Thread by @lymanstoneky

It's #NBERday! And we've got some good ones!

First off, I want to do something I don't usually do and mention that I just HAD to break my own rules and see who wrote something. As a rule, I try not to read who the authors of papers are and definitely not to click on their bio. #NBERday

But there has just been a spate of very cool and even-handed papers about welfare reform the last couple years and especially inter-generational effects, and I had previously noticed they shared authors, and today's does too, and kudos to that team! #NBERday

This group of authors (not always the same group but a lot of overlap) has previously found that welfare reform reduced minor crime rates for youth (but not serious crime), and was a "turning point" triggering falling female crime rates especially. #NBERday

So yay welfare reform!

But they have also found that welfare reform led to deteriorating parent-child relationships, worse behavior at schools, and ultimately worse educational outcomes especially for boys.

So booo welfare reform!
#NBERday

What's actually going on here is researchers systematically slogging through every bit of data they can on welfare reform and intergenerational linkages to try to paint a broad picture of what effect welfare reform had on a stated goal: ending a "culture of poverty." #NBERday

My read of the literature so far is that welfare reform *probably did* increase material wellbeing for the average impacted person. Effects were heterogenous, with women vastly better off than men. Intergenerational effects however were mixed at best. #NBERday

In general, welfare reform probably worsening a lot of "noncognitive" issues especially for boys, and did nothing to address serious drivers of deep poverty, but may have helped "cognitive" skills for some kids, especially young girls. #NBERday

So, today's paper:

Q: Does being exposed to welfare reform as a kid cause you to experience LESS food insecurity as an adult?

A: YES. Welfare reform reduced the odds a child grew up to become food insecure! #NBERday nber.org/papers/w29054

This effect was driven mostly be positive effects for women, and specifically for women who completed high school. It was driven by exposure *before age 5* as well. #NBERday

Sorry, good question. These are shorthand for meaning "stuff that you can't really track on a standardized test." Behavioral issues, attention and hyperactivity issues, skipping school, poor social skills, etc. #NBERday

https://twitter.com/freakyguisan/status/1419669757230473223

So if you have a program that mostly operates by taking poor women off the welfare rolls and nudging them into work, and it only helps kids if they're young girls who complete high school, and also you know this program shifted care environments.... #NBERday

Leading to more kids in institutional care at earlier ages, where we also know boys tend to do worse than girls *in general*, it really feels like the story here is just that we have a good explanation for male underperformance in the last decade or two. #NBERday

We did a massive experiment in shifting poor kids into institutional care environments and it worked out pretty well for girls and badly for boys. Same thing happened in Quebec. Surprise surprise, little boys don't thrive in the typical daycare environment. #NBERday

Note that I find it implausible that there's a role model effect operating here on 2 year olds that just vanishes for 10 year olds. #NBERday

I think this story will make nobody happy. Welfare reform critics will dislike the idea that welfare reform may have actually succeeded at reducing intergenerational transmission of some economic hardships. Welfare reform enthusiasts will dislike the family implications. #NBERday

For me, this literature all makes total sense. Heterogeneity in optimal care environment seems intuitive, and the tradeoff involved in getting women with little education into low-wage jobs instead of welfare seems pretty obvious! #NBERday

Was this tradeoff worth it? That's going to depend on your value set and the wider social circumstances! #NBERday

Let's stay on welfare!

One problem in welfare studies is we often end up staring at the same datasets over and over and over and you never know if the relationship you're observing is "general" or just "this dataset." #NBERday

That's why the research agenda I described above is persuasive to me btw: they show effects in tons of different datasets. #NBERday

I really love when papers develop new datasets. So....

Q: At the grassroots level, is it just more difficult to apply for welfare in some states than others?

A: Yes. There's huge variation in the length and difficulty of phone calls to apply. #NBERday nber.org/papers/w29055

These researchers had their RAs call every state's contact line for UI, income tax, Medicaid, and SNAP, and try to get a real person on the line to talk to. These graphs show, out of 10 RAs calling each state, how many successfully got a person for each program. #NBERday

Evidently, it's a ton easier to get a person to talk to you about your taxes than about SNAP! #NBERday

There's some correlation across programs within states but it's messy. But here's the overall number (out of 40 calls) that reached a real person.

I see this as sort of a cool measure of "quality of governance." #NBERday

So NH, WI, TN, MS, VT, NC, LA, AR do well. NJ, GA, CT, HI, IL, NM, MD do poorly.

It's important to realize this isn't a "policy" per se. But these kinds of barriers are hugely important in terms of who gets what benefits! #NBERday

Nor is this a digital/phone thing. The RAs also visited all the websites to see if they could locate a live chat feature, and how well it worked. States with crappy phone help did not have substantially better websites. #NBERday

More data like this pls! #NBERday

Keeping on this broad theme, it's impossible not to notice Heckman papers due to his distinctive style, so here's a Heckman paper where he says the Perry Preschool Program produces intergenerational benefits. #NBERday nber.org/papers/w29057

Just want to remind everybody though that Perry Preschool participants (i.e. the treatment group) are hopelessly nonrandom. Among the control group, 31% of moms worked. Among the treatment, it is ***9%***, because they let people swap some. #NBERday

So what they really studied was "what happens if kids are in a high quality preschool and have a nice lunch made for them by mom everyday and also she picks them up and asks how their day went and makes a nice dinner for them." #NBERday

I'm being sarcastic but seriously it's a bad enough problem they had to literally invent a new method and make up a new word to describe the problem in their new-ish paper trying to tackle the issue. #NBERday nber.org/papers/w27738

The most significant effects Heckman finds of course are for *marriage*. Treated kids (kids who did the Preschool, who remember were also 3x as likely to have a stay-at-home-mom) were vastly more likely to marry young and not do crimes. #NBERday

Here's the graph. You can see the marriage effects crop up WAY before the income effects, and even before the crime effects. #NBERday

So what happened is the Perry kids married younger, which kept them out of crime, which meant they got into a stable job, which meant they had higher income, so that when their kids were born, they invested more into their kids, so their kids did better. #NBERday

But did *preschool* make them marry more, or the huge difference in family structure? #NBERday

Honestly would love to see the results broken out by maternal work status. They show some results broken out by *paternal* work status, but not maternal! #NBERday

But that's not all! There's a second Heckman paper! And you know it's a Heckman paper because this is the interpretation of a simple linear OLS model of survey respondents: #NBERday nber.org/papers/w29072

Yesterday I said after people get a Nobel they should be prohibited from speaking again to preserve the dignity of the honor, in reference to Krugman.

Today, Heckman proves the wisdom of my view.

That's how a Nobel laureate FOR CAUSAL INFERENCE describes OLS. #NBERday

And they are interpreting it as causal btw. They refer to athletics as "generating" benefits, or benefits "flowing from." It's just straightforwardly claiming that an association is causal. #NBERday

Legitimately reads like an undergrad econometrics paper.

That's the only formula in the paper y'all.

They're using restricted-use data none of us can access and didn't report coefficients of *any* controls.
#NBERday

If you believe playing high school basketball can increase your adult wages by ***15%***, you might have a Nobel prize in economics! #NBERday

It appears the actual argument that they're making is that athletics are a vehicle for poorer and minority kids to get scholarships, and this improves their subsequent performance in life. But that isn't what they show! #NBERday

They don't show scholarship receipt.

Y'all they refer to an "Appendix C," when the paper doesn't even have an Appendix section.

They have this footnote.

What on earth. #NBERday

Seriously is this an undergraduate paper. #NBERday

NOPE. NOT AN UNDERGRAD PAPER.

An economic consulting firm!

Who got the idea for the article.... *literally from the NCAA's lawyers*!!! #NBERday

And when they say "the authors weren't paid for this"

what they mean is THE AUTHORS WERE ABSOLUTELY PAID:
compasslexecon.com/professor-jame…

#NBERday

Folks, the argument Heckman is making is that the benefits of getting scholarships worth LESS than what athletes would be paid are bigger than the benefits of getting paid. #NBERday

And he then published this paper (mostly a copy-paste of a court filing, hence the explaining-it-for-idiots tone to it) in NBER, laundering an academic reputation for it. #NBERday

In the meantime claiming no financial interest despite the fact that expert witnesses are usually paid and the co-author literally works for an anti-trust defense consultancy contracted by the NCAA. #NBERday

Y'all. This research is shoddy quality and also I'm not sure this is even ethical. #NBERday

I remind you that Heckman literally won a Nobel prize for causal inference connected to policy advocacy, and this is what he's doing. #NBERday

I am a paid economic consultant who works for corporations that not everybody loves. I have zero beef with that. That's fine.

Don't take your court filings helping them try to avoid paying black kids a fair wage and repackage them as economic research! #NBERday

I have had consultancy-research crossover. But that's why you have to go over the top on disclosure! I've had editors delete my disclosure statements as being excessive because I was identifying prior survey question draft wordings to show how backers influenced wording. #NBERday

But that's what you have to do to be conducting research ethically. Don't act like you're not conflicted! Communicate the conflict! #NBERday

Also does anybody know if Heckman is a paid consultant for early childcare programs and if so how much and is it disclosed in relevant publications? #NBERday

Again, zero beef with experts being paid consultants! Hi I'm Lyman that's my whole career!

It's when you act like you're doing academic research but you're literally rebundling your paid expert testimony organized by the NCAA's anti-trust defense team. #NBERday

If I were publishing a paper about like breastfeeding prevalence among mothers on WIC, I would probably mention that I am Extremely Paid To Have Opinions On This by companies sell infant formula. #NBERday

Even if they didn't pay me FOR THE PAPER, I literally cannot imagine publishing on that and just casually not mentioning that I have a huge financial stake in moving tons of that sweet sweet fluffy white powder to the kiddos. #NBERday

The worst part is HE HAS A NOBEL PRIZE. He can get paid $$$ just to like show up and say "Hello, economics is good, have a nice day."

He even wrote a SECOND paper here:
compasslexecon.com/wp-content/upl…

#NBERday

Anyways. If you didn't already know that Heckman's work was overrated from his laughable response to the cost-benefit literature debate or his eternal return to the Perry program, here's another example of Post Nobel Brainworms. #NBERday

Heckman: we should promote intergenerational mobility of poor black kids!

Also Heckman: unless it means schools have to actually pay poor black kids for their work. that's a bridge too far. #NBErday

Speaking of helping people...

Q: Can giving local communities institutionalized democratic control of "development" funds give good long-run development outcomes?

A: Yes.
#NBERday
nber.org/papers/w29079

So, backstory here. In the 2000s, responsive to a lot of critiques of conventional big-budget development programs, donors like the World Bank started experimenting with "community driven development" (CDD). #NBERday

So you can read an exhaustive (and exhausting!) account of an early such project in Indonesia here: amazon.com/Contesting-Dev… #NBERday

There have been a lot of these programs running now. So how do they work?

In concept it's simple. Instead of the government hiring a bunch of engineers to build a dam, they hire a bunch of community organizers to set up meetings. #NBERday

These organizers set up structured meetings at neighborhood, village, county, or higher levels where people collectively decide on what kinds of development projects they would like to have. They also get allocated a budget for such projects. #NBERday

In some settings, they are collectively deciding on a *submission to a competitive grant process*. In some settings they're allocating a budget. Doesn't make a huge difference for the purposes of this discussion. #NBERday

The point is, the government (or international donors) spends a bit less money on a big project, and instead spends some money on building institutions. Donors usually demand these institutions include special processes or quotas to include marginalized groups. #NBERday

This is all part of the big shift in economic development think towards INSTITUTIONALISM. Basically, nowadays experts think long run prosperity comes more from "participatory and inclusive institutions" than from "big dams and good roads." #NBERday

So the theory is if you build a bunch of roads and string a bunch of electric wires but you don't address peoples' ability to assert their rights and make their voices heard politically, you won't really get very far, developmentally speaking. #NBERday

Big names associated with this big view of development might be folks like Darron Acemoglu, Amartya Sen, or William Easterly. They're all making variants of this kind of argument and have been major critics of donor-led development focused on the BIG PROJECTS. #NBERday

These institutionalist thinkers basically won the argument and so while Big Projects continue (esp. financed by less scrupulous actors like China: lowkey this shift did open the door for China to step into the World Bank's shoes), participatory approaches are growing. #NBERday

The only problem is that early reads of these projects often didn't find very good effects. The book above about an Indonesian project found some good stuff, but the project just didn't deliver the big gains exponents hoped for. #NBERday

Likewise, studies of big projects in Sierra Leone and the DRC have also turned up disappointing results.

So what gives? If all the theory says BUILD INSTITUTIONS is what works, how come institution-building doesn't really seem to work? #NBERday

This article provides an answer. It's by the same authors who previously argued Sierra Leone's project didn't turn out very well. But now with an extra 5 years of follow-up data and improved methods, they find.... it did work! #NBERday

Infrastructure quality improved about as much as various Sierra Leone's people and outside experts anticipated, and institutional change (stuff like electoral participationg) improved about as much as experts thought (tho less than Sierra Leonians hoped) #NBERday

So an institutionally-focused program succeeded in meeting expectations for the concrete and wires and pipes that people want development projects to achieve, *and also* achieved a modest institutional benefit as experts believed they would! #NBERday

This is very good news!

So why did this study find better results?

1) Institutional investments probably take a lot of time to yield returns. You need *decades* of follow-up.

#NBERday

) Measuring benefits can be hard, and economists need methodological advancements which are better at dealing with the unique quirks of "soft" data around institutions. #NBERday

In particular, while benefits to infrastructure seemed to fade somewhat over time between 2009 and 2016, benefits in terms of public trust, participation in government, etc, actually grew over time! #NBERday

Places with CDD projects even had slightly better performance on *Ebola* management and knowledge! In particular, CDD communities were more likely to have implemented formal Ebola management protocols. #NBERday

However, this program did *not* succeed in reducing crime. This is a problem because reducing crime and violence is a huge area where development projects want to make a difference as the payoffs are very large. #NBERday

Q: So how can we reduce crime in developing-country contexts?

A: Pay families $$$ to keep kids in school.
#NBERday
nber.org/papers/w29056

Basically, Colombia has like a "predictive analytics for poverty" program. They take a bunch of household traits for a person and score it. If it's low enough then they are likely a "good candidate" to get more transfers. #NBERday

But there's a sharp cutoff at the score of 11. Below that you get access to this pay-parents-for-kids-at-school program. ABove it you don't. For families below the cutoff, about 30% receive it. For those above, virtually none. #NBERday

So that's a regression disconintuity woooooo baby put your causality hats on! #NBERday

The key thing here is that while people juuust on one side or the other of the line are *very* similar, recipients of the cash are actually *poorer*, so arguably *more* at risk of adverse outcomes like crime, teen pregnancy, etc. #NBERday

Nonetheless, they find that receiving the cash is associated with much lower adverse outcomes across tons of different measures.

Huzzah! #NBERday

Okay, now we're into odds-and-ends of #NBERday

Q: Does Uber reduce car accidents?

A: Yes!
#NBERday
nber.org/papers/w29071

This paper uses proprietary data from Uber on the volume of rides to see what happens to traffic deaths when rideshare volumes increase. What they find is that more ridesharing is associated with lower deaths.

But.... hmmmm.... #NBERday

This does not seem correct. It seems like they're just comparing rideshare volume among places with ride sharing without any kind of exogenous shock. Uber rollout is exogenous, *volume of use* is not. #NBERday

In principle they could exploit something like variation in daylight hours to approximate night-life and thus rideshare demand, but they don't do that. #NBERday

So it seems like in general places and times with more ridesharing have lower traffic deaths but this isn't really causally proven. Sorry my intro was so optimistic! Mea culpa! I misunderstood the mechanism. #NBERday

Sharing this paper because I have no dog in the debate here, but I'm just laughing because this paper overturns a prior paper.... written by the same authors. #NBERday nber.org/papers/w29062

Here's the prior paper. It's not old! It's very recent! And they're now saying, nah, we were wrong, there wasn't a turning point in 1992. #NBERday nber.org/papers/w25697

Last #NBERday paper:

Trump's surprising victory in 2016 spared highly-Trumpist counties from the fertility decline that occurred in other counties. #NBERday nber.org/papers/w29058

The intuition to use counties that are both very-Republican AND ALSO which had a big swings toward Trump is to isolate places where support for Trump was especially enthusiastic and locally dominant. #NBERday

"Morale effects" have been observed before, but it hasn't always been clear what caused it. The assumption embedded in this specification is it's about a specific viewpoint being locally ascendant. #NBERday

The authors also show that Hispanic fertility fell slightly around and just after Trump campaign visits, and use Google search volumes as a robustness test for pregnancy timing, which are nice tests. #NBERday

i.e. we shouldn't see this as just "Trump supporters became more optimistic so had more babies" but also "People threatened by Trump became more pessimistic and so had fewer babies." #NBERday

Reason 9,786 why race-pronatalism is unworkable in a country that's already racially diverse! #NBERday