Thread by @lymanstoneky

There's an #NBERday paper out today arguing that because COVID interrupted access to contraception and abortion, low-income women's fertility may have actually RISEN in 2021.

I... am skeptical. Let us discuss. nber.org/papers/w29722

So first off, this is not a study of, say, vital statistics by education, or something like it proxying for social class or income. We have that data for 2020, and we have it for 2021 for a few states, so we'll get to it momentarily.

Rather, this study uses data from Planned Parenthood centers in Michigan + a longitudinal study of contraceptive usage by low income women in Michigan to directly look at how contraceptive usage changed during COVID.

Here's a description of their method. You'll notice there are a *lot* of parameters in here estimated from empirical data where there is definitely a lot of variance.

That's a lot of assumptions. So for example, they assume that abortion ratios were stable or perhaps even fell, i.e. that abortions got less common. Of course we actually *have* abortion data for 2020 for many states now and that turns out to be wrong: abortions rose in 2020.

For all the hue and cry about abortion clinics being closed, it turns out that *increased demand* for abortions overwhelmed any effects of *reduced supply*.

Similarly, several PMA rounds abroad have found contraceptive usage *increased* in 2020 in poor countries!

What's going on here is a story many in this space deeply believe, namely, that there is huge pent-up demand for contraceptives and that the main reason women don't use them is barriers to access or costs.

It will be no surprise to learn I disagree with this view.

My view is that in *most* (not all!) cases and countries, most people actually do have pretty low-cost (and often no-cost) access to various kinds of contraception, and non-use is usually *more* related to differences in knowledge and attitudes, i.e. demand, not access barriers.

Which explains why contraceptive use rocketed upwards in poor countries in 2020 despite tons of public health programs being closed: because *demand* rocketed upwards.

I've discussed this in great length for the international context elsewhere.

This paper is about the US context. Since the US has free distribution of condoms by almost every county health department in the country, many school systems, and...

... has low-cost or, on many insurance plans, completely free oral and sometimes other kinds of contraceptives, I am very skeptical that shocks to supply, like clinics being open or closed, are a big driver of contraceptive use today.

They may have been 40 years ago! But today, I'm very skeptical. But this paper, basically using a kind of "proximate determinants of fertility" framework, says that shutting clinics boosts fertility.

How do they get that result?

Well, they show that on average women who visit a Planned Parenthood office adopt more effective kinds of contraception. Using standard estimates of contraceptive failure rates, this yields a possible predictable number of births.

Now, they don't deal with endogeneity. Maybe women at Planned Parenthood don't adopt better birth control because they went to Planned Parenthood; maybe they went to Planned Parenthood *because they wanted to adopt better birth control*.

The authors have prior work using exogenous tests of clinic opening and closure and contraceptive uptake focusing on as I recall the 1960s-1980s, but it's not clear how to walk those estimates onto microdata today.

Okay, so, that's their argument. Visiting Planned Parenthood leads to more effective contraception, so closing Planned Parenthood leads to not-as-much adoption of more effective contraception, which leads to more contraceptive failure, which leads to more babies.

The authors suggest this will lead to just a 1% decline.

But from what baseline is unclear. From 2020? That's a bit tricksy since *2020 was already impacted by COVID*, as we now know, via increased abortions and other dynaamics.

Or is it *from some baseline accounting for economic conditions?* Maybe, but also a bit tricksy since they don't provide any formulas for that baseline, and the "Click here for the online appendix" link.... appears to be not actually a link? I can't seem to locate said appendix.

But look, quibbling over model parameters is boring.

Let's ask a simpler question:

Have births for poor women changed more NEGATIVELY or more POSITIVELY than for other women?

Just offhand, I know of a few states that provide data on this question. We'll start with Florida, which has data for all of 2021 out now. We're going to use "birth shares," i.e. the share of births in Month X occurring to mothers of Category Z.

When birth shares RISE, then a group has had MORE POSITIVE birth changes than the average (i.e. more babies). When they fall, it's more negative.

This paper predicts that the birth share to poor moms should RISE: they should have lost contraceptive access and had more babies.

Because richer moms have more options for access, richer moms should be less impacted. So the theory here is poorer moms should RISE as a share of births in 2020 and 2021 vs. 2019.

Now look, this is a dual-axis graph, but they are scaled identically just with different zeroes, and purely because it's hard to show them both at the same time another way. But broadly, we don't see poor women rising as a share of Florida moms.

In fact, in Feb-May 2021, both less-educated and also WIC-receiving women had a considerable *decline* in their birth shares, suggesting poorer women had a *bigger* decline in births than richer women!

So actually, rather than rich women successfully avoiding undesired births and poor women losing contraception and thus having undesired births, it's the opposite story!

Rich women had more kids, poor women fewer!

I won't belabor this too much but if I break out by detailed educational group, it's basically linear: the more educated the group of women the more positive their birth trends. 2021 was a gangbuster year for Florida women with PhDs to have babies!

And the reason for this is absolutely blindingly obvious:

Poor women in service sector jobs poor the brunt of COVID.

Richer women in white collar jobs went remote, spent more time at nice homes with family, etc, etc.

Other research from before COVID suggests remote work increases fertility along the educational gradient, so that seems to be what happened here.

But the point is, DEMAND drove the story, not access to contraceptives.

Also, remember, we actually have data on abortions in 2020, and we know that 2020's abortion rate was the highest in about a decade. So the idea that access *actually declined* is perhaps suspect to begin with.

But maybe Florida is weird. Do other states have the same trend?

Well, we can look at New Hampshire data too. It's formatted a bit different and they don't give WIC status so we just have education.

But look, same story! In spring 2021 (the dark red line), less-educated women make up an unusually SMALL share of New Hampshire births. Far from a boom in unintended births, we actually see that poor women avoided these births MORE than usual!

Unfortunately, that's *about it* in terms of birth microdata for 2021 with the necessary detail to explore this theory. Few states report this stuff in a proactive way.

But the key thing to understand is that the authors made a fairly explicit prediction: interruptions to clinic services would reduce effective contraception use and thus increase births to poor women in 2021.

We can test that. It seems... not right so far?

Which perhaps suggests that "going to Planned Parenthood" is not causal of "adopting contraception," but rather that "desiring to avoid conception" is causal of both.

Gotta draw them DAGs!

What happened in 2020 was almost all women acquired compelling reasons to avoid conception in spring 2020. But by summer/fall, richer women benefiting from remote work acquired reasons to think it was actually a very good time to have a baby, whereas poorer women did not.

As a result, beginning in March-May 2021, poorer women fell considerably as a share of all births.

Over time, everyone's fertility is returning to prior trend, with poor women on a slower trajectory.