Thread by @lymanstoneky

This is a very cool paper doing really excellent work. However, I think the author's interpretation of the results may be slightly off.

https://twitter.com/florenciatorche/status/1382341959528837120

So first off, I don't have access to the full text, so I'm working from the very nice presentation that @florenciatorche gave here:

Midway through, a questioner asks if she has data on fathers who are absent. She says she does not have data about cohabitation, and shows a graph showing that the cohabiting share of births to single moms is probably more-or-less stable.

This matters because @florenciatorche is arguing that the change in infant outcomes is *due to* normativity, NOT selection factors. Showing no-change in union structure among unmarried moms is a good sanity check on selection issues! It's the right thing to do!

But what the questionner actually asked was not just cohabitation, but *father present on the birth certificate*.

And as she shows at about 48 minutes in, that data DOES exist!

When we think about single moms, there are several ways to divide them.

One is "non-cohabiters" and "cohabiters."

But that's not the only way.

ANother way is "moms with kids *who have a legally acknowledged father*" and "moms with kids *with no legally acknowledged father*".

Legal acknowledgement may matter a lot. I dunno Chilean rules, but in the US it matters for child support, alimony, etc.

Furthermore, kids born with no legally acknowledged father probably represent "core single motherhood." There is no father in the picture *at all*. Non-cohabiting-but-acknowledged dads (i.e. non-custodial fathers) might have a big influence on kids' lives!

So at 48 minutes, @florenciatorche shows a graph I can't show here because screenshotting YouTube is grainy, but she shows there has been ***no change*** in the gap in infant outcomes when comparing married moms to moms ***with no paternity listed for their baby***

The argument made is that because *unacknowledged paternity* has not changed as a share of *total births* (no change in normativity), we should expect no change in infant outcomes for this group.

On its face, that's reasonable!

But to a first approximation, the massive changes that the World Values Survey shows in Chilean attitudes about single parenthood should have impacted those infants *just as much if not more* as infants of single moms who DO have legal paternity established.

Why would a shifting norm about single parenthood only improve outcomes for kids ***who have two legally recognized parents***?

There was a massive decline in the share of births to unmarried births without paternity. No-established-paternity births rose from 31% of unmarried births at the start of the sample to 12% at the end.

Okay, now, somebody with journal access sent me the paper. And here it gets weird.

The presentation says they compared those with unknown fathers to married people.

The paper text says they compared unknown fathers to known fathers.

Which thing they compared might be consequential.

But I think the mere fact that there *was* a large change in paternity acknowledgement among unmarried is probably indicative of selection factors being important.

Here's what we can say.

In Chile, mom's marital status became less predictive of neonatal health outcomes over the same period where single parenthood was becoming more normative.

At the same time, the composition of single motherhood did change: there was a large decline in unacknowledged paternity (and also btw there *was* a small increase in cohabitation as a share of unmarried moms).

There was NO convergence in neonatal health outcomes between babies with no acknowledged paternity and other moms. The penalty for being *fatherless* did not change over time.

When we don't know is the combination. How did unmarried-but-acknowledged-paternity change vs. married? Because THAT is actually closer to a proxy for "benefits accruing to formal marital status."

Finally, just a note: measuring differences is a good way to capture an effect.

But we should probably at least pay attention to levels too.

And what happened is NOT that unmarried neonatal health improved to reach the level of married.

Married neonatal health *got worse*.

What you'll notice is preterm delivery and low birthweight got more common over time.

Now, when we look at "small for gestational age," which is a nice vector of the two, we do see improvement in both cases.

But 1) size is not the *only* problem with preterm delivery and 2) what this is probably *actually* showing is is that there was a change in *stillbirth and miscarriage rates*.

That is, if you delivered babies earlier, but they were on average at a HEALTHIER weight-for-age, it's plausible you may have just been doing things like medically indicated C-sections to deal with common causes of late-stillbirth.

But if you have a change in that sort of thing, then you have to wonder if excluding stillbirths from the sample might be missing part of the effect channel.

I happen to know from other studies that Chilean access to contraceptives experienced several major shocks during this period, which DID cause changes in stillbirths, so it's not implausible that there could be an informative trend here.

All that to say:

This is all really excellent and interesting work! I have no critique at all of methods or approach; I think it's a great study!

I just don't think it shows that if we all think more nicely about single parenthood that not having a dad won't hurt kids anymore.

And of course the measure of "normativity" here is actually not a measure of attitudes at all but just prevalence of behavior, so they aren't technically arguing that either. i.e. it could be structural shifts in decisions hospitals make based on unstated beliefs about risks!

But the casual citation of "norms" in the wider literature links them to beliefs and attitudes, and the authors specifically note the importance of ideational changes, so it's not absurd to think the intention is to talk about that stuff.

Oh and I tagged @florenciatorche because she was who I saw the presentation by, however I should also tag co-author @aleabufhele