Case Studies and Theory Development in the Social Sciences - Alexander L. George [15]
Consequently, statistical research is frequently preceded by case study research to identify relevant variables and followed by case study work that focuses on deviant cases and further refines concepts.43 For example, after a range of statistical studies suggested that democracies do not fight other democracies, case study researchers started to explore which aspects of democracy—democratic values, democratic institutions, the transparency of decision-making in democracies, and so on—might be responsible for this apparent “democratic peace.” Should these case studies indicate, say, that transparency is an important causal factor whereas universal suffrage is not, then revised and new statistical tests are performed.
DERIVING NEW HYPOTHESES
Case studies have powerful advantages in the heuristic identification of new variables and hypotheses through the study of deviant or outlier cases and in the course of field work—such as archival research and interviews with participants, area experts, and historians. When a case study researcher asks a participant “were you thinking X when you did Y,” and gets the answer, “No, I was thinking Z,” then if the researcher had not thought of Z as a causally relevant variable, she may have a new variable demanding to be heard. The popular refrain that observations are theory-laden does not mean that they are theory-determined. If we ask one question of individuals or documents but get an entirely different answer, we may move to develop new theories that can be tested through previously unexamined evidence.
Statistical methods can identify deviant cases that may lead to new hypotheses, but in and of themselves these methods lack any clear means of actually identifying new hypotheses. This is true of all studies that use existing databases or that modify such databases only slightly or without recourse to primary sources. Unless statistical researchers do their own archival work, interviews, or face-to-face surveys with open-ended questions in order to measure the values of the variables in their model, they have no unproblematic inductive means of identifying left-out variables. Even statistical methods of “data mining” necessarily include only those variables that a researcher has already thought to code into a data base. Deductive theorizing can also identify new variables, but with the exception of purely deductive theories, inductive field research methods typically lie behind every newly identified variable.
EXPLORING CAUSAL MECHANISMS
Case studies examine the operation of causal mechanisms in individual cases in detail. Within a single case, we can look at a large number of intervening variables and inductively observe any unexpected aspects of the operation of a particular causal mechanism or help identify what conditions present in a case activate the causal mechanism. Our definition of causal mechanism (see Chapter 7) notes that such mechanisms operate only under certain conditions. Statistical studies, which omit all contextual factors except those codified in the variables selected for measurement or used for constituting a population of cases, necessarily leave out many contextual and intervening variables.
Researchers can also use theories on causal mechanisms to give historical explanations of cases. Historical explanation is quite different from the development and testing of variable-centered theories from the statistical study of a large number of cases. As statistical researchers frequently point out, correlation does not imply causation. If a prosecutor