Introduction

The approach we present in this book is not the only way to select cases for pathway analysis. The literature suggests two alternatives: the variable-based approach and the residual-based approach. This chapter reviews these methods, critiques them, and compares their application to our method using several textbook examples drawn from John Gerring's Case Study Research: Principles and Practices (2007). We use examples from Gerring for several reasons. They are simple and heuristically useful. These examples have been used to demonstrate the utility of the prior case selection approaches, and therefore they provide a benchmark against which to demonstrate the advantages of our approach. In subsequent chapters, we apply our approach to more complex examples, but many of the basic points – such as the need to select comparative cases, the usefulness of considering both the expected X1/Y relationship and variation in case characteristics, the utility of visualizing patterns within the data, and using case control strategies – can be seen even in this chapter's simple examples.

Overview of the variable-based approach

The variable-based approach has the advantage of simplicity; it urges researchers to seek cases with extreme values of X1 and Y (see Seawright and Gerring 2008). The assumption of this approach is that cases with large X1 and Y values are most likely to feature the X1/Y relationship and thus provide a promising case for exploring causal pathways. The values of X1 and Y are potentially useful pieces of information for those interested in pathway analysis, but focusing on the X1 and Y values alone is problematic for a number of reasons. First, the variable-based approach does not deal with the problem of potential confounds; it simply assumes a straightforward X1/Y relationship. Second, if we select cases with extreme values of X1 and Y, it is difficult to assess whether these cases are outliers in the absence of a strong theory about the nature of the relationship.