Fiachaire

I'm fairly certain it isn't. It's phrases are repeated, certain nouns or other parts of speech are used way more than I would imagine a neural network would be limited to.

Just a reminder: This article is an ad for a web crawler/content marketing service. The purpose of the article is to be viral. Data is selected and presented to that end. See especially "The Tragic Data Behind Selfie Fatalities" which charts data from a particularly poor Wikipedia list and what appears to be a google search.

If your clients come to you for help with student loans, aren't most of these figures blatantly obvious or hopelessly misleading? Think of it this way:

Only 1% of people over the age of 53 needed our help with student loans and also live with their parents.

Almost 25% of people living with their parents in need of our help with student loans have only a high school degree...for now. The company is only 3 years old.

Also, why is the Earnest article titled "Math, Engineering Majors First to Leave the Nest"/"Humanities, Psych Majors More Likely to Live at Home"? Math and engineering majors seemed middle of the road, and in the chart there is no indication that they even factor in graduation. If I have a high school degree, and a student loan and am a client of Earnest I may or may not live at home, but I definitely have a major.

And why do they have data for 48-50 (less than 10 != 0) states if they only operate in 41 and DC?

The 'report' is drawn from Earnests clients. Earnest bills itself as the bank of tomorrow, but seems to profit mainly by student and personal loans. As such, one of the requirements to get a loan/be used in this totally scientific advertisement is to be in the US...although they appear to only operate in 41 states, so perhaps there's a statistically significant exception.

Just a reminder that the Priceonomics blog is self-promotion designed to show clients they can get people online to viral-market the website/brand/service for them. The data and how it is presented may or may not be true, but it will definitely be about profitable content. As most of the consumers here seem to respond to the content by clicking through, upvoting, adding personal anecdotes, (perhaps sharing on FB or Twitter?) I'd guess this will be a successful advert.

Perhaps somebody trained in handling data could explain why 10 responses is enough to reflect the rest of a states population. Is it significant that out of 60,000 private data sets the company didn't have 10 from Vermont or South Dakota? If we're drawing conclusions based on geography, shouldn't we be able to determine which states have over 1000 responses, and which have less than 100?

My guess is the math works out. As they said $10 isn't a great amount of people, and if you're after less than 4% CTR then the overlap is probably negligible. While Priceonomics makes an effort to appear as though it respects your intelligence, and its adverts are cited by almost every major news source, it is ultimately a showboating supertrawler. Check out 'The Tragic Data on Selfie Deaths' which makes a long form argument about a viral topic while getting its data from a google search and a particularly bad wikipedia list.