A 2015 paper by Kate Cockcroft et al., compares the scores of 349 British middle class university undergrads to 107 lower class black South African undergrads on the WAIS-III (UK edition).

The results were as follows:

The UK students averaged a full-scale IQ of 106.95 (UK norms) while the South Africans averaged IQ 93.27. However because this study was published 18 years after the UK WAIS-III was published, we should adjust for the Flynn effect.

The single best source on recent Wechsler Flynn effects is Weiss et al., 2015  which found that full-scale IQ has been increasing by 0.31 points per year, at least in U.S. children. If we assume it’s the same for UK adults, then the UK students have an adjusted IQ of 101 and the South African students have an adjusted IQ of 88.

What’s more interesting to me is how they did on the culture reduced tests since that’s the more fair comparison.

Test:	Digit Span scaled score	Flynn adjusted Digit Span  scaled score	Flynn adjusted Digit Span IQ equivalent	Block Design scaled score	Flynn adjusted Block Design scaled score	Flynn adjusted Block Design IQ equivalent	Compoite IQ based on adjusted scores on both tests
UK undergrads	9.5	9.32	97	10.66	9.76	99	98
Black South African undergrads	9.35	9.17	96	8.67	7.77	89	91

So on a composite score of the most culture reduced spatial & non-spatial test (Block Design & Digit Span), Black South African undergrads average IQ 91. This is 11 points higher than the average Black South African seems to score on the same culture reduced tests.

As of 2013, only 16% of South Africa’s black young adults were attending higher education (compared to about 55% of whites, 47% of Indians and 14% of Coloureds). Thus, simply attending university puts one in the top 16% of this demographic, with the median South African university student being in the top 8%. If there were a perfect correlation between IQ and education, the median South African black university student would have an IQ 21 points higher than the average black South African. In reality his IQ is only 11 points higher, suggesting a correlation of 0.52 (at least on the most culture reduced tests).

This is similar to the 0.57 correlation between IQ and education observed in the United States.