Nobody’s home.

Where Is Fertility Low, and Since When?

Some Basic Statistics About Having Babies

My specialty is migration. But you can’t do one piece of a demographic balance sheet without at least a passing professional familiarity with the other pieces. So today I’m going to present some basic descriptive statistics about fertility in the U.S., mostly because my recent post about Hartford seemed to surprise people on one key point: Connecticut’s low fertility rate. I’ll be honest; it surprised me that CT’s low fertility was surprising to people. It shouldn’t be surprising. The Blue Tribe places a lower priority on producing offspring than the Red Tribe, so, controlling for age differences as TFR does, Blue States should usually have lower fertility. The exception is that immigrants of have higher fertility than natives of either tribe, and many immigrants live in blue states. But my suspicion is that immigrants in red states may even have higher fertility than immigrants in blue states, but that’s mere speculation.

So. Let’s get started. To begin with, let’s look at just crude demographic balances. Here’s some estimates of births/deaths in the USA since 1900.

Demographics purists will note that I’m representing the crude birth/death rate as a percentage here, rather than per 1,000 people. That’s a purely stylistic preference; the mathematical significance is identical.

The key point here is birth and death rates are both falling in the long, but birth rates are falling faster, and death rates may have reached their bottom. Given the frailty of the human frame, death rates cannot decrease forever, and they’ve mostly fallen thanks to lengthening lifespans for some very large generational cohorts. As those big cohorts die off, death rates must rise for a period of time. Birth rates, meanwhile, have no theoretical lower limit; I mean c’mon we’ve all seen Children of Men.

But the point is, in the crudest demographics terms, there is no fundamental “replacement rate” of fertility. Rather, as long as more babies are born in a given year than there were deaths in that year, voila, you get population growth, no matter whether fertility or mortality are individually low or high. And, as you can see, U.S. births have far exceeded deaths in every year since 1900, although the percentage point gap was the smallest in 2016 that it has ever been.

Okay, so the U.S. has plenty of babies, but that’s changing quickly as mortality rises and natality falls. But, for now, we’re fine. Right?

Well, it’s a bit more complicated than that.


Different Measures of Fertility

There are various ways to measure fertility. The most common statistics are either the crude birth rate or what’s called the “Total Fertility Rate,” or TFR. TFR is a hypothetical figure. Basically, each year, you look at the share of women having babies for each year of age, estimating the “natality risk” for each age bracket. Then you look at today’s 15 year olds, and extrapolate out, if current age-specific natality rates remain constant, how many babies will that 15-year-old girl have over her whole life. It’s a neat stat because it controls for age differences, so the population getting older won’t cause the birth rate to fall, unlike the crude rate, where population composition changes can cause major changes.

So let’s compare TFR back to 1960, the longest time series I have on hand right now, to the crude birth rate.

Now, let me be clear, though these units are similarly scaled, they have different meanings. Crude birth rates are babies divided by population. TFR is the hypothetical number of babies that will be born by a woman who turns 15 in the specified year over her whole lifepsan.

As you can see, the two trends do have some co-movement, especially through the 1980s. Since then, TFR remained about stable as the crude rate fell, reflecting a larger and larger share of the population that wasn’t of reproductive age, either because they were too old (boomers) or too young (echo boomers, meaning Millennials). But as Millennials have come of age to have kids, TFR has fallen sharply.

But hold on. TFR is hypothetical. What if what we want to know isn’t hypothetical lifetime births, but actual lifetime births? This is my favorite measure of fertility, which we called Completed Fertility (CF). Until 2012, CF the Census Bureau surveyed women about CF, asking women aged 40–45 how many kids they’d had in their whole life. In 2012 and after, thanks to more late-in-life pregnancies, they bumped up to surveying women 45–50. My time series below uses the highest age definition available for each year, and fills in gap years with simple linear extrapolations.

Completed fertility looks different. It peaks way after the others, and its been more stable since. What’s going on?

Well, simple. Women who are 45 in, say, 1960, had, on average, 3.654 kids. But when did they have those kids? Not when they were 45! Probably in their late teens, 20s, and 30s. So the “completed” fertility reported in a given year should reflect TFR in years decades earlier. And crucially, because TFR reflects just “that year” figures, the timing of births can cause TFR to jump around more erratically. A recession can supress TFR, even if those women are just shifting a birth by a few years. So completed fertility helps us see what womens’ actual experience was.

But saying “women who hit middle-age in year X had Y babies” is kind of boring and hard to conceptualize. Let’s do this another way. Let’s approximate, using the 5-year windows specified by the Census data, birth-year-specific completed fertility rates, and then line up TFR with those cohorts’ plausible peak fertility years, and see how accurately TFR captured realized fertility. This is a hit-or-miss activity because I’m getting approximations of specific birth-cohort CF, and because don’t have direct measures of when those womens’ actual births occurred with respond to historic TFR. But this should give us a first approximation of how well TFR predicts CF. Also, note that because of how I compute my cohorts, some years at the beginning and end have incomplete coverage in the data, so I’m less confident about them. Additionally, I mark where Census Bureau changes in the top-year of completed fertility could impact results.

I also mark a very important factor: the replacement rate of fertility, or about 2.08 kids per woman. When fertility measures like TFR or CF are below 2.08, it means that, in the long run, mortality is likely to exceed natality, and thus, without immigration, the population will shrink through demographic decline.

As you can see, my TFR measure time-shifted and averaged out to coincide with expected peak fertility for each birth cohort roughly corresponds to their experienced fertility. However, TFR still overstates fertility of women born before 1940, and understates the fertility of women born from about 1940 to 1960. In other words, TFR failed to capture some important changes in when fertility occurred across these womens’ lifespans. For women born after 1960, TFR during their peak fertility years has tended to overstate experienced fertility.

But the most important takeaway from this chart is that no generation of women in America has birthed above-replacement-rate since the women born in the mid-1940s.

That is to say, declining fertility isn’t new. When TFR got close to replacement rate in the mid 2000s, it wasn’t necessarily because women were actually going to have more babies across their lifespan, but maybe was just time-shifting. However, we don’t have full completed fertility data from after the 1966 birth cohort, and I’ve got no data for those after 1971. Maybe completed fertility will show some divergence from what the estimated TFR in those birth cohorts’ peak fertility years would suggest. It doesn’t seem likely that completed fertility would exceed TFR by enough to get over replacement, but maybe!

Broadly speaking though, this raises an important point. American population growth has been driven by immigration and by lifespan-lengthening health improvements, not by above-replacement fertility. Women in America have been experiencing below-replacement fertility for at least two or three generations.

We can break down this data by some different categories. Here’s estimated completed fertility by birth cohort, broken out by lifetime marital status.

As you can see, there are several interesting trends here. Most obviously, never-married women have way lower lifetime completed fertility than ever-married women. But while the gap is large, fertility for never-married women is rising steadily. Unmarried childbirth is less stigmatized than in the past, abortion rates are declining since the 1980s highs, and women who might have been married in the past are opting out of marriage now, but not out of childbirth.

But the other very interesting trend is that married fertility has actually managed to get above replacement rate for some cohorts. Married women born from about 1963–1970 managed to reproduce at or above replacement rate. Kudos to you, women of Early Generation X!

But of course, despite rising fertility for never marrieds and near-or-above replacement rate fertility for ever marrieds, fertility has fallen. Why? Because the never-married share rose! If we chain the marital demographic composition to that experienced by the 1932-birth-cohort, we get substantially different realized completed fertility.

As you can see, the marriage-state-chained estimate is substantially nearer replacement rate than the actual completed fertility. The dotted line is total completed fertility estimated from the same sample from which I derive marital status; slightly different from my main sample. As you can see, the marital-status sample actually has slightly lower fertility in many years, so the marital adjustment is could be bigger than just the comparison to my headline figure may suggest.

Adjusting for changes in the marital status composition of the population accounts for more than half of the gap between completed fertility and the replacement rate.

Now, a caveat. It’s possible that whatever social phenomena could have kept people married would have caused married people to have fewer kids. That is, maybe if people who selected into the never-married category had remained married, they would have had fewer kids, more similar to the never-married group, anyways. That’s entirely possible. I’m willing to believe a pure marital status adjustment overstates the impact of the rise of never-marrieds. But it seems almost certain that higher incidence of marriage would create some significantly higher realized fertility.

I’d love to break this data out by native/foreign born as well, but CPS doesn’t make much of that data readily available. Completed fertility is only available by nativity status from 1994–2010.

As you can see, foreign-born women have above-replacement-rate fertility for all birth years, while native-born have below-replacement. Native-born are falling, foreign-born were rising at least as of those born in 1970.

Unfortunately, Census Bureau hasn’t updated native/foreign born completed fertility since 2010 and thus can’t give us comparable statistics reflecting completed fertility up to age 50, so we can’t make a time series to compare to our total realized fertility. But we can do a decomposition by components for native/foreign fertility completed as of age 45, showing how much of that difference in fertility is accounted for by nativity.

As you can see, without the increased foreign-born share of women since the 1949 birth cohort, completed fertility, the orange line, would have been appreciably lower. The effect is not as large as the effect of changed marital composition, but it’s still meaningful.

Without rising immigration, completed fertility likely would have been about 2% lower, which means about 8% further below replacement rate.

Now, a caution. My method here is two totally separate estimates. You can’t just add up marriage and immigration and say the combined effect was X%. I don’t know how immigration interacts with the marriage rate. It’s also possible that if the immigrant population had not risen that immigrant fertility would have followed a different trajectory, just as was the case for marriage. But that seems less likely.


Where Is Current Fertility Low?

Completed fertility is a life-cycle measure. It’s hard to pin it to a location. Even for national completed fertility: immigrant mothers may have had their children prior to immigrating, so it would never show up in the USA’s TFR. We definitely cannot meaningfully assign completed fertility to state geographic areas based on where women live when they’re 45 or 50; childbirth may have occurred in numerous other locations.

But we can assign total fertility rates (TFR) to states. And indeed, it’s been done. Here’s a map of state TFRs in 2015:

If it’s not green, it’s got below-replacement TFR. As you can see, most states have below-replacement fertility rates. Exceptions include, of course, Utah, but also South Dakota, Alaska, Nebraska, North Dakota, Idaho, and juuuuust barely Texas.

Meanwhile, we can look at where TFR has changed, say, since 2007.

As you can see, TFR is falling pretty widely. Declines are not mere convergence either; the standard deviation of TFRs is roughly consistent across the period, and the coefficient of variation actually rises. The point is, this isn’t just high-fertility places undergoing a later demographic transition. We have decline everywhere, but the rate of decline is variable.


Conclusion

Fertility is low now, and has been low for quite a while. We’ve managed to keep births above deaths by extending lifespans and making deaths rarer. Without immigration, births would have been substantively lower. But even with high inflows, we’re well below replacement rate. Had marital demographics stayed the same over the last few generations, fertility would be much greater. But again, still below replacement, although it’s worth noting that fertility among women who have ever been married has mostly hovered around replacement, occasionally exceeding it.

There are many reasons for falling birth rates. Healthcare has gotten pricier, while wages have risen less. Education has become more necessary for a middle class lifestyle, so fertility is delayed and, in being delayed, is reduced. Norms and preferences about marriage and childbearing have changed as well. Assumptions about what degree of spending is “necessary” on a child have changed as well; affordable child-rearing has been increasingly regulated and stigmatized out of existence. Womens’ wages have risen, thus the economic losses to taking time off work have risen. The price of other forms of desirable consumption, like travel, entertainment, food, or apparel, has fallen consistently, meaning rational utility maximizers will shift spending away from “having kids” towards “going to Italy.” Extended families may have become less tightly knit, raising the cost of childcare. More individuals are abstaining from childbearing for ethical or political reasons, i.e. environmental concerns. Marriage is less common. Non-reproductive sexual activity has replaced much potentially reproductive activity thanks to more lesbian and gay individuals being free to publicly marry and love whom they wish.

I could go on and on with factors contributing to declining childbearing. But the point is, it’s low. Completed childbearing isn’t falling particularly fast, which suggests much of the shift in TFR may just be a timing question (though maybe not!). But crucially, the population free-ride we got from lengthening lifespans is running out. The nation’s biological clock is ticking. While environmentalists may look forward to a smaller American population for some reasons, anyone who values liberalism, democracy, economic dynamism, and mobility should look with some trepidation on a rapidly-approaching future with fewer Americans and less population turnover.

Check out my Podcast about the history of American migration.

If you like this post and want to see more research like it, I’d love for you to share it on Twitter or Facebook. Or, just as valuable for me, you can click the recommend button at the bottom of the page. Thanks!

Follow me on Twitter to keep up with what I’m writing and reading. Follow my Medium Collection at In a State of Migration if you want updates when I write new posts. And if you’re writing about migration too, feel free to submit a post to the collection!

I’m a native of Wilmore, Kentucky, a graduate of Transylvania University, and also the George Washington University’s Elliott School. My real job is as an economist at USDA’s Foreign Agricultural Service, where I analyze and forecast cotton market conditions. I’m married to a kickass Kentucky woman named Ruth.

My posts are not endorsed by and do not in any way represent the opinions of the United States government or any branch, department, agency, or division of it. My writing represents exclusively my own opinions. I did not receive any financial support or remuneration from any party for this research.