Jim Manzi writes:
I just read your post on the recent paper on the causal effect of daughters and sons on political orientation. It struck me as really weird that the sex ratio among respondents to both questions was almost exactly 1.1. Isn’t it very stable around 1.05 for the US population? If so, couldn’t there be a very biased sample of people who are willing to respond to these questions that could account for the difference? (Though it does seem weird – though not inconceivable – that such a bias would be almost identical for both questions.)
My reply: I dunno, perhaps some differences in the rate at which people give up kids for adoption? Or maybe it’s some sort of nonresponse issue. It would make sense to see if something similar is happening in other surveys that ask about the sex of respondents’ children, and it would also make sense for the researchers mentioned in the cited post to do some sensitivity analysis regarding these selection issues.
I have looked at this some in other data sets. Daughters seem to leave the home before sons, which will bias the sex ratio. To deal with this, I and others limit the sample to families whose eldest child is younger than say 12.
I did some investigation with the GSS today, and found that the proportion of children female has a significant association (t=2.9) with Republican party ID, and a marginally significant (t=1.65) association with conservative political views. Then I looked at opinions on 8 or 9 political issues, including both social and economic issues. There was an association with support for the death penalty (t=1.9); nothing else was close. This absence of any clear pattern of association with views on specific political issues leads me to wonder if the apparent connection to party ID might just be sampling error. Of course, there are a lot of questions in the GSS I didn't consider, but I did consider a pretty wide range (redistribution, abortion, welfare, religious views…). So looking at other surveys would be informative, but most political surveys don't seem to ask about the sex of one's children–at least I don't know of any that do.