A measure adequate for description becomes incomplete once used in causal hypotheses.

Evidence

Many World Bank consumption surveys are motivated in large part to analyze poverty. One might be descriptively interested in consumption in the sense of what people are spending their money on. If this is the case then the concept is really consumption. However, once one is interested in concepts like poverty or wealth, or one begins to use such concepts in causal hypotheses, then consumption is fundamentally an incomplete concept. Hence bipolarity in this example means that consumption ranges from almost none to huge dollar amounts. It is incomplete as a measure of poverty, one end of the continuum of consumption. — Gary Goertz, Social Science Concepts and Measurement, 630

Categorization

• DEEPTHINKING: The strongest idea in the chapter: concepts are tools selected by the causal question, so the same indicator is complete or incomplete depending on what you’re testing. Ragin-style set theory meets measurement theory — worth sitting with.
• XREF: Resonates with the software-engineering insight that a variable is only well-defined relative to a query, and with calibration problems in physical measurement. — connects to: The implicit self-esteem hypothesis requires an indirect measure that did not yet exist. — Anthony G. Greenwald and Mahzarin R. Banaji, Implicit Social Cognition: Attitudes,…, 173