Regression to the Mean

Here’s a thought experiment. Randomly pick two White individuals who have IQs of 140. They marry and have children. What is the adult IQ of their children? Research on the heritability of adult intelligence generally find the following pattern: genetics ~0.75, shared environment ~ 0.0, unshared environment ~0.25.

This means that of the above average IQ scores of the parents, about three fourths of it is accounted for by the parents’ genetics, this is passed down. None is accounted for by the family environment the parents raise the children it, so there’s nothing to pass down. The remaining one fourth of it is accounted for by “luck,” this is not passed down from parent to child. So the children of those two parents should have IQs averaging 130.

Now what happens when the child of those two 140 IQ parents, who has an IQ of 130, marries another White person with an IQ of 130 and has children. You might think the same calculation should be done: 130-30*.25 = 122.5. But it has been already determined that one of the parents has only average “luck.”(“Luck” can help or hurt.) So the calculation that should be done is 130-30*.125 = 126.25. You can continue this over many generations and you’ll see that regression to the mean occurs but it happens quite slowly:


What if both of the 130 IQ individuals had parents with IQs of 140? Then there would be no regression to the mean in their children. If you took a bunch of people with IQs of 140 and segregated them on an island and then measured their descendants’ IQs they should average 130.

In the real world the individual with an IQ of 140 would have married someone of merely above average intelligence. If he married someone with an IQ of 110 their children should have IQs averaging 118.75.

This entry was posted in Genetics/HBD. Bookmark the permalink.