Finishing education and fertility

An analysis of NLSY97 data showing that education might be bottlenecking fertility

Note: This is a repost of a post that was previously published on my website.

One of the big trends in fertility patterns across advanced economies is that of “postponement”, i.e. having children later and later in life. According to Our World in Data, the average age of mothers at first birth in the US (where my data analysis will focus on) reached a low of 22.2 years in the 1960s and has risen to 27.7 years in 2021. Compared to other high-income countries, this is still relatively young. For example, South Korea now has an average age at first birth of over 32 years and in many European countries, it is now over 30 years.

I think there are three good reasons to try and slow down or reverse increasing postponement of fertility: First of all, we know that the ability to have children decreases strongly with age, especially for women (Geruso, LoPalo, and Spears, 2023). That means that whenever the average age at first birth rises, there will be more couples who want children who will suffer from infertility. Second, postponement might also contribute to falling birth rates overall due to the decrease in fecundity at older ages and the limits that this puts on family sizes. Third, the older people have children the less of their life they will share with their children and especially their grandchildren. It also makes it harder for your children to have children in the first place if you are less able to support them with childcare.

Addressing the postponement of childbearing is of course not easy because there are many complex reasons for its increase. From gender equality, to educational attainment and dating culture: Many aspects of life have changed considerably in the last 60 years. Additionally, there can be a self-reinforcing effect of trends like later childbirth because they are also considerably influenced by social norms. If all your peers have their child later, you will also be more likely to have a child later which further reinforces the norm.

One aspect that I have recently seen discussions on is whether we should try to make education more efficient so that it is easier for young adults to have children. Longer educational trajectories are probably a main driver of postponement (Ní Bhrolcháin and Beaujouan, 2012) and it’s intuitive that more time in education means that it takes longer to reach financial stability and also delays feeling like a “proper adult”. I realised that I don’t actually know what the patterns of childbirth around education completion are and was curious to take a closer look. Basically, is it the case that people really are “bottlenecked” by still being in education and then fertility jumps up discontinuously when finishing? Or is it more that increased education has contributed to the shift in norms but it’s not as directly tied to when you start having children? For example, we could imagine we have shifted to 30 being the “right” age for having your first child and then it doesn’t matter that much whether you finish your degree at 25 or 26 years old.

In order to have a better sense, I decided to do a quick data exploration with the National Longitudinal Survey of Youth 97 (NLSY97). This is a younger cohort of the NLSY, with respondents born between 1980 and 1984. The most recent available wave is from 2021-2022 meaning that respondents are between 36 and 42 years old at the time. At this age not all of them will have completed their families yet or even had their first birth. However, I expect that most of them will have completed their education and that a significant fraction will have had their first birth. Even though we can’t analyse completed fertility, this should still give us a pretty good sense of how fertility is influenced by education completion.

The data has detailed information on education and relationship timelines, meaning that I can not only look at fertility but also cohabitation and marriage. The sample that I’m working with is roughly 6,700 respondents. Throughout the analysis I’m using the sample weights from the last survey wave to make the results more representative of the US population. Keep in mind that this is still just a quick and dirty data exploration and you probably shouldn’t draw strong conclusions from this. If you want more details on my methodology or even download the data yourself and play with my code, you can find everything you need in this github repository.

To get a sense of the data, I start looking at some descriptives and general patterns of fertility by education level. The graph below shows the distribution of the highest education attended. The most common category is 4-year college and women have on average higher education than men.

The next graph shows age at first cohabitation, first marriage, and first birth based on the highest level of education respondents have. We can see that women are on average younger at each life event than men and that the age at each life event is increasing with the education level. This is particularly pronounced for age at first birth where we can see the steepest increase with education.

Comparing the age at different life events within the same education group, we can also see that the age at first birth is younger than the age at first marriage for those with no college education or a 2-year college degree. For those who attended 4-year college or a graduate program, age at first birth is higher than age at first marriage although the gap between the two life events is bigger for the more educated group. For example, women who attended a 4-year college on average get married at 25.7 years old and have their first birth at 26.1 years old, implying a gap of less than half a year between getting married and having a child. Women in a graduate program are on average 26.8 years old when they first get married and 28.6 years when they have their first child. This implies a larger gap of almost 2 years between getting married and having a child.

We can also see the overall postponement of fertility when looking at age-specific fertility rates (ASFR). These count the number of births in an age group per 1,000 people who are in that age group. We can see that for both men and women, the ASFR curves shift successively to the right with higher education levels. At older ages, fertility is then actually higher for the higher education groups. This is what demographers call the recuperation phase because those who postponed earlier on are then catching up on fertility later

The above patterns follow what we would expect from the existing literature on education and fertility. Now we can finally turn to our main question: Does finishing education make fertility jump up? The graph below shows that indeed there is quite a discontinuous jump! Fertility seems to plateau in the years before finishing a degree and then jump up considerably within the first two years after degree completion. The graph shows a fitted smoothed probability density function which you can think of as a smooth histogram. We’re looking at the distribution of first births relative to the time of education completion and the density tells us when first births are more or less likely to occur.

Doing this separately for all three education levels, we can see that the trend we identified is present for all of them. The effect seems strongest for those who attended a 4-year college. The probability of having a child is very flat before degree completion and then jumps up by a lot very soon after finishing education. For those who attended a 2-year college, the peak after education completion is smaller. We can also see that for women, there is actually a relatively high probability of a first birth before education completion that goes down in the two years before (which is roughly the time of actually doing the degree). For the highest education level, a graduate program, the peak after finishing is also present but there is already a continuous rise in the probability of a first birth before education completion

These results seem to point to an actual bottleneck that is created by education, with fertility rising immediately and discontinuously after that constraint is lifted. This is confirmed when looking at the patterns of first marriage relative to education completion. Especially for those in 4-year colleges and graduate programs, we see much more of a continuous rise in the probability of getting married before degree completion. The highest probability is still around the time of education completion but the fact that there is much less of a discontinuity implies that being in education is not as much of a barrier for getting married as it is for having a child.

Overall, it looks like policies that make education more efficient and therefore reduce the time until degree completion might actually be a promising tool in decreasing postponement. Given that fertility rises so discontinuously after finishing education, an earlier degree might lead to a shift forward in first births.

Of course, these are only descriptive results and can’t actually establish causality. (I’m working on that for a research project in which I exploit a quasi-experimental setting in Germany where the length of high school was reduced so stay tuned for a paper on that in a year or so!) The patterns that we see here are all the results of endogenous choices. For example, someone who wants a child earlier in life might decide to not pursue a graduate degree because of that. That means that the results above don’t necessarily apply to a situation where the length of degrees suddenly changes.

Another thing to keep in mind is that the best way to deal with postponement might not necessarily be to shorten educational trajectories but instead to make it easier to have children while still in education. It’s also important to note that social norms would have to shift as well. As long as it is considered “weird” to have a child in your early or mid twenties and nobody in your peer group is doing it, it’s unlikely that shortening degrees would decrease the age at first birth much. A completely different angle to the problem is that we shouldn’t try to “fix” postponement at all and instead work on technological solutions like innovations in egg freezing and fertility treatment that would make it a lot easier to have children at older ages.

Thanks to Ellen Munroe for the discussion that sparked this data analysis and to Denise Melchin, Lauren Gilbert, and Jeff Fong for helpful comments on an earlier draft of this post!