Indicators of a single dimension can correlate poorly yet still be combined.

Evidence

Normally this means that the indicators should be highly correlated. In fact the OECD indicators of quality of life are frequently not very highly correlated (see OECD 2011, table 1.A.1, “Correlations between the headline indicators included in How’s life?”). The conceptual and methodological question is what to do when these indicators of the same thing significantly disagree: Does that mean that they are not indicators of one secondary-level dimension? The OECD uses them anyway in spite of low correlations. — Gary Goertz, Social Science Concepts and Measurement, 253

Categorization

• QUESTION: What justifies retaining low-correlation indicators — aggregation convention, or a claim that they still measure one latent dimension?
• SHARE: A perfect tidbit for anyone who has eyeballed OECD How’s Life sub-indicator correlation tables and quietly wondered why they are averaged.