Researchers log data to fix skew, not to preserve conceptual meaning.

Evidence

The most common rationale for logging, however, has nothing to do with semantics. If one were to ask why a scholar is logging data in a statistical analysis, almost certainly the answer would be because the data are skewed, as seen in the top panel of figure 4.9. The big majority of cases are toward the low end and the long tail is typical of skewed data. This kind of skewed data is very common in the social sciences. — Gary Goertz, Social Science Concepts and Measurement, 731

Categorization

• QUESTION: If the log’s only justification is distributional, what happens to concepts whose meaning is genuinely nonlinear in the raw variable?