In Part 1, I showed that the clock is ticking for guys who want to start families too, and that childlessness isn’t just something that happens to other people. It is quite likely, if you are an unmarried twenty- or thirty-something, to include you.

I pointed out that, because average relationship age gaps are small and perpetual bachelor-types usually aren’t the ones who land the large age gaps. Because of this, men should feel the same urgency as women when it comes to forming a family.

But what is that level of urgency?

A 2015 study gave estimates as to how old a woman needs to be upon starting to try for children. It doesn’t factor in male age, but this matters too! Male age does reduce fertility, see Part 1!

But anyway, let’s assume the guy is roughly the same age.

Here I’ve made a graph that shows how to give yourself a 75% chance of having either two or three children:

But what does this mean for preparing to have a family?

Under our current norms, it takes time to build a family.

If you or your prospective spouse are sufficiently like the average person, you likely think that a reasonable amount of time to have been together [in this thing they call the romantic relationship] is two or three years (according to polling). There is good reason to think this is slower than optimal! As the phrase goes, you know by six months (and if you don’t know, then you know). But I’m drifting away from a firm evidence-backed position here, so let’s get back to the stats.

Assuming that you, like the average person, will want to wait two to three years before committing to your partner and trying for a baby (in contravention of historical norms and in the absence of evidence that this new norm of prolonged courting is optimal), this means the following:

If you would like to have a 3 children (and have a 75% chance at achieving this goal): you should be with your spouse/future spouse at 28 years old.

If you would like to have 2 children (and have a 75% chance at achieving this goal): you should be with your spouse/future spouse at 31 years old.

Finding a spouse

But does that mean you can afford to wait until your late twenties to find someone? Unlikely! Of course it would be incredibly risky to wait until your late or mid twenties to start looking for a spouse, because the timelines are tight and if the relationship you enter at 28 doesn’t work out this will likely have very real effects on your future-family.

You probably also won’t be very good at communicating and attracting the kind of person you would like to be with, so this should also be accounted for when thinking about timelines.

And, unfortunately, IVF only adds one or two years of extra time! Even when using IVF, a woman should start trying by her early-to-mid thirties to have a 75% chance of having two or three kids.

In any case, IVF is expensive if you don’t meet the standards for public healthcare availability (which you probably won’t in the UK if you already have one child) and also carries harsh emotional costs.

To give a more complete and accurate picture of what the stats/ numbers/ data/polling show I should add that many people end up, after becoming parents, wanting more children than they originally thought they would.

So this might mean that you should give yourself time and space for the possibility of having another child after you have had your desired amount. It seems like most people didn’t know exactly what they wanted before having children!

When you consider that women graduates only consider the timing of having children a pressing issue at the age of 33, it is no wonder that so many are on course to be childless against their wishes.

How many women who are currently 25 or 30 years old will end up becoming mothers?

Rates of motherhood will drop to unprecedented lows. Women who were 30 years old in 2024 (the last year with available data on HFD) who have not already become mothers faced a one in four chance of ever becoming mothers according to this model. Those who were 25 years old faced roughly squarely between a one in five and a one in four chance.1

You can find more sensationalist figures which claim that a woman who gets to 30 years old has a 50:50 chance of having a child. Those figures are true historically. My projection is more optimistic because it inputs changing age of mothers at first birth, i.e., it takes into account that family formation is getting delayed anyway (I include descriptions and graphs for my projections in the last section of this piece). This includes good news and bad news: in the good news it looks like the numbers are less dire than a straight read across of historical rates would suggest; in the bad news, it looks like family formation will fall to unprecedented lows even after accounting for changing timing. And the risks isn’t that much lower for 25 year old women than 30 year old women precisely because the trend is downwards.

You might be thinking “well, no one in my circle is even coupled up”. But consider that many young people are coupled up. In the US 55% of people aged 25-34 year olds were living in a couple in 2023 (whether married or cohabiting). Clearly, those folks are going to be the ones disproportionately unlikely to end up in that one in four who don’t go on to have children.

So to those in their late twenties in social circles in which folks are not yet coupled up, I mean to say: this is now likely an indicator that many in your circle will never become parents.

Reading and thinking about fertility rates in the abstract is not a substitute for getting serious in one’s personal life!

These figures betray real human outcomes, and—since many people still want to have a family—real human tragedies.

The next baby boom

I said at the beginning of this piece that the proportion of people who become parents will fall to unprecedented lows. And it will, if the model continues to predict rates of motherhood well and given that TFR1 continues to fall. But there is one other cohort on record in which more than 18% of the women born that year never became mothers: 1918.

What happened after 1918? Women born from 1919 spent most of their twenties in the baby boom, complete with increased marriage rates, younger age of marriage, higher rates of parenthood and higher rates of going on to have 2nd, 3rd and 4th babies.

In England and Wales, Canada, Ireland, Czechoslovakia, Finland, New Zealand, Switzerland, Denmark, Sweden, Norway, Iceland, Spain, Japan, Italy, Portugal (you get the point), their baby booms can be entirely attributed to an increase in the number of people getting married. Marital fertility actually fell in all of those countries during their baby booms.

Given most of the decline in fertility in the developed world in the past ~20 years can be attributed to declines in coupling and rates of becoming a parent, a marriage boom today would likely also achieve a baby boom. This is in theory possible! It happened right after a low period in rates of parenthood and overall fertility.

But the marriage boom didn’t just happen; it came in the context of vast home building and improvements in medicine surrounding the risks of childbirth (to name just two likely factors). Similar societal changes today may well create such a change in conditions too!

In this case, the night could well be darkest before the dawn, but only if we, as a society, adequately respond to the challenges of the time. The challenges of the time include housing, but also technological shocks (including dating apps and social media in particular).

Model workings

My colleague Lyman Stone has found that the total fertility rate (TFR) in a given year is a great predictor of the average number of children who will be born to women currently 25 years old. We can do something similar for the number of first borns (and therefore the proportion of women who become mothers).

TFR1, the fertility rate of first born children, is a bit more tricky to work with than TFR, because it is quite volatile. People regularly bring forward or push back becoming a parent by a few years based on societal conditions, so TFR1 fluctuates far more (in relative terms) than the overall TFR.

We can average TFR1 over multiple years to smooth out some of this volatility. I found the best predictive results were found at smoothing over a 5 year window and lagging by 20 years (in the US data)… i.e., the TFR1 in a given 5 year window following women aged 18-22 years old can be used to pretty well match how many of those women become mothers [if the relationship maintains in the future].2

People are often worried about tempo effects screwing up the predictive effect of TFR in general, and TFR1 in particular (for the fluctuation reason I mentioned above). But the period examined above includes women who had children in the baby boom, during which time the average age of becoming a mother fell, and women who had children during the sexual revolution, during which time the average age of becoming a mother rose. Bar temporary slight over- or under-shooting, the predictive model ended up pretty closely fitting the proportion of women who become mothers.

However, another way of predicting things is to see trajectories in fertility rates of first births by age group and see how they add up. But this ignores the fact that, when people are delaying family formation, age-specific fertility rates may fall across younger adults for some time whilst being stable for older adults followed by a time of rising age-specific fertility rates in older ages. This graph below shows the possible beginning of steep rise in ASFRs in post-covid era.

An improved model would therefore consider how trends in age-specific fertility rates of first births (ASFR1s) and changing average age of mother at first birth have predicted the proportion of women who become mothers in previous cohorts. To level up the model one extra, it might also consider breakpoints in rates of first births (TFR1) and in average age of first birth, so that trends can be measured from breakpoints. This is important because we are in a period of transition, the US has gone through a major trend change in TFR1 in recent years, and the task is trying to forecast the future in a new reality. A model which uses a simple linear function of trends in ASFR1 since the latest trend-break in TFR1 and trends in average age of motherhood since its latest trend-break can also pretty well match rates of motherhood in past cohorts:

This model has slightly better fit to my simple TFR1-rates one. But the best fit is found when we take the average of the two! You can see each of the models, plus the average model, compared to what rates of motherhood ended up being in this graph:

The lines all follow each other pretty well! So, there’s not too much in it. But the average of the models does win the best fit by a sliver, so let’s use that. I also [very slightly] checked the models’ predictive power by seeing how well they fit the proportion of women born 1975-79 who became mothers by the age of 45 (which tends to be approximately the same as the proportion who became mothers by the age of 50 for obvious reasons), and a) they all performed pretty well, and b) the average still performed best.

So, what does it tell us about the percentage of current women who will become mothers? Here’s the projection given by the average model:

About 18% of all US women born in 1999 are projected never to become mothers. The orange line is a projection which relies on adapted carry-forward age specific fertility rate model with slightly less good predictive power (i.e., it’s better to carry-forward rates when women are 25 than from when they are 20), but the model still performed pretty well. It says that about 19% of all US women born in 1999 are projected never to become mothers.

Please note two limitations to this model:

• I assume that good predictive power in past cohorts will translate to good predictive power in present ones.

• Post-covid, trends could have changed. It’s too soon to identify if this has really happened, but if it has, that could make my projections overly optimistic or pessimistic.

I used HFD data on fertility rates by parity, cumulative cohort fertility rates by parity, completed cohort fertility by parity, and parity progression ratios for the fertility figures. I pulled US Census Burea data for the coupling rates.