Thread by @lymanstoneky

I like @jburnmurdoch but this entire post is wrong and detached from the scholarly literature. I appreciate the friendly cites to me, but the thesis is totally wrong. ft.com/content/838eeb…

First of all, the lack of citation to either Bergsvik review is a serious misstep, e.g. here: or here: onlinelibrary.wiley.com/doi/full/10.11…
papers.ssrn.com/sol3/papers.cf…

We also know that many of the studies Bergsvik reviews in finding that policy does matter were themselves not correctly designed, and in a way that biased effect sizes downwards: nature.com/articles/s4159…

Meanwhile, a retraction-bait piece is cited. Every actual scholar of time use data knows this graph is fake. It is based on a trivially small number of points and built into the model is the assumption that parenting changes over time are similar across countries.

I dunno if John is using the actual study estimates or did his own replication, but several of the study estimates can no longer be replicated because the cited data is no longer available at the cited location.

Also, it should be noted that consistency in time use data is low: countries ask it in very different ways and there are huge cultural changes in how people conceive of and report childcare time.

Now it's true parenting time has risen, I don't dispute that!

But actual policy studies using high-quality causal inference are passed over for a mostly-fake correlational study of mediocre data????

Finally, there's a deeper theoretical issue here.

Yes, culture is hugely important.

The government shapes the culture.

It's not ambiguous why marriage rates have disproportionately declined for poor people! It's because governments pay them not to get married!

Young people are disproportionately poorer and so more exposed to the marriage penalties in means-tested programs. This effect accounts for a very large share of the total marriage rate decline.

Because virtually every welfare state on earth punishes marriage, this also works well to explain the highly correlated decline in marriage across countries: they all rolled out similarly anti-marriage welfare states!

But the key point is that this "policy or culture" or "economics or culture" debate is just insane. Doepke, who is cited, wrote a whole book about how crazy this argument is! That book is cited in the article! Culture isn't a separate domain!

There is a deeply endogenous relationship between culture and policy. Consider Israel. We know for a fact from causally informed studies POLICY explains a VERY LARGE share of Israel's higher fertility, including among secular Jews.

But WHY DOES THAT POLICY EXIST?

Because a culture exists to give it a voting constituency.

But what does that culture exist?

Partly because of a policy that incentivizes its existence.

Efforts to untangle "policy" and "culture" are academically interesting but they miss the point: we can intervene on both and many places have done so quite successfully. Policies intervene on culture, not just cost-benefit tradeoffs.

It's true that popularly-tried pronatal policies have a limited scope of effectiveness. The recent IHME forecasts in the Lancet correctly noted that currently-common policies can't plausibly push TFR up by more than 0.2 kids/woman.

But

1) THAT'S A HUGE EFFECT! That effect size means the US would have tens of millions for people after a few decades!

2) Emphasis on "currently common policies." We can invent new ones or make currently uncommon policies common.

In particular, one reason pronatal policies might not be as effective as policymakers wish is because of cowardly subtlety. Policymakers can't just quietly nudge costs and benefits. They must stand in a visible location, wave a flag, and yell "WE'RE PAYING YOU TO HAVE BABIES!"

Cultural interventions, where I agree the biggest bang-for-the-buck exists, require the creation of compelling cultural narratives. That may not literally be flag-waving lol but the point is you have to actually give a compelling pronatal cultural narrative, in public.

So I have zero sympathy for the argument that policy doesn't work when 99% of the cases of policies we study are cases so small and so non-public that most of the "treated group" didn't even realize policies had changed, or else didn't think increased fertility was a policy goal

Now the big exception case here is Hungary. I've written about why I think Hungary hasn't gone the way people might have expected. We'll see what happens long term there.

As an aside, I want to note that childcare is a uniquely lame pronatal policy. The literature already yields a much lower Babies/$ tradeoff for childcare than cash transfers. And the reason why is obvious!

https://x.com/robinhanson/status/1773726888147476631?s=20

Free childcare subsidizes babies, yes.

But it also is a tax on babies. In particular, it's a subsidy (free) you only get if you pay a tax (less time with your kid). Because spending time with your kid is the point of having kids, and because the people most willing...

... to have more kids are people who would like to spend more time with kids, offering a subsidy only available to people willing to reduce time with kids is throwing money at the absolute least efficient part of the fertility decision tree.

There may also be a workplace attachment dynamic wherein childcare reduces one cost (cost of care) but increases another (opportunity cost of next birth), whereas "career exit to care for kids" internalizes a massive care cost but yields much lower future opportunity costs.

This sort of dynamic is at the heart of this re-analysis of maternity leave policies:

https://x.com/lymanstoneky/status/1773711518648930766?s=20

The key point is: yeah, even I who am enthusiastic about pronatal policy would agree that childcare spending is probably a very low-margin way to spend money!

I want to give credit where credit is do, talking directly with @jburnmurdoch , his intended point here is rather different than I and others read it: he's correct that currently-common pronatal policies are unlikely to effect long-term stability of birth rates.

Here's a mental model. I and others took @jburnmurdoch to be denying both the blue and red lines, whereas I believe he intended to be only rejecting the red line and affirming the blue one. But that affirmation didn't come through.

The key point here is that currently-common pronatal policies are mostly fiddling around the edges and none of them seriously tackle the underlying issue of skills-biased technical change. But they do increase fertility! They just do it mostly via a shift in level not slope.

I have a paper in review at JOPE showing an intervention with a HUGE level effect but not that much slope effect, and I have a paper in review at IFS showing an intervention with little level effect and a big slope effect, but slope effects are definitely rarer.

That said, slope effects are also harder to find. Level shifts are obvious and provoke closer scrutiny. Slope shifts are a lot less obvious, especially counterfactually.