Logging is standard practice yet its effect on semantic validity goes unexamined.

Evidence

Logging a variable (i.e., using the natural log of the variable) is very common and often recommended in statistical research. To take a classic example, the decision to log GDP per capita is rarely controversial. … In a cross-national study, easily 50 percent of the continuous variables could be logged. Does logging increase or decrease semantic validity? — Gary Goertz, Social Science Concepts and Measurement, 728

Categorization

• QUESTION: The chapter’s own open question — does the ubiquity of logging corrupt the match between variable and concept? Worth tracking the book’s answer.