Can women really have it all?
Can women really have it all?
A response to Babu's "Beliefs that kill birthrates"
I recently came across “Beliefs that kill birth rates” by
after Scott Alexander linked to it with the following summary:What beliefs correlate with low fertility rates? You might expect to find socially liberal beliefs (like that women need to focus on their careers), but Aria Babu says the data don’t support this. Instead, the biggest driver of low fertility seems to be a belief that taking care of kids is a lot of work and you’ll screw them up if you cut any corners. Victory for Bryan Caplan and genetic determinism?
The potential link between thinking kids are a lot of work and lower fertility reminded me of a recent conversation between
and where they talked about how a belief that pregnancy is really, really bad might make women more reluctant to have kids. But still, I find it pretty hard to believe that there’s a causal link between thinking it’s best for mothers to stay home with pre-school aged kids and lower fertility rates. And in Aria’s previous post she noted that within the US the trend doesn’t hold for conservative values in general given the positive relationship between voting for Trump and fertility rate.
Stone, Lyman. “The Conservative Fertility Advantage.” Institute for Family Studies, 18 Nov. 2020, https://ifstudies.org/blog/the-conservative-fertility-advantage.
Aria’s first post used data from the European Values Study, and considered the relationship between total fertility rate (TFR) and the percentage of the population that agrees that pre-school children suffer with a working mother.
I used data from the World Values Survey to recreate the European plot below as it was easier to get in an excel format. The TFR data used in the below plots is from the World Bank. You can access the excel files I used to make all of the below plots and tables here, and each file includes links to where I sourced the data.
The plot looks basically the same as the one she showed:
Excel file used for plot. Source file.
There’s a roughly -40% correlation between these two series with the European data.
In her second post, which Scott linked to, she considered whether the relationship still holds with OECD countries using data from the World Values Survey. I also recreated that plot below:
Excel file used for plot. Source file.
The relationship still holds but the correlation was significantly weaker, only -22%. Israel, an outlier on fertility with a TFR of 3 in 2021, wasn’t included in the World Values Survey data. Note that my version of the OECD plot looks more noisy than Aria’s because she used larger y-axis limits (0-3) than was needed for the data range. That axis limit choice was also inconsistent with the small y-axis range of (1-2) that she showed in the European chart. Regardless, she concluded that the relationship still holds. This is technically true, but if you exclude the European OECD countries and only look at the 10 remaining the correlation basically disappears:
Excel file used for plot. Source file.
Non-OECD ex Europe is a very small set of countries… but I think it’s a bit misleading to say that the trend holds up outside Europe when the trend seen in the OECD sample is entirely driven by a subset of the same European countries included in the first chart.
The first comment on her post, from
suggested that maybe this correlation was due to higher immigration in the more progressive countries and a higher TFR among foreign born women. I took a look and while foreign born women in progressive countries do tend to have a higher TFR than native-born women, the overall TFRs are only mildly affected by this. This was pointed out in the reply from and is also shown in this study. France is the most significantly affected by higher immigrant TFRs in the study, but even so the difference between native born and overall TFR is small (1.8 vs. 1.9):
Looking at a few of the other countries in the upper right quadrant of the first plot, you can see that while there’s a higher TFR among foreign born women in this sample, the native-born vs. overall TFRs are still pretty close:
France source, Sweden source, Great Britain source, and first and second US source. Excel file used for table.
But as Aria noted, there is a positive correlation between fertility in the US and voting for Trump. It could be that, as she suggests:
…the birth-rate-destroying belief is not really a conservative belief about women and gender. It is actually a belief about babies. My guess is that if the World Values Survey asked whether “It takes a lot of work to raise a child well”, then we would find an even stronger correlation.
Except that we have data, not only on voting patterns and fertility, but on this exact question within the US from the General Social Survey. I was able to use their data explorer to look at how the respondent’s average number of children relates to their answer to “preschool kids suffer if mother works”. And this time it’s totally in line with my prior.
Excel file used for plot.
There’s also an association between the answer to this question and age of the respondent: the 5% who strongly agree averaged 52.3 years, the 22% who agree averaged 51.3 years, the 52% who disagree averaged 47.9 years and the 21% who strongly disagree averaged 46.7 years. But still, the relationship seems pretty strong such that I don’t think it’s just about older people having had more babies and more conservative beliefs.
So we have a situation where there’s what I think is a pretty surprising correlation between average expectations of mothers and TFR between countries in Europe. There are obviously a lot of possible confounding variables which could lead to this correlation without there being any causation. And, to be fair, Aria does test a bunch of the ones that come to mind and finds that they don’t explain it. However, looking at the non-European OECD countries we find that the relationship doesn’t hold. And then looking at the intra-country data for the US, where there’s fewer potential confounders, we find the opposite relationship. Given all of this, I’d say there’s very weak evidence if any for the sort of causal relationship that Aria asserts.
Perhaps I’ll get a chance to look at some of the other questions in the World Values Survey for a future post. And in the meantime, if anyone finds other intra-country data comparable to what I found for the US or has ideas about potential confounders please share it in the comments!