Something to remember - the median family in the '70's had one income. A median family today has two. So the margins (single income households) are in more dire straits today than they were in the past.

Measuring in households and not individuals hides a large number of changes in how people are paid. It also doesn't account for things like pensions, which were common in the '70's and earlier.

There are also more single person households now[0], and the average household size has decreased by about 15%.[1]

[0] https://www.statista.com/statistics/242189/disitribution-of-...

[1] https://www.statista.com/statistics/183657/average-size-of-a...

There are also more two-person households, so perhaps part of the change is young people not living with their parents for so long? Or perhaps it is old people living alone. I don’t think I have very good priors about this. I think the kind of people I know are not going to be a great cross section of US society (and I suspect this is true for basically anyone too).

It's actually the opposite, we are at an all time high for young people living with their parents, for longer.

Approximately 50 percent of young adults 18 - 29 live with their family now[0].

Initially that was thought to be a transient externality of the pandemic but it seems to be sticking[1] thanks to the outfall of the bull market crashing.

[0]https://www.pewresearch.org/fact-tank/2020/09/04/a-majority-...

[1]https://www.pewresearch.org/fact-tank/2022/07/20/young-adult...

Which entirely disqualifies the proposed explanation of the original comment, which considered household income. thus bundling the parents and children's income together.

The consumer price index (CPI) is untrustworthy. It incorporates adjustments into things like technology that don't actually reflect utility. For example, if you look at the section on "Computers, peripherals, and smart home assistants" you'll note a nearly 300-fold decrease in the consumer price of computers from 1997 to 2022. How the heck does that work? Computers today are cheaper than in 1997 but surely not 300 times cheaper.

It turns out that they use a "quality adjustment" that takes into account the relative performance of a modern computer compared to previous years. The bulk of the 300-fold decrease in price is due to the fact that modern computers are many times faster than those made in the 1990's. But there's a problem with that: a consumer doesn't get hundreds of times more utility out of a computer. You can't have a hundred people in your house sit there and work on the same computer and achieve the productivity scaling which would warrant this adjustment.

In reality, you're paying for one computer an amount which is nominally within an order of magnitude of what you would have paid in 1997. So this adjustment has the effect of drastically skewing the CPI statistic in this sector. Furthermore, computers have only grown in importance. What was a luxury in 1997 has become a necessity today. Families often have multiple computers. Parents may work from home and use their personal computer for paying bills, applying for jobs, doing taxes, budgeting, or applying for government services, while children may use their computers for doing schoolwork, research, playing games. Everyone uses their computers and devices for communications. However, none of these tasks (besides playing games) has markedly improved or been made more efficient by the technological advances in the computers themselves.

In practice, the way the poor live is to always look for a used, refurbished, or pre-owned option.

So while if you buy new you can't get very much computer, if you buy a bunch of junkers, you can get enough to get by. But they have to be in a sweet spot of "old enough" to have depressed price and "new enough" to not be useless for current applications. This was actually a problem during the 2010's long reign of Intel, because they didn't add a lot of generational performance improvement, so old systems retained their price for longer. Since other architectures have caught up on leading performance, pricing has finally started discounting these older CPUs again.

Still, it's not a 300-fold difference and the bottommost price brackets tend to need a DIY approach. But it's definitely more feasible than 1997.

Your numbers don't take into the age bands (you include older, wealthier families).

However, the comparing CPI to inflation adjusted incomes (as the article does) doesn't work either. Clearer explanation of the error I think the article made (and that I think you're pointing out):

For the sake of argument, say CPI is always proportional to income (which it would be if purchasing power was constant). To make arithmetic easy, say the CPI and median nominal income were both 1 in 1970, and are now both 8.

The inflation adjusted income in 1970 would be 8. (If we assume CPI moves with inflation, we multiply the 1970 median income by the ratio of 2022 CPI / 1970 CPI.)

So, by the author's calculations, a median family would have had 8 units of purchasing power in 1970, but only 1 unit of purchasing power in 2022, even though, by construction, CPI and nominal income went up in lockstep, and real purchasing power should be the same.

The original article was not based on incomes in any age ranges either. If they had wanted to do a more rigorous analysis, they would have compared say, 18-34 year olds in 1970 and today.

Anyway, the point is the first half of the article is fundamentally mistaken, There are ways to do more interesting analyses, but this one was mostly incoherent.

This one's a whopper. The annual salary in the first graph is adjusted for inflation, while the CPI is not adjusted for inflation. Rather the CPI is how inflation is calculated.

Wildly misleading.

I've lost quite a bit of respect for some of the HN community here - this should be obviously wrong, we obviously aren't buying 86% less today than we were in the past.

The title's premise is so outlandish it deserves at least a little bit of critical thought... But because it confirms some previous biases (even if at a completely different and more extreme scale), more than I expected are acting as if this is true.

Why should hackers have any idea about economics? Not everyone can be a subject matter expert on everything, and it's easy to be swayed on topics like this where emotions run high (rightly so) and anecdotal evidence is everywhere.

As for the actual number being outlandish, I don't think most people have a sense of what "86% less purchasing power" even means. People see "<surprisingly big number> + less purchasing power" and that gets the eyeballs. "Hackers" are just as susceptible to this as any other non-experts.

Plus, plenty of people here do know what they're talking about, enough to call out the obvious problems in the article and charts.

There have been methodology changes between 1970 and today.

Okay but two wrongs don't make a right. If there were actually methodology changes that caused CPI to be not representative of purchasing power then accounting for CPI twice doesn't really fix anything, it just adds more noise into the result. To do it properly you need to explain what the methodology changes were, estimate their impact on the "real" CPI, and use your revised CPI to account for inflation. The author hasn't done that, so it's far more likely she has no idea what she's doing.

I don’t know if you’re thinking of it but there’s this site called shadowstats which claims to calculate CPI/inflation with the old method and shows totally wild inflation numbers but the way it actually works is that the guy who runs the site takes regular CPI and adds in adjustments to his liking to get inflation numbers that he thinks look right.

My understanding is that you get pretty similar numbers with either methodology.

See more here for example: https://fullstackeconomics.com/no-the-real-inflation-rate-is...

That doesn't make any difference; any methodological differences are just swamped by the erroneous double application of the price deflator in the comparison.