Charles Murray has a paper in press at Intelligence analyzing data from the National Longitudal Survey on Youth (NLSY) relevant to the claim by Dickens and Flynn that the IQ gap between whites and blacks has narrowed during the past thirty years (see also here). The children of the very large and representative sample of American women beloning to the same 1979 cohort (blacks and low-income whites were oversampled to give 3652 whites and 2557 blacks) were given standardized tests of ability and achievement periodically over the subsequent twenty years. The analysis of these data is mildly complex, but the findings are more or less simply conveyed:

The black sample sizes for the NLSY children are very large compared to those used in IQ standardization samples, which typically consist of just a few hundred children (e.g., the black sample for the 2002 standardization of the WISC-IV consisted of 343 subjects [Prifitera, Weiss, Saklofske, & Rolfhus, 2005]). Most of the subjects were tested more than once. The testing procedures identify and test children who are easily missed in surveys that rely on children being present at school on a day designated for testing. The same versions of the tests were used across time. Cognitive test scores were available for the mothers, along with extensive family background data. While no single database can provide dispositive results, the database for the NLSY79 children has many strengths, and it yields consistent evidence, robust across alternative specifications and samples, that the B-W difference did not diminish on either academic achievement or cognitive tests for children born from the mid 1970s through the mid 1990s.

Recall the tentative conclusion from the discussion in our previous post on this issue that the gain by blacks relative to whites seems to have been concentrated in those blacks born between 1955 and 1975 (hat tip to Steve Sailer). Murray tentatively endorses this interpretation:

The inconsistency between the evidence from the NLSY and from the IQ standardizations [on which Dickens and Flynn base their conclusions] remains, but it is not necessarily irreconcilable. The Dickens and Flynn evidence for a narrowing gap was based on means and standard deviations for age groups. The hypothesis suggested by the data from the NLSY children is that if individual data from the IQ standardizations were analyzed by birth cohort, they would reveal that the effect that Dickens and Flynn found was concentrated among subjects born before the late 1970s. Such a finding would inform a variety of theoretical explanations of the B-W difference, but elaborating on those implications is premature until the hypothesis is tested.

Murray's paper has been uploaded to GNXP Forum. When Psychological Science finally publishes the original papers by Dickens, Flynn, Rushton, and Jensen, we may make them available in the Forum as well.