On Coding

Conceptual analysis is an important part of political philosophy. Sometimes done explicitly with arguments devoted to analysing and defending a specific way of defining or treating a concept such as freedom or power. At other times, done less explicitly where terms are used within a broader theory of the good life or constitution of society. Many claim that normative terms are essentially contested since different definitions are equally plausible versions of an original exemplar with no way of choosing between them. Another reason why conceptual debate seems interminable is that concepts are vague and bringing precision gives different precise conceptions of the vague concept. In this paper, we argue that moral and political concepts are vague and the drive for precision can be seen in terms of coding decisions for concepts in empirical work. There are two main ways that coding occurs in empirical political science: the analytic and the inductive. Analytic methods turn qualitative concepts into numerical data through coding frames listing characteristics of empirical political phenomena such as ‘type of legislation’ into specific categories. Here, human coders surface details of the cases deciding which category to place them. Hard cases are identified and discussed, and sometimes the coding is altered or added to in order to accommodate them, or decisions made about how handle the hard case. Where we use multiple coders, techniques exist to examine inter-coder reliability. Failures in inter-coder reliability might occur because (i) mistakes are made and items are miscoded, (ii) different decisions are made over hard items. The first, ‘objective error’, creates noise in analysis but may not be a problem if miscoding is normally distributed over the coders. The second, ‘subjective error’, is more of a problem since it is highly likely bias will occur because of systematic disagreement. The inductive method is more recent where computer algorithms find patterns in data automatically coding (usually written) words into categories that share lexical, compositional or psychological similarities which correlate with a dependent variable. These patterns are then interpreted as to their meaning in the context of the research question. We suggest that conceptual analysis for vague political concepts can be treated as an analytic coding problem. Different coding frames can lead to different analyses of the same phenomenon. Often the different analyses are convergent in the sense that coding frames vary only at the margins. Where they do conflict predictive veracity and theory lead us to prefer some frames over others. Sometimes different frames do not matter since they are designed to answer different research questions. Complex though vague political concepts might also be coded differently for different moral or political purposes. We shall argue that again sometimes this does not matter since they vary only at margins, or they are used for different non-rival purposes, but sometimes their utility in analysis and theory should guide us as to which we should prefer. We will also briefly consider analogies between inductive empirical methods and what that might teach us about moral and political theorizing.