Toward a Pluralistic Vision of Methodology

Abstract

Both data-set observations (DSOs) and causal-process observations (CPOs) are important for causal inference. DSOs – located in the standard “rectangular data set” of statistical analysis – make their contribution through a quantitative logic of comparison, frequently using different forms and extensions of regression analysis. They are deservedly a major research tool in political science. Yet when based on observational data, they must be analyzed with great care. Problems such as cofounders, for example, can be hard to address. DSOs are therefore valuable, but serious limitations must be recognized. CPOs provide a different form of analytic leverage. They are carefully selected “nuggets of data” that are a foundation of qualitative research and can have great probative value in evaluating hypotheses. Achieving rigorous inference with CPOs depends on exact specification both of the inferences to be evaluated, and of standards of evaluation. If such specification is inadequate, CPOs likewise have serious limitations. A more powerful option – which can help address limitations on both sides – is to use DSOs and CPOs together in a single study. As our examples show, major studies have indeed adopted this multi-method approach, thereby achieving strong inferential leverage.

Keywords: methodology, causal inference, multimethod research, qualitative methods