Share this page!

Enter Twitter Thread URL to Unroll

Needs to be the full URL like: https://twitter.com/threadreaderapp/status/1644127596119195649

How to get URL link on Twitter App

  1. On the Twitter thread, click on or icon on the bottom
  2. Click again on or Share Via icon
  3. Click on Copy Link to Tweet
  4. Paste it above and click "Unroll Thread"!
  5. More info at Twitter Help
Crémieux Profile picture
@cremieuxrecueil
May 26 • 6 tweets • 2 min read Twitter logo Read on Twitter
Scrolly
"[Scholars] argued... poor structural factors—poverty, ethnic heterogeneity, population mobility—cause crime.... Our first lesson is that the relationship between structural factors and crime may be spurious."

Yes. Neighborhood deprivation does not appear to cause violent crime. Image
Nor does poverty. Image
Or teen motherhood. Image
Or a person's mother having them when she was young. Image
And of course, there are replications of the lack of effect of things like family income, parental unemployment, parental welfare usage Image
or a parent doing their own criminal offending. Image

• • •

Missing some Tweet in this thread? You can try to force a refresh
　

Keep Current with Crémieux

Crémieux Profile picture

Stay in touch and get notified when new unrolls are available from this author!

Read all threads

This Thread may be Removed Anytime!

PDF

Twitter may remove this content at anytime! Save it as PDF for later use!

Try unrolling a thread yourself!

how to unroll video
  1. Follow @ThreadReaderApp to mention us!

  2. From a Twitter thread mention us with a keyword "unroll"
@threadreaderapp unroll

Practice here first or read more on our help page!

More from @cremieuxrecueil

Crémieux Profile picture
May 27
This pattern denotes publication bias, but it can also demonstrate problems with intervention scaling

The absence of this pattern in scenarios with strict preregs speaks to it not being driven by failing to scale

Like this: preregistered money priming effects are precisely zero Image
But sometimes you get something like the publication bias pattern with data where

- Preregistration is required
- Reporting is required
- Analyses are evaluator-independent
- Samples are realistic

Ed studies that received grants from the Education Endowment Foundation and the
National Center for Educational Evaluation have those requirements. For realistic sampling, the NCEE imposes the requirement and the EEF only uses it for categorization.

The results for these agencies do seem to show larger effects -> larger SEs: the classic pattern Image
Read 8 tweets
Crémieux Profile picture
May 26
American researchers are really only getting huge Census data recently, whereas Scandinavians and Dutchmen have been using theirs for a while.

Americans can learn a lot from the excellent reporting practices that seem to be common with Scandinavian register study researchers.
Causal inference can be done in the same ways on the data from these places, but if you perform the same analyses and you don't report your results clearly enough to interpret them or use the provided statistics so the results can be qualified, your results are almost useless.
With Chetty, I know he can do good reporting. In fact, I've seen it! I also know that in his most popular papers, his reporting is undergraduate-quality and he knows better. I don't know why this happens; he is so clearly brilliant and when he wants to, he can do some of the best
Read 8 tweets
Crémieux Profile picture
May 26
A new NBER preprint by @brittanyrstreet and @Econ_Mike includes one of the first U.S. sibling control estimates of the effects of neighborhoods on kids'

- crime
- employment
- teen parenthood
- death

But we don't know the descriptives, so it's unclear how to interpret coefs! Image
For criminal charges by age 26, the coefs are all null within sibling pairs, but they're also similar in size or larger than the other results, which do not appear to just be between-pair ones

Suggestive: maybe no effect, or at least not one that's large enough to show up in 99k
pairs? individuals? It's not clear what the observation number refers to, but I would really like to find out so I can at least figure out the power with different possible variances.

For W2 employment, there's a possible effect with young exposures. It's not clear how large
Read 9 tweets
Crémieux Profile picture
May 26
Emily Oster had this insight back in 2020.

She exploited the fact that a NEJM piece published in 1993 showed vitamin E had major health benefits and another article published in 2004 showed that vitamin E's benefits were overstated and it might actually be harmful. twitter.com/i/web/status/1… Image
Not everyone jumped on board the vitamin E train.

In fact, it was educated, high-income people who exercise, don't smoke, and have good diets who jumped aboard. Image
The effect on the relationship between vitamin E consumption and heart health was that the newly-introduced, highly selective uptick in vitamin E consumption created a correlation between vitamin E consumption and heart health and bolstered one with mortality risk. Image
Read 8 tweets
Crémieux Profile picture
May 26
Stu is fighting the good fight.

The argument that breastfeeding has major benefits is based on little evidence and the continued insistence it matters so much does little more than shame women who struggle to breastfeed.

Let's look at the causal evidence for IQ.

Here's some: twitter.com/i/web/status/1… Image
But you may object, we need replications. I agree. There was a replication of Der, Batty & Deary's result (bmj.com/content/333/75…).

The replication by Evenhouse & Reilly was an almost-explicit fishing expedition where they just tested the sibling control effect on everything.
They ended up finding a p = 0.04 effect on Peabody score using the variable "Months Breastfed" and a p = 0.07 for "Ever Breastfed".

There was no correction for the multiple comparisons they sought to engage in, so obviously this is a nothingburger (ncbi.nlm.nih.gov/pmc/articles/P…). Image
Read 9 tweets
Crémieux Profile picture
May 26
A new preprint added to the case for herpes causing dementia.

How did they do it? They used a Herpes zoster birth date vaccine eligibility cutoff for a regression discontinuity.

Dementia diagnoses were down 3.5% or 19.9% in relative terms for people born a week later. twitter.com/i/web/status/1… Image
The results are hopeful, but very weak. They actually showed some troublingly unexplained heterogeneity by sex Image
and with a sample as large as the one they had, their result is suspicious (p = 0.019) - it might end up being a prime example of publication bias.

We're also not sure how broad the vaccine effects are. They shouldn't be broad, but studies on this intervention
Read 4 tweets

Did Thread Reader help you today?

Support us! We are indie developers!

This site is made by just two indie developers on a laptop doing marketing, support and development! Read more about the story.

Become a Premium Member ($3/month or $30/year) and get exclusive features!

Become Premium

Don't want to be a Premium member but still want to support us?

Make a small donation by buying us coffee ($5) or help with server cost ($10)

Donate via Paypal

Or Donate anonymously using crypto!

Ethereum

0xfe58350B80634f60Fa6Dc149a72b4DFbc17D341E copy

Bitcoin

3ATGMxNzCUFzxpMCHL5sWSt4DVtS8UqXpi copy

Thank you for your support!

Follow Us on Twitter!

Send Email!

:(