American Fertility Still Runs in Families: A Short Update

Excluding fantastic answers substantially strengthens my original results

While re-reading my “American Fertility Still Runs in Families” I noticed some sloppy work on my part. In the General Social Survey, the maximum reported number of siblings is 68! Perhaps someone is counting a massive number of polygamous half-siblings, but it’s far more likely that this respondent is lying or misunderstood the question. The same goes to a lesser degree for all of the sky-high sibling responses.

(Note: The GSS caps responses to “number of children” at 9, so this variable doesn’t have the same credibility problem).

Question: What happens to the results if you exclude fantastic responses?

Answer: The estimated tendency of fertility to run in families markedly increases.

See for yourself. Here are the original results for 1972-2022, counting all respondents:

Here are the revised results after excluding the roughly 3% of respondents who claim to have more than 10 siblings.

The magnitude of the key coefficient rises by 13%!

My original results are broken down for the 20th versus 21st centuries. These, too, get stronger once you discard respondents with more than 10 siblings.

20th Century - Original Results

20th Century - Updated Results

The key coefficient is 11% higher.

21th Century - Original Results

21th Century - Updated Results

Fully 17% higher after throwing out the hard-to-believe respondents.

Would I have posted these revised results if they undermined my original thesis? Yes, conscience requires it. Once I realized my initial oversight, the only scenario where I wouldn’t have written a full new post is the one where the results barely changed. In that case, I would have simply posted an update at the end of the original essay.