[onlyrealcuzzo]

So... Let me get this straight.

Because people who had iPhones during the AT&T exclusive period has less kids...

They think there is no other possibly explanation besides the iPhone, because they looked at similar groups on different networks and in different areas that didn't yet have coverage for iPhones?

It definitely couldn't have been due to richer people having iPhones and having less kids, or people preferring iPhones who weren't going to have kids anyway??

Why definitely not? And why definitely iPhones or Smart Phones or whatever?

[FabHK]

I suggest you read the abstract, at least. The fact that only AT&T had the iPhone back then resulted in a natural experiment: It was only available in certain regions. You can thus compare regions where it was available and where it wasn't, while controlling for "richer people" or "people preferring iPhones".

As a rule of thumb, if you look at something for 3 minutes and have some obvious questions, the scientists that looked at it for several years of their life in great detail might have had those same obvious questions as well.

[hansonkd]

Not to be pedantic and I agree with what you are saying but:

> As a rule of thumb, if you look at something for 3 minutes and have some obvious questions, the scientists that looked at it for several years of their life in great detail might have had those same obvious questions as well

This does not mean that just because they had those obvious questions that they were properly resolved. Human history has a long track record of people who knew better but chose to ignore. In science there is an incredible pressure to have positive results rather than negative ones (IE nobody would care or know about this study if the title was "we looked and iphone doesn't explain 33-52% of fertility decline"

[makeitdouble]

> natural experiment: It was only available in certain regions.

This study treats ATT doing market research and progressive rollout through prioritized markets as a "natural experiment".

We could at least agree it's specifically chosen population, whatever ATT marketing dept had in mind when they planned the rollout.

[tylervigen]

Right but the “certain regions” were affluent metro areas, the control group was rural counties, and the timeframe was the Great Recession.

Even the study authors acknowledge the severe risk of the huge confounding variable.

[virgilp]

That's for the good studies. Let's not pretend that all published studies are honest. Unfortunately it is quite reasonable to be skeptical about extraordinary claims such as this one.

[rayiner]

It's not reasonable when the skepticism is disproven by looking at page 11 of the paper that's in the link.

[FabHK]

It is reasonable to be skeptical, absolutely. But responding to a study like this with "haha what about if only rich people got iPhones" or "bro don't you know that correlation does not imply causation" is juvenile.

[virgilp]

To be fair, their control variables treat the first objection (wealth), not the second (brand preference; and yeah there's some correlation but one doesn't imply the other)

[ilija139]

Not every rich person got an iPhone. The rich people without an iPhone did not had equal amount of less kids. There are two groups, one has an iPhone, the other has not. The assumption is that two groups are big enough to have equal amount of people from any other group that can explain the decline in fertility, i.e. equal amount of rich/poor, educated, etc. They can control for this because they know which people had access to the iphone based on the AT&T network coverage.

At the end of the abstract they state the likely explanation of this seemingly spurious correlation: > National-survey evidence on time use and sexual behavior is consistent with the iPhone reducing in-person interactions, increasing pornography use, and reducing sexual frequency.

[pbhjpbhj]

>the iPhone reducing in-person interactions, increasing pornography use, and reducing sexual frequency.

Why would iPhone _particularly_ do that? I can see greater social media use, greater access to porn, would do those things. But that's common to smartphones in general.

[ilija139]

The iPhone was the only smartphone in 2007, and the one that they could track based on AT&T coverage.

[mike_hearn]

It wasn't the only smartphone. And it's not like the internet appeared simultaneous with the iPhone. It was just a different way to get access to the same websites, back when it launched, most of which didn't support the iPhone's screen size anyway. And the only difference was that it made it easier to reach those websites ... outside. On the move. Where you are more likely to meet people.

Their causal explanation here does not ring true in all sorts of ways. Nobody was deciding not to get married and have kids because they were too busy watching porn on an iPhone in 2007, a time when video on smartphones barely worked. Social media predated the iPhone by years. And this practice of controlling for some ad-hoc list of things and then announcing causality is a major cause of spurious false positive claims in academic science. They really should know better than to make strong claims of causality in a case like this. Sociology just isn't capable of explaining things like this no matter how many stats you throw at it.

https://pmc.ncbi.nlm.nih.gov/articles/PMC4816570/

There's much better work out there on this topic that is less heavy on the cute regressions and more heavy on the historical analysis+common sense. It's often a lot more convincing.

[raincole]

> Let me get this straight.

Let me get this straight, I believe one needs to read a paper to get it straight.

But I fully understand your knee-jerk reaction. That was my reaction when I read the title too. However, it seems to be a surprisingly well-thought analysis where all your points are answered (controlled).

If I read it more thoroughly I'll likely find flaws on the statistical methods. But it's not like the authors didn't have common sense.

Edit: unfortunately, enough people had the same knee-jerk reaction to get this thread flagged. We really need a way to vouch a thread.

[munificent]

Read section 5.

"Table 1 documents that treated counties (those with >90% AT&T 3G coverage) are substantially more urban, White, Republican-leaning, and affluent than control counties. To address this imbalance, we apply the entropy-balancing reweighting of Hainmueller (2012), which solves for the entropy-minimizing set of control-county weights that equalize the treated and reweighted-control means of a specified set of covariates."

[jsw97]

Nobody at this level writes a paper like this, asserting a specific causal relationship, without considering exactly the questions you raised. The authors address your concerns. It's possible they did so poorly. But that is the case you would want to make. I'm tired of reading these low-effort takes on HN.

[jfengel]

HN sees an awful lot of low-effort "research", which merit only low-effort takes.

This comes from the National Bureau of Economic Research, a well-respected organization. Even if you read only the title, the presence of "nber" in the URL means that it's worth more than just a hot take.

If the link were to substack, or phys.org, or various other content mills, then the HN kneejerk reaction would be warranted. Or rather, what's warranted would be silence, but at least the kneejerk reaction wouldn't necessarily be disproportionate to the quality of the work.

[rayiner]

The paper has a section listing their controls, including income, poverty rate, race, etc: https://www.nber.org/system/files/working_papers/w35310/w353.... It's on page 11.

[spwa4]

Because correlation is not causation. If A and B correlate there's 3 options:

1) A causes B

2) B causes A

3) C causes both B and A (in some order)

4) your correlation figure is bullshit (hence not counted in the 3 options, but certainly with news these days, it must be mentioned)

A famous way to illustrate where this goes wrong is to show a map which libraries that loaned out Harry Potter books, and a map of where poodles got raped. Very high correlation, and obviously an example of the 3rd option.

(obviously both were caused by population density, which leads to both library creation and poodle-related crimes. And probably non-poodle-related crimes)

[wavemode]

The 5th option is random chance.

That often results from p-hacking. In a world of infinite variables, if you look hard enough you are guaranteed to eventually find two completely unrelated variables that correlate with each other over a statistically significant period of time.

[virgilp]

That's the 4th option

[wavemode]

I guess it could be? I interpreted what the parent commenter wrote like "the variables aren't actually correlated" (which definitely does happen sometimes)

Whereas my point is moreso when, the variables really are correlated but it's purely due to random chance. Not bullshit, per se, just bad luck (or possibly, p-hacking).

(Though the solution to both is the same - you shouldn't trust a study until it's been independently replicated on new data.)