Performing case studies

Case studies come in several forms, ranging from the single case to a large number of cases used for making comparisons. In my earlier work, I distinguished among four approaches: the enhanced case study, time-series analysis, focused case comparisons, and aggregate or large-N case comparisons.

Enhanced case studies. This research approach consists of viewing a case through the lens of a theoretical framework. Key concepts are used to interpret the way a conflict process unfolds or how it is resolved. Many published analyses of negotiation and mediation use this approach: examples are Haskel’s (1974) use of the concept of power symmetry/asymmetry to understand Scandinavian market negotiations, Winham’s (1977) emphasis on issue complexity as the primary challenge in multilateral trade talks, and the studies by Cameron and Tomlin (2000) and Druckman et al., (1991) on turning points in NAFTA and the INF talks respectively. The books edited by Rubin (1981), Zartman, (1994), and Cohen and Westbrook (2000) provide examples of how particular cases - Kissinger’s shuttle diplomacy, European Community talks, Bronze-Age diplomacy with Egyptian pharaohs - can be viewed from a variety of theoretical perspectives, such as game, decision, coalition, or leadership theories.

My 1995 study with Green (Druckman and Green, 1995) on negotiations between the Philippines Aquino regime and the National Democratic Front shows the relevance of the concept of “ripeness” for decisions to negotiate and “formula” for opportunities discovered or squandered during the process of negotiating. Further insights into that negotiating process came from superimposing concepts from the literature on the sociology of conflict, particularly with regard to the interplay between values and interests, onto the unfolding discussions. (See also the other case studies of internal conflict in Zartman, 1995.) These enhanced case studies call attention to the relevance of theory for capturing the essential processes of complex cases. They do not, however, provide sound explanations for relationships among variables in the conflict setting, process, or outcome. Explanatory goals are addressed by the comparative methods to be discussed in the sections to follow.

Time-series analysis. This approach to analysis captures the dynamics of conflict processes. It consists of a family of techniques that analyzes a sequence of events that occur over a relatively long period of time in the context of one or more cases. The techniques include experimental before-after comparisons, correlation/regression analyses of trends, probability forecasting, and qual­itative process tracing. An example of each of these applications is provided in this section.

Before and after comparisons of trends are illustrated by our study of mediation during the conflict over Nagorno-Karabakh (Mooradian and Druckman, 1999). Nearly 4000 events from 1990-1995 were coded on a six-step scale ranging from significant action toward peace (+3) to significant violence directed at an adversary (-3).

Another attempt to explore the effects of mediation on violence is the study by Schrodt and Gerner (2004). Using various indices of cooperation and competition, they examined ten cases of mediated dyadic conflicts. For each conflict, they cross-correlated mediator behavior with the level of violence over time. This technique is useful for examining long­term, time-delayed effects of interventions. Of particular interest is their finding about the sticks-or-carrots issue: Is mediation more or less effective when mediators use rewards (including side payments) or coercive mea­sures such as sanctions? Their technically sophisticated machine-coding approach to analysis produced a complex, contingent find­ing: violence was reduced when mediation was accompanied by both coercive measures and by cooperative incentives directed at the weaker party in the dyad.

Probability forecasting with Bayesian tech­niques is illustrated by asking whether a peace agreement will hold through time. Probabilities can be estimated from historical data showing that agreements hold for at least five years in one third of the cases examined. The question is whether this is sufficient information for projecting the longevity of future agreements. The Bayesian answer is that other factors, referred to as conditional probabilities, must also be taken into account.

Qualitative time-series often takes the form of process tracing. This consists of searching an historical record of events for evidence about whether an hypothesized process did or did not occur. An example is the paths developed from cases of negotiation in my study of turning points (Druckman, 2001). Focusing on departures in trends, I traced the relationship between causes, referred to as precipitants, and consequences following the occurrence of a departure, known as a turning point. Paths were traced for each of three types of negotiations, those over security issues, environmental concerns, and trade matters. The case-specific paths were combined by discovering the most frequent precipitant (inside or outside the negotiation), departure (abrupt or non-abrupt), and consequence (escalatory or de-escalatory). A particularly interesting finding is that outside factors were needed for departures in the security while inside factors (either new ideas or procedures) were usually the cause of change for the environmental and trade cases.

An interesting application of this three- part framework is Dougherty’s (2006) study of social movements in Northern Ireland. Her detailed historical traces of the school integration movement provided insights into the factors that precipitate the transformation of social conflicts. The link between collective action and transformation was shown to turn on the development of a critical mass of key members of the movement. Other examples of the tracing of conflict dynamics are Carstarphen’s (2003) analysis of dialogues about prejudice among small groups of men from different backgrounds and Pruitt’s (2005b) chain analysis of two cases, the Northern Ireland peace process and the talks between Israel and the PLO in Oslo.

Focused case comparisons. Another enha­ncement to the traditional case study is the structured, focused case comparison. Similar to the enhanced case study discussed earlier, the focused comparison uses theory to guide case selection and analysis. Unlike the single­case study, a focused comparison consists of a matching of similar cases. It is an attempt to impose the logic of experimentation on a small number of cases. (SeeFaure’s 1994 discussion of the Most Similar Systems Design [MSSD].) One of the early examples of application is Putnam’s (1993) comparison of a northern and southern Italian province in terms of a number of performance indicators. The provinces were similar in most aspects of their political systems. They differed only in terms of economic development, with the north outpacing the south, and the development of a civic culture. These were the independent variables in the focused comparison. If, as Putnam argued, the provinces differed only on economic and civic development, then either or both of these variables can be used as an explanation for differences in performance. Using partial correlation, he showed that civic culture was the better explanation: When levels of civic cul­ture were controlled, economic development made little difference in performance.

One of these studies is Allen Nan’s (1999) analysis of coordination among conflict­resolving organizations. The research focused on processes of complementarity and coordi­nation among NGOs in three former Soviet republics: Abkhasia, South Ossetia, and Transdniestria. These cases were matched on types of disputing parties, interests, time, and power. A key difference however was whether they initiated a process of long-term unofficial facilitated joint analysis among negotiators (LUFJAAN) of the contending groups: only Abkhasia did not entertain this process. The cases differed also on the key dependent variable progress toward a cease fire. Progress occurred in South Ossetia and Transdniestra, but not in Abkhasia. Because the cases were carefully matched on other variables, she inferred that there was a relationship between the LUFJAAN process (the independent variable) and the reduction of violence toward a cease-fire agreement (the dependent variable).

Additional coding of the South Ossetia negotiations by Irmer (2003) revealed that the process consisted of more problem solving than competitive bargaining state­ments, leading to comprehensive (rather than partial) outcomes. Further analyses showed that trust between the negotiators increased through the course of the talks. These effects may have resulted from the long­term facilitation process used in this case as documented by Allen Nan. Considered together, the two studies illustrate the value of combining methods, MSSD and statement coding.

Another small-n case-study approach emp­hasizes differences rather than similarities between cases. Referred to by Faure (1994) as the Most Different Systems Design (MDSD), this approach is useful for developing typolo­gies. An example is Weiss' (2002) study of mediator sequencing strategies. Three cases of peace processes were used as exem­plars of alternative sequencing approaches: gradualism (Mozambique), boulder in the road (El Salvador), and committee (Angola). Content analyses of the negotiation process in each case confirmed the approach used and identified a number of factors that distinguished among them, for example, type of most contentious issue, type of reasoning, trust building, and ripeness in the process. These factors were used to construct a profile for each of the three types of mediator sequencing strategies.

Another MDSD study, conducted by Druckman and Lyons (2005), compared Mozambique with Nagorno Karabakh. The different profiles for these cases resembled the distinction - made by Galtung (1969) - between negative (Nagorno Karabakh) and positive (Mozambique) peace. Additional coding showed that the cases differed also on various indicators of distribu­tive (competitive) and integrative (problem­solving) bargaining. Further insights about these cases emerge when we consider findings obtained in other, related studies. Concerning Mozambique, the gradualist mediation strat­egy discovered by Weiss may have encour­aged problem-solving and forward-looking rhetoric coded by Druckman and Lyons (2005) and by Irmer (2003). Concerning Nagorno Karabakh, the 1994 war between Armenia and Azerbaijan (see Mooradian and Druckman, 1999) may have led to the distributive bargaining process and partial, cease fire outcome also coded by Irmer (2003). The larger picture for these cases emerges from studies that employed several methodologies - MDSD, content analysis, and interrupted time series. I will return to the topic of complementary methods in the sections to follow.

Aggregate case comparisons. Compara­tive case studies are also performed across a large number of different cases. Referred to as the method of concomitant variation (Faure, 1994), large-N case studies are another form of MDSD. Unlike the controlled com­parisons discussed in the previous section, these studies consist of sampling of cases from a larger universe. When carefully done, the sampling increases confidence in the generality (or external validity) of the findings. Random sampling of cases from a defined universe reduces the possibility of selection biases (or self-selection) which pose the dilemma of alternative explanations for findings (see Sprinz and Wolinsky- Nahmias, 2004, pp. 368-369). The large number of cases also increases the scope of the comparisons and permits the use of statistical techniques for evaluating relation­ships among variables. This tradition of case­based research in international relations can be traced to McClelland's (1976) analyses of the World Events Interaction Survey. In anthropology, this tradition has been strongly influenced by Murdock's (1957) World Ethnographic Sample and by Whiting and Child's (1953) cross-culture analyses of child training. It has been refined by Naroll (1962) and by the authors in Naroll and Cohen's (1973) handbook. Building on these foundations, the method has gained popularity in research on conflict management.

Parallel aggregate case research streams have been developed by Bercovitch and collaborators on international mediation and by Druckman and associates on negotiation. The former have compiled a useful data set of 333 international conflicts since 1945. A variety of coded features include infor­mation about the parties, the conflict, and the conflict management agents. Guided by a conceptual framework that connects these categories, the authors have identified con­ditions influencing mediation success. Their statistical analyses make evident contingent relationships between independent (concern­ing the dispute, the parties, the mediation process, and the mediator) and dependent (mediation success) variables. For example, short conflicts with few fatalities are more amenable to mediation than other kinds of disputes; conflicts involving parties with the same amount of power are more likely to be successfully mediated than conflicts between parties that differ in power or available resources (Bercovitch and Trappl, 2006). Further aggregate analyses of mediation showed that third parties were more likely to be used in complex (many issues) negotiations (Druckman, 1997).

The large-n analyses of international nego­tiations have also produced a number of interesting findings. A key dimension that ran through 23 negotiation cases was whether the talks were bilateral or multilateral. More treaties result from talks with a smaller number of parties (Druckman, 1997). In an analysis of 30 cases of negotiation sampled from a listing of 176 cases in the Pew Case Studies in International Affairs, Druckman et al., (1999) found that the cases were dis­tinguished in terms of a well-known typology of negotiation proposed by Ikle (1964): profiles of features differed depending on whether the talks were primarily distributive, integrative, normalizing, extension, or side effects. Another analysis examined turning points that occurred in 34 cases, distinguished in terms of issue area - security, trade, or environmental. Themost frequently occurring trigger for the security cases was an external event, such as third-party assistance; triggers of turning points were from within the talks (as substantive or procedural decisions) in the trade and environmental cases.

These differences highlight the risk-averse approaches taken by negotiators in the area of security (Druckman, 2001). Risk aversion is also highlighted in the Ember and Ember (1992) study of resource security and warfare in 186 societies (using the Human Relations Area Files). They showed that fear of others and of nature were the strongest predictors of war. Specifically, fear of the future rather than current problems of scarcity appeared to be the primary motive for war. Warfare was used for plundering resources in order to protect against an uncertain future.

Interesting findings about conflict have been obtained also from a number of other aggregate case studies. Diehl et al., (1998) multidimensional scaling of expert judgments showed that different types of peacekeeping missions clustered around two dimensions, roles and processes. Peacekeepers in some missions played primary roles (e.g. col­lective enforcement, state/nation building). In other missions (e.g. election supervi­sion, arms control verification), they are third parties. Missions characterized primarily by distributive processes include collective enforcement, sanctions enforcement, and pacification. Those where integrative or problem-solving processes are emphasized include election supervision, nation building, and observation. These findings contribute to the development of taxonomies with implications for training.

Compelling findings about the factors that lead to the initiation (but not the duration) of ethnic conflict within societies were reported by Gurr and his colleagues (2005). Their sophisticated statistical analyses pro­duced a probability (odds ratio) ordering of various drivers of internal conflict: state-led discrimination, ethnic diversity, regime type, spillover or neighborhood effects, recency of ethnic wars, and youth bulge. Similar factors were shown to drive conflict in predominantly Muslim societies. This study is notable for the careful analyses performed as well as the policy implications derived. Factionalism, more than regime type, led to a larger risk of ethnic conflict. This suggests the strategy of institutionalizing political participation in a manner that cuts across communal lines with policies aimed at easing discrimination against minorities. Another implication from the findings concerns transi­tions to democracy. The authors suggest that attempts to democratize autocratic regimes in Muslim countries increase the short-term risks of political instability. This would seem to be a result of the persisting influence of under-development and being in “bad neighborhoods.”

Another recent aggregate case analysis with policy implications was performed by Downs and Stedman (2002). They were interested in the factors that influenced the successful implementation of peace agreements nego­tiated during the early 1990s. Specifically, they examined the relative impact on the implementation outcome of the difficulty of the conflict environment and the willingness of other states to intervene in the conflict. A difficulty index was constructed as an aggregate score of eight measured factors: for example, the number of warring parties, hostile neighboring states, number of soldiers, wars of secession. A willingness index was composed of three factors - major or regional power interest, resource commitment, and acceptance of risk to soldiers. These indices were regressed on an implementation out­come judged as a success, partial success, or failure. The difficulty index was a stronger predictor of outcomes than the willingness index: a partial correlation of -.66 between difficulty and outcome actually increases to -.76 when willingness is controlled; the correlation between willingness and outcomes is.33. The key factors in the conflict environment were the existence of a spoiler, the presence of disposable resources, and a neighboring state hostile to the agreement.

Taking the Downs-Stedman analyses a step further, Druckman and Albin (2008) included the extent to which each of these cases included principles of distributive justice in the text of the agreement: each document was coded for the number of principles included - equality, proportionality, compensation, need. They found that justice moderated the rela­tionship between difficulty and the implemen­tation outcome: the more justice principles included in the agreement, the less severe the effects of difficult environments. Justice principles can offset the negative effects of the conflict environment; they can also reinforce the positive effects of an environment that is more conducive to peace. These findings were bolstered by the results of complementary qualitative (focused comparisons) analyses and will be re-visited in the section below on multi-method research.

The four case-study approaches entail somewhat different data requirements. For enhanced case studies, detailed descriptive information is needed. Event chronologies must be developed for both qualitative and quantitative time series. Comparable units must be constructed for matched-cases in focused comparisons and generic codes (e.g. cooperation-competition scales) are essential for aggregate analyses. These data and mea­surement requirements turn on the availability (and quality) of information. Often, the information comes from archival sources such as the Pew Case Studies in International Affairs (1999). Missing from most archival documentation, including the Pew cases, is a common framework that organizes the diverse case material for analysis. This shortcoming was noted by the authors in the special issue on data sets edited by Telhami (2002). They recognized the value of conceptual frameworks that capture processes, dynamics, structures, relationships, and issue areas for analyses of negotiation in particular but also for studies of conflict management more generally. Such frameworks should guide (primary analysis) rather than follow (secondary analysis) data collection. With the advent of web-based technologies, case researchers can access information directly from conflict actors, even before settlements or resolutions have occurred, reducing their reliance on secondary analyses of informa­tion gathered at an earlier time for other purposes (Druckman, 2002). The discussion of documentation continues in the next section.