Word tests are used as part of good, high quality IQ tests (e.g. WAIS and Stanford-Binet, though obviously not Raven’s). They are in fact one of the most useful and predictive parts (I.e. they are most g-loaded). They are biased (i.e. not measurement invariant) if you compare subjects from different countries/cultures, but they are not biased if you compare subjects from the same country/culture. For example, the score on word analogies sections has exact same predictive validity for white Americans as it does for black Americans. See Arthur Jensen’s “Bias in mental testing” for a very comprehensive treatise on the issue.
