Do jobs run in families? | Hacker News Reader

Do jobs run in families?

research.facebook.com

68 pointsironmantra30 comments

This data is pretty interesting. Of course, it's hard to separate out or even infer the causes as nature vs. nurture (which the article is careful about) but the part about twins was the most striking for me:

"Twins' tendency to choose the same occupation, at 24.7%, is even more striking."

A possible explanation was provided above:

"Whether identical or fraternal 2, twins are also more likely to be raised in a similar environment — parenting styles may differ as a parents add more children to their brood, but twins will likely be exposed to a similar parenting style."

Also, being raised in the same time period would also expose you to the same cultural/economic/societal factors - all of which can change with time, and also may affect your choice of career. (i.e. the proportion of math_CS parents could be different from the proportion of math_CS children just because of the different time periods in which each grew up)

> Of course, it's hard to separate out or even infer the causes as nature vs. nurture (which the article is careful about) but the part about twins was the most striking for me:

If you look, they already give you enough to get an idea of the relative contributions:

> 15% of siblings share an occupation, which is higher than the 8.6% rate for any two same-gender, same-age individuals in the population. Twins' tendency to choose the same occupation, at 24.7%, is even more striking.

So siblings have twice the rate of random people, and identical twins have twice the rate of siblings. Since twins have twice the genetics as well, this points to high heritability and (probably) lower shared environment.

There are twins separated at births studies that show _striking_ similarities between identical twins. Even when twins are brought up in different environments (income, ethnicity) their choices, intelligence, and divorce rates have a higher r^2 than non-twin siblings brought up within the same household.

Here's an interesting video that takes that into account.

Twins, even those that grow up separately, are remarkably similar.

https://www.youtube.com/watch?v=KDFh74eENuw

A little uncomfortable that a company that spends $10m lobbying the government in its interests, is also now doing social policy research using their near-monopoly on a particular sort of people's data. (This one, for instance, ties into social mobility which is always an argument that has currency with politicians.) Quite what they're arguing for behind the scenes is unknown ... but then that's part of the problem.

There's no "policy" to this research

They're using this data for incredibly accurate targeting, allowing advertisers to effectively target race/ethnicities: http://thenextweb.com/insider/2016/03/18/facebook-showed-us-...

I wrote a proof of concept on it last year too: http://caseysoftware.com/blog/social-apis-for-social-evil

I think father-daughter and mother-son is interesting too.

Did they run something like TF-IDF to normalize for popular occupations?

https://en.wikipedia.org/wiki/Tf%E2%80%93idf

This certainly makes sense: "To conclude, we see that people within a family are proportionally more likely to eventually also choose the same occupation." Still, I wonder if it has become less common with time: the idea of a parent (usually a father) passing the family trade on to the child isn't culturally common, as it used to be during the guild era that got started in the Middle Ages.

On an anecdotal level, I got started as a grant writing consultant because my parents started the business when I was a kid (I wrote a little more about that here: http://seliger.com/about/). But while I became interested i the business as a teenager, neither of my siblings did, and they both work in unrelated fields.

I also feel like I've met a disproportionate number of doctors whose parents were doctors. It does seem like they'd pick up a fair amount of useful information just listening to their parents talk.

I think it can be less subtle than that. For instance a piano solist will 'test' children inclinement to play music, or more directly to play piano. I know other engineers who also expose their kids to programming like games to see if there is potential.

It doesn't mean the trade will be passed down as is, but if a kid seems predisposed for a field, the parent with expertise will help it grow further as long as the kid enjoys it.

Related: Gregory Clark, The Son Also Rises

https://www.worldcat.org/title/son-also-rises-surnames-and-t...

(And yes, Clark's got a thing for very bad Hemmingway puns.)

Is it just me or is the data presentation completely terrible

I thought the data visualization was pretty good. How would you improve it?

A profession by profession table with the raw numbers would be more useful. Perhaps a Hinton Diagram would be a better visualization.

It's very hard to get a sense of how more or less likely a child is to pursue a given profession than normal. You just sort of have to eyeball the bars before and after, which tends to visually accentuate changes to common professions, even when there might be a much larger percentage change for an uncommon profession.

It depends on what you want to convey. For example, as a layman and someone who likes structured data, I want to find out "what are the top professions for the offspring of profession X"? To me, this screams like a sorted table of some kind with a selection.

The relative sized bars are cool in terms of animation but man, they are a pain to read.

This makes total sense -- you do what you know. As a child growing up, you emulate your parents in a lot of ways. If your parents are factory workers, they probably didn't push you towards college so you will end up in a similar job. Likewise, if your parents are doctors, you not only have a path to medical school laid out before you, you likely have connections that would give you a better than average chance of acceptance.

Good to see this proven out with data; but I'd consider this a pretty intuitive conclusion.

Interesting data. I think we'll get very interesting data from Facebook if they keep it up, considering they're one of companies who have the most data on people.

https://lornareiko.wordpress.com/2009/10/08/identical-twins-...

http://articles.philly.com/1990-11-04/news/25926164_1_identi...

I imagine the march of technological progress and America's shifts in employment mean it is a lot less prevalent than it was a generation or so ago - a lot fewer children following their parent into manufacturing jobs.

That said, my Dad worked in computing - not as a coder, just as a manager, but it meant that we had an early proto-PC, a 2,400 baud modem and a CompuServe connection. It was probably inevitable that I would end up doing something with computers...

The interactive visualization is very neat. Interesting data, and an interesting way of presenting the data.

Why are all the resultsets segregated by gender? The description of the problem they were looking to solve gave no indication that gender was of interest to them in analyzing the results.

That yellow when you hover over medicine/office contrasts very poorly w/ the white background.

that is very cute: fb doing research!

I'm not sure what you're saying here; Facebook does more research than some universities. Their propensity for experimenting on their users is well-documented, and while I disagree with their ethics, they do produce significant, relevant and valid results.

If you're prepared to toss ethics overboard it's not that hard to come up with significant, relevant and valid results. The whole trick is to get those results while acting ethical.

I think there's another issue that being willing to be completely unethical requires such distorted thinking (for humans, anyway) that it introduces its own sort of spurious results. Just thinking about a lot of callous and inhumane psychobabble that was promulgated not too long ago, mostly dealing with people whom the researchers thought of as not-quite-as-human.

Indeed, that's another risk. But for now FB seems to have steered clear from research in that vein. Even so I can't read about FB conducting 'studies' without thinking of this:

http://www.zdnet.com/article/facebook-unethical-untrustworth...

And like in that study (which went so far as to actively mess with the subjects emotional states) I miss the 'consent' element.

Where's the ethical concern here? Familial relationships can be explicitly provided as can occupation. If it's all self-reported data then what's the problem?

The problem with the emotion manipulation study was that they were effecting people's psychology, not just click behavior. Which is creepy, especially because it apparently worked.

I find it interesting that you don't see the problem. I suspect that is a large part of why companies like FB and the various advertising networks are utterly blind to what will eventually be their downfall.

Here is a nice example of a sample consent form the kind of which one would expect to fill out before having one's data be used for any kind of study:

https://www.irb.cornell.edu/forms/sample.htm

Not being informed about the fact that your data is going to be used in a study, that the results of that study are going to be published is un-ethical in several ways, for one it presumes consent where non exists, for another it is a breach of the law (because it violates the original purpose of collection). Not that Facebook would care about any of the above but companies and institutions that play by the rules would require the researcher conducting this study to obtain consent from the subjects.

Consent is given through the click-through user agreement and the data policy https://www.facebook.com/policy.php . It's no different, and anything more is impractical. There are literally hundreds if not thousands of tests being done simultaneously on overlapping user populations.

Reply on news.ycombinator.com