IQ was originally devised by Binet as a way to measure the 'retardedness' of children. It is found by dividing the child's mental age by their chronological age. 'Mental age' is found by subjecting the child to a series of tasks generally thought of as being a good indication of intelligence.

IQ was never meant as a test for people of regular intelligence - Binet himself complained bitterly over this application of his test.

Nowadays, IQ has come to be almost synonymous with intelligence. The problem is that intelligence is not inherently measurable or even meaningful. The IQ test works by picking a bunch of tasks which the scientists of those times found easier than the average person.

Who were these scientists? Almost all caucasian males, with analytical minds. Is it any wonder, then, that caucasian males with analytical minds score higher on this test than other people?

The problem is that the IQ test only measures ability to perform IQ tests! (thanks to iain, [and also ThePope], for this statement.)

It is perfectly possible to create an intelligence test with a bias towards the negroid race, or females. (And presumably negroid females).

See also multiple intelligences. (For another nail in the coffin.)

N.B. 1) As for the idea of different races having different 'styles' of intelligence, it is possible, since the races developed in separate environments and thus have a different genetic history. However it is far from a proven fact.

N.B. 2) The idea of males and females having different 'styles' of intelligence is far more likely. Neuroscientists have shown that there are differences between male and female brain structures, so Marilyn Vos Savant aside, females should do worse in male-biased tests. A female-biased test might have the testee having to constantly remember facts from previous questions, rather than each question being separate. And there might be less focus on those nasty 'three-dimensional shapes' problems, more on vocabulary ones.