiators. They also include the simulation/case study analyses by Beriker and Druckman (1996) on multilateral post- WWI peace talks and by Ozcelik (2004) on multilateral environmental regime talks. Similarly, my research on turning points and models of responsiveness, discussed earlier, shows how findings obtained from single-case studies can be generalized by conducting statistical analyses of many cases.
Another example of combined methods comes from a study on mediating international crises. Wilkenfeld and his colleagues (2003) explored the same research questions with two methods: the International Crisis Behavior events data set and an experimental simulation based on the Ecuador/Peru border dispute of 1981. The historical events data analyses showed that mediated crises were usually characterized by compromise among the disputants, more likely to end in agreement, and indicated a tendency toward long-term tension reduction.
The simulation findings confirmed these findings while also adding other results: mediation leads to crises of shorter duration and to greater satisfaction with the outcome. In addition, the simulation data revealed that a manipulative mediation style produced more benefits from crisis termination than a facilitation style.These findings illuminate several advantages of a multi-method approach. One concerns the validity of results: confirmation with another method increases confidence in the results. Another concerns additional discoveries: new findings extend our understanding of mediation impacts. And, a third advantage is the opportunity to explore questions that cannot be addressed with events data: information on crisis duration and on mediator style were obtained from the laboratory study. Going further, the laboratory allows for imaginative construction of conditions that did not occur in the case but have theoretical relevance as illustrated by this study and that conducted by Beriker and Druckman (1996). (See Sprinz and Wolinsky- Nahmias, 2004, for more on counter-factual analysis.)
These studies illuminate the value of multimethod research on negotiation. Conceivably, this topic is suited to these sorts of complex analyses. One reason is that there is a treasure trove of well-documented cases for content analysis.Anotheris that the processes in many of these cases are relatively easy to simulate. And, a third reason is that the literature is well developed in terms of theories, models, and empirical research. For these reasons, among others, we have learned much about processes and strategies of negotiation. Among the remaining research challenges are connecting negotiation to the sources of conflict that gave rise to the issues and to the aftermath of the implementing phase where relationships between disputing parties undergo change. These problems would also benefit from studies that view them through a multi-method lens.