Process-Tracing Methods: Foundations and Guidelines

Chapter 4. Working with Theories of Causal Mechanisms

45 Chapter 4 Working with Theories of Causal Mechanisms Theoretical concepts are not self-explanatory. Nor are causal mechanisms self-evident from causal theorization of relationships between X and Y. Instead , causal theories need to be transformed so they offer a clear hypothesized mechanism describing how a type of outcome is produced. Therefore, theoretical concepts and causal mechanisms need to be carefully defined before they can be employed in process-tracing analysis. While there are numerous methodological texts relating to conceptualization techniques in political science, existing guidelines are not always applicable when we are using process-tracing methods.1 This chapter adapts existing guidelines to each of the three variants of process-tracing. We start by discussing some of the common challenges shared by the three variants of process-tracing methods. First, we discuss how concepts should be conceptualized, arguing that defining key concepts in settheoretical terms is more compatible with process-tracing than viewing concepts as variables. We then elaborate on the common theoretical elements that causal mechanisms share in the three variants—in particular, the need to elaborate the content of each of the parts of the mechanism that are composed of entities that engage in activities. Activities are explicitly conceptualized in process-tracing to capture how causal forces are transmitted through a causal mechanism to produce an outcome. We then discuss how causal mechanisms differ across four types of theoretical explanation, the analytical level at which they operate, the degree of contextual specificity, and the time span in which they are theorized to operate. The second section illustrates the specific challenge in working with theories in each of the three variants of process-tracing: theory-testing, theory- 46 Process-Tracing Methods building, and explaining-outcome. We show that the conceptualization phase of theory-testing process-tracing involves the mostly deductive task of using existing theorization to flesh out the causal mechanism between X and Y, whereas theory-building derives theories from a more inductive analysis of empirical material. In contrast, conceptualization is an ongoing and iterative process in explaining-outcome process-tracing, involving multiple stages of analysis of evidence and theoretical reformulation (see chapter 2). 4.1. Common Challenges Shared by All ThreeVariants Defining Key Concepts Key to any research design is the definition of the central concepts that form the basis for theoretical propositions. Adcock and Collier (2001) refer to this as the translation of abstract theoretical concepts into what they term systematized concepts. Most important theoretical concepts are contested, with multiple plausible meanings, and they are often quite ambiguous and vague at the abstract level. We therefore need clear systematized definitions that distinguish between what is included and not included in the concept that allow us to know it when we see it. Defining systematized concepts is not an “anything goes” process. For most concepts, a considerable amount of theoretical work describes what the concept should include and not include, and our systematized concepts should be faithful to this work (Adcock and Collier 2001: 532). Conceptualization involves defining the constitutive dimensions of a concept and how they relate to each other (Goertz 2006). For example, a systematized definition of democracy could include the two characteristics of civil rights and competitive elections. These secondary characteristics obviously should not rule each other out. It is also important to note that concept formation in qualitative case study research such as process-tracing is attentive to the details of cases, resulting in more context-specific conceptual definitions that have a narrower scope than are commonly used in large-n analysis (Ragin 2008: 78–81). There are important differences in how concepts are defined depending on whether they are considered to be variables or conditions. In the understanding put forth by King, Keohane, and Verba (KKV), theoretical concepts are variables. Causal relationships are then formulated in terms of an independent variable (X) (or set of variables) that are theorized to Working with Theories of Causal Mechanisms 47 cause variation in the dependent variable (Y) (King, Keohane, and Verba 1994). This type of theory usually describes a probabilistic causal relationship , where an increase in X raises the probability of Y occurring in a given case (Gerring 2005: 167). The term variables means that they must be able to vary. If we are conceptualizing democracy as a variable, then we also need to take into consideration whether democracy is a dichotomous variable (democracy—autocracy) or an interval scale variable (for...