In the discussion on this post about interpreting historical trends, several people pointed out that the relationship between fertility rates and women’s employment rates is not simple, and has changed, at least in the rich countries. I made some charts using international data about that, which I will show below.
But first a figure from this paper by Rense Nieuwenhuis and colleagues, which he linked from the comments. In that 2012 paper they show that the negative association between motherhood and employment weakened in OECD countries from 1975 to 1999. Still, at the individual level, in almost every country and every year, the odds of being employed are lower for mothers, as this figure shows (dots lower in each box indicate a bigger employment gap between mothers and non-mothers; click to enlarge):
It’s a very interesting paper I should have recommended earlier.
The fact that mothers are less likely to be employed than women without children doesn’t mean that countries — or time periods — with lower fertility rates necessarily have higher women’s employment rates (see Nieuwenhuis’s comment for a few other papers on this). So it’s good to look at individual as well as macro-level patterns.
Anyway, those are all rich countries. What about poorer countries? Because of the unbelievably good archive of census data (freely available, thank gov) at IPUMS International (74 countries, 238 censuses, 544 million records, and counting), it’s possible to ask questions like this.
Looking for censuses that recorded the number of children ever born to women, as well as their employment status, I sampled 10,000 households each from 89 censuses in 29 countries in Latin America or the Caribbean, Asia, and Africa, ranging in time period from 1960 to 2010. I limited the samples to women ages 25-44, and counted their children up to 7. The countries were:
- Latin America / Caribbean: Argentina, Bolivia, Brazil, Cambodia, Chile, Colombia, Costa Rica, Ecuador, El Salvador, Haiti, Jamaica, Mexico, Nicaragua, Panama, Peru, Uruguay
- Africa: Burkina Faso, Ghana, Guinea, Kenya, Malawi, Morocco, Rwanda, Senegal, Sierra Leone, South Africa
- Asia: China, Indonesia, Vietnam
Here’s what I found. Overall there is not a strong correlation at the country level between mean number of children born per women and employment rates (correlation = -.09):
Closer inspection reveals a pretty strong relationship in the Latin America / Caribbean samples, as well as the three Asian countries, but not the African samples. But this scatter doesn’t show the time trends. If I limit it to the 9 countries that have at least 4 censuses (8 from Latina America, plus Indonesia), they almost all show the pattern I started with: falling fertility and rising women’s employment rates. The arrows track each country’s censuses in chronological order, so moving up and to the left fits the historical pattern:
The country-level association is not the same as an individual-level association, because it can’t confirm that women with more children themselves are the ones who aren’t employed. To gauge that I estimate a linear regression within each census, measuring the association between number of children ever born and employment, controlling only for age. These are the results from those 89 regressions. The x-axis is still the mean number of children in each sample, but now the y-axis is the statistical effect of each additional child on the probability of being employed: below 0 indicates that having had more children reduces the probability of employment.
In 15 of the 89 samples, each additional child is associated with a greater chance the woman is employed, but in 74 samples the effect is negative*. Furthermore, it appears that countries with lower fertility rates have a stronger negative association between children and employment — each kid reduces the odds of employment more. Consider, though, that a reduction of .11 in the probability of employment for each kid has a lower total effect in a country with two children per mother than a reduction of .05 in a country where people have three kids each**.
If we go back to the 9 countries with at least 4 censuses each, we can compare the trends in fertility to the child effect on employment:
Most of these countries (Chile, Colombia, Indonesia, Panama, and Mexico) show the pattern in which the child effect strengthened while the fertility rate fell. Uruguay and Argentina show falling child effects and little fertility change.
Two possible conclusions:
- Although it may seem prosaic, this reminds me that the long-run, modern movement of women into the paid labor force is closely associated with the decline in fertility (as well as, incidentally, the decline in marriage). I think of that as indicating that women’s labor is increasingly diffused outward from their own children through market (or otherwise socialized) mechanisms. As the prototype, think of a woman with 2 children teaching 30 children in school (while her own kids are in another classroom) instead of spending the day caring for 6 children at home (while growing food, etc.).
- The trend toward a smaller employment gap between mothers and non-mothers is a recent, selective, rich-country phenomenon associated with very low fertility rates and (as the Nieuwenhuis et al. paper nicely shows) state policies designed to encourage mothers’ labor force participation (and, they hope, increase fertility).
* I didn’t bother with significance tests because these were arbitrarily small subsamples from each census; we could always go test them with the full samples.
** I could test a total motherhood effect, like Nieuwenhuis et al. did, but in almost all of these are samples 80% or 90% of women have children, so the kid/no-kid comparison is not as salient.