This post has been updated with 2022 data, here.
I previously reported that the name binary in the U.S. was weakening, as indicated by the tendency of babies assigned male and female at birth to be given different names. I’m updating that now.
UPDATE: Now with England and Wales added at the end!
First, an important note on the dynamics of name androgyny. In a 2000 paper, Lieberson and colleagues had argued that nonbinary names were “unstable,” that is, the trends for a given androgynous name tend to tip rapidly one way or the other, presumably because parents don’t want ambiguity in their children’s names. And that had become conventional wisdom. But since my last post, in a very cool paper, Seguin, Julien, and Zhang have pretty well convinced me that’s not how it works. In fact, “androgynous names, although rare, are stably androgynous.” So some conclusions in my previous post are probably wrong.
Anyway, Seguin et al. don’t deal much with the long-term trend toward androgyny, and I don’t see that anyone else has devoted much analysis to explaining the trend described here. Unless I’ve missed it, I think someone should! (my data and code for all this are shared here).
Today’s update is motivated by the revisions I’m working on for the fourth edition of my book, during which I noticed that androgyny was still climbing rapidly, and I thought you needed to know that right away. Here is the trend, by two measures. The first line is the “index of androgyny” proposed by Lieberson et al. who describe it like this: “For the name of every individual girl born in a given year, the boys’ proportion of children with that name. The index gives the average of all these proportions.”1 The second line is the proportion of babies born each year who get a name that is between 20% and 80% single-sex. This is pretty rare, now 6%.
Those two measures are highly correlated, and the pattern is clear: name androgyny is at a record high in the history of the Social Security Administration data series, and has risen a lot in just the last few years.
To look at it a different way, here is the histogram of name gender for two points 50 years apart (gulp!): 1971 and 2021:
I like this figure because you can see the declines at the tails — 0% female and 100% female (statistically, the kurtosis has fallen from 1.4 to 1.1). You can also see that there has been more change on the left end of the distribution, so the change appears to be coming primarily from fewer babies getting all-boy names rather than fewer getting all-girl names.
Finally, if you’re curious about the more truly androgynous names, here are the most common names in 2021 that were between 40% and 60% single-sex. Altogether just 1.7% of babies were given names in this range, and the 20 on this list account for 1% of babies born.
Note that in this I’m not combining names by pronunciation, as some people have done, so “Charlie” is not combined with Charlee, Charli, Charly, Charleigh, and Charley (or with non-nickname versions, like Charles or Charlotte). Someone on Mastodon pointed out that “Ari” is still a boys name for Jews, so there might be an ethnic element to these. (I wrote about Jewish names here.)
On a tip from Martin Rosenbaum, I went and got the name data for England and Wales, which only goes back to 1996 but is otherwise very similar to what the US puts out – complete registry of names (with at least 2 per year). After getting the data into the same shape as the US files, my quick take is that name androgyny is much less common in England and Wales (about half as high), but is also increasing, at least since 1996. Here is the comparable figure:
Here are the most common androgynous names in England and Wales, 2021:
Someone could look at the overlap in androgyny between the USA and England/Wales. I put the data and code for England and Wales up on OSF, too.
For the complete series of posts, follow the names tag.
1 The index of androgyny ranges from 0 to about .51 (depending on the sex ratio at birth). If you like this kind of thing:
“where G is the total number of girls born in a given year, gi is the number of girls with a given name, bi is the number of boys with the same given name, ti is the total number of children with the name.” With the magic of computers, this is two lines of Stata code,
gen lterm = (g/G) * (b/(t))
bysort year : egen lieband = total(lterm)
where the total number of girls born in each year (G), the number of girls (g) and boys (b) with a given name in that year, and boys or girls with that name (t), are all in the same row (so, one row per name per year, N=1.9 million).