A quick note on liberal family values

The ISSP data

The International Social Survey Program has a nice module on gender attitudes. That lets us focus on social liberal attitudes to the family specifically. I’m still using only European countries; the US and Asia are too different and need separate analysis. I’ll also just use the latest wave, from 2022. Here’s the ISSP participant countries from Europe:

There are 11 “value” questions which cover enough countries, and don’t have too much missing data, to give a reasonable sample. (Missingness is a big issue, but not something I can do much about.) Here they are:

Working mom can establish as warm relationship as a not working mom
Family life suffers when woman has full-time job
Men’s job earn money, women’s job look after home
Women or Men more responsibility for home and family
Marriage if people want kids
Couple living together without marriage ok
Single mother can raise child as well
Mother, father: who should take care of child on daily basis
Mother, father: who should teach child how to behave
Mother, father: who should take time to listen to and advise child if proble
Mother, father: who should try to be role model for the child

The first principal component of these questions gives a measure of social liberalism: it loads in the expected direction on all of them, and explains about a quarter of the variance of all the questions. I’ve called this “liberal family values”.

Without further ado, here’s how it relates to completed fertility.

The dotted line shows replacement fertility. The grey area is a 95% confidence interval. This is the same picture that we saw before for social liberalism in general:

Fertility and social liberalism in Europe

though in this 2022 survey, differences in fertility seem not to have flattened out so much as in the latest European Values Survey wave.

That said, there’s an important technical nuance. I’ve weighted respondents to match their country populations, which means big countries count for much more than small ones. If we weight all countries equally, we get a different line:

Weighting all countries equally, the relationship between liberalism and fertility is much flatter and seems to show a U shape. Why the difference? Well, there’s a lot of variation in fertility between countries, and there’s probably also variation in the fertility-liberalism relationship. Really, the two lines are answering two different questions. Weighting by population treats the whole of Europe as a single unit, and takes the average result over the whole population. Weighting countries equally is more like taking the average of all different countries. I’ll focus mostly on population-weighted results, to match what I did for the European Values Survey, but be aware that this reduces our effective sample size.¹

Let’s look at the different countries. Each dot below shows the effect on fertility of a 1 standard deviation change in liberal family values, estimated separately in each country:

No countries show a significantly positive relationship, and most countries (including all the big ones) show a negative relationship, often significantly so. Not surprisingly, weighting the big populations more will show a more negative relation across Europe overall.

A couple more interesting points. Here’s the graph again, with extra lines dividing survey respondents by social status (self-reported on a subjective 1-10 scale, which I’ve split into three groups).

I also added the population densities of liberalism for each status group, below the main plot. The top plot shows that liberalism predicts fertility more strongly for low-status people than for high-status people: the low-status slope starts off higher and ends up lower. The bottom plot shows that high status people are a lot liberal.One possible interpretation: low-status people are more conservative on this value dimension because they do worse from adopting liberal beliefs.

These differences aren’t huge and may not be significant, but if we divide people by education, we seem the same story more clearly:

Degree-holders’ fertility seems not to depend on liberal family values, unlike non-degree-holders. And again, from the density plot at the bottom, degree-holders are much more liberal.

How much of this is driven by marriage? Well, conservatives on family values are certainly more likely to be married. And that is still true, even within the group who are living together with a steady partner:

So, what happens if we split fertility by marital status?

Actually, the line for married people is more or less the same as for the whole sample. Neither previously-married people or never-married people show a negative association between liberal values and fertility. As a side note, married people are clearly more conservative than the other two groups.

Causality?

One explanation for this data is that liberal family values causally reduce people’s fertility. But there’s another very plausible story: having children makes you more conservative on family values! If you’ve had children, you probably know how much it changes your view of the world, and of how family life works in particular. It would not be surprising if this shifted people’s views. In particular, I think it is much more likely that having children makes you conservative on family matters, than that having children makes you more conservative in general.

And, unlike other non-causal stories for the link between liberalism and fertility, this one would affect cultural evolution. As I said last time, if some confounding factors explain the link, it can still be true that conservatism increases over generations because conservatives have more children. But what if things are like this: as soon as you have your first child, you become more conservative on average; but your children, until they have children of their own, are just as liberal as you used to be. Now we see an association between conservatism and fertility, but conservatism doesn’t spread at all.

We can do a bit to check this by looking at the exact number of children people have. Here’s the proportion of people having 0, 1, 2, 3, 4, or 5 or more children, by their liberal family values:

Conservatives are less likely to be childless, but they are also more likely to have large numbers of children: 2 or more, 3 or more, and so on.

That’s interesting in two ways. First, it makes the reverse causality story seem a bit less likely. Having your first child may well change your values, but it seems less ptobably that having a third or fourth child systematically shifts people’s values in one direction or another. I mean, I guess you value a big car more.

Second, cultural evolution still works even under reverse causality, so long as larger families have more conservative parents. (For cultural evolution not to work, children from larger families would have to shift back towards liberalism from their parents’ values more than children from smaller families, which seems very unlikely.)

Still, this is just a loose argument. I can’t deny that reverse causality might explain some of what we see. But let me remove my analyst nerd hat, put on my conservative gadfly hat, and prong family values liberals on the horns of a dilemma. Maybe your values causally reduce fertility; that doesn’t seem great. Or, maybe liberal values aren’t bad for fertility, but when people actually have children, they revise them — which suggests that those values are not very compatible with reality?

Or maybe it’s something else! I don’t know. Just planting a little seed of doubt.

Really, though, what we need to do is track people over time. Then we can actually see if their values move after they have children.

Bonus

Finally, just to pique your curiosity, here’s a global chart using all the ISSP countries. I’ve split out two separate countries with really huge populations. “Other” is mostly Asian countries, plus South Africa and Israel.

This is still the same Eurocentric scale I used before, i.e. the principal component estimated using Europeans only. But liberal values still seem to predict lower fertility worldwide… with the possible, interesting exception of India. (Why not? Maybe they’re just picking up an unrepresentative sample? Or is this driven by income?)

And the other thing to see, from the population density plots in the bottom margin, is just how conservative the rest of the world’s family values are compared to Europe and America…. If you’re European, you’re weird, and if you’re university-educated too, you are really weird.

Code for the graphs.

1. To understand why, think of an extreme case, where only one respondent has a non-zero weight. Then our effective sample size would be exactly one.