iiinova — where your digital self keeps evolving

John Voorheis @johnvoorheis.bsky.social · 8h

The TL;DR version is: take generational comparisons of wealth/homeownership with a large grain of salt if they are based on household head demographics (as in HVS and SCF based statistics) -- apparent trends may (likely) be due to changes in headship, not actual changes in the distribution of wealth

071

John Voorheis @johnvoorheis.bsky.social · 8h

A simple illustration from a simulation show this: if you take 20 successive cohorts who have wealth drawn from the same distribution, and let headship rates decline over time faster for lower wealth people, you get exactly this pattern

121

John Voorheis @johnvoorheis.bsky.social · 8h

In this case, the median/mean of the household wealth distribution you can calculate using the household heads will deviate from the true distribution of personal wealth in a positively biased way -- its essentially a selection bias issue, as the heads you see in the data are higher wealth

131

John Voorheis @johnvoorheis.bsky.social · 8h

But if headship rates decline *and* this decline is related to the underlying distribution of wealth you are measuring, then things can go haywire. This can happen if people of a given age group are living with different age group heads (e.g. parents) in a way that's correlated with personal wealth

141

John Voorheis @johnvoorheis.bsky.social · 8h

If people of a given age group were equally likely to be household heads over time, you don't have a problem -- the aggregation is working the same across cohorts, so you can do apples-to-apples comparisons. and if headship rates are around 50%, per capita HH income ~ personal income.

131

John Voorheis @johnvoorheis.bsky.social · 8h

For some survey-based measures, this is all the survey collects and that's a sensible enough way to aggregate things to capture cross-sectional distributions of income or wealth. The issue can come when you want to compare these age-specific distributions over time.

121

John Voorheis @johnvoorheis.bsky.social · 8h

Apropos my earlier householder point, here's a quick summary of the issue. The issue arises when you use households (or families, or primary economic units but I'll just say HH) as a unit of analysis and assign the head of the household's demographics to the household, as in "wealth at age 25-34"

131

John Voorheis @johnvoorheis.bsky.social · 10/10/2026

Should note I'm being cheeky and not actually trying to impugn @ernietedeschi.bsky.social 's character -- this householder (or PEU head per Ernie) issue is something that comes up constantly, people get wrong (imo) and it grinds my gears

2101

John Voorheis @johnvoorheis.bsky.social · 10/10/2026

And since non-PEU head wealth is invisible to the survey the medians are not really comparable over time. I'm being dramatic, and of course the sucky thing is there's no better alternative; only approach that works is to acknowledge the stat isn't capturing what we want it to

181

John Voorheis @johnvoorheis.bsky.social · 10/10/2026

I'm too lazy to make a "they're the same picture" meme, so impute one. if you replace PEU head for householder the critique goes through and you end up in the same place. The distribution of household/PEU heads is different from the distribution of individuals in a way that varies across generations

191

John Voorheis @johnvoorheis.bsky.social · 10/10/2026

The bummer thing is that there's not really an alternative that can measure the thing people want

240

John Voorheis @johnvoorheis.bsky.social · 09/10/2026

I don't ever know if this is old wives tale or not but all the gardeners I talk to insist that black walnuts don't really compost and will kill gardens because they have this toxic chemical, which is not great given how many get buried all over my yard by overeager squirrels

110

John Voorheis @johnvoorheis.bsky.social · 09/10/2026

Bad chart, everyone who posts it loses chart guy points

3211

John Voorheis @johnvoorheis.bsky.social · 09/10/2026

This chart is telling you about median wealth *for household heads* across generations -- that's a mix of actual distributional changes and compositional changes in headship. It is *not* the median household wealth of each generation!

2242

John Voorheis @johnvoorheis.bsky.social · 09/10/2026

This is a bad chart for one specific reason: the SCF is a household based survey, so it only can tell about the demographics of household heads. Householder status has varied across generations, and has become more positively selected. Same issue as the official homeownership series.

23912

John Voorheis @johnvoorheis.bsky.social · 09/10/2026

Young adult householders are more positively selected than in previous generations, so the SCF is just not the right way to measure this. Bad chart, you all should know better!

020

John Voorheis @johnvoorheis.bsky.social · 07/10/2026

Hard to tell the difference between being confidently wrong and refusing to budge re: signaling and purported belief in signaling being an ex post justification for racism is the thing

040

John Voorheis @johnvoorheis.bsky.social · 04/10/2026

(a similar issue is emerging with educational polarization, which also intersects with the SF-specific problems)

040

John Voorheis @johnvoorheis.bsky.social · 04/10/2026

Also, this whole deal is, like many things downstream of housing policy. A maximally NIMBY metro area dominated by a single high value added industry will inevitably displace everyone who could tell the remaining computer boys they've gone insane and are being morons

280

John Voorheis @johnvoorheis.bsky.social · 04/10/2026

And Bullshit Jobs (and/or the ambient ideology that is downstream of it) justifies this incuriosity.

130

John Voorheis @johnvoorheis.bsky.social · 04/10/2026

I do think that "Graeber is the 21st century's biggest monster" is a bit overdrawn (because there is demand for something like "Bullshit Jobs" that would have been supplied anyway), but it's not totally wrong. Implicit or explicit in all AI scenarios is a confident incuriosity in what jobs even are.

190

John Voorheis @johnvoorheis.bsky.social · 01/10/2026

I think this is in line with at least the LinkedIn part of the data -- people report job losses on LinkedIn because they are looking for work, but then when they get a job maybe they don't update much anymore

110

John Voorheis @johnvoorheis.bsky.social · 01/10/2026

An ideal speculation: because Revelio and other similar commercial data are sources at least in part from visible parts of the labor market (what people on LinkedIn report), it is both biased and also picking up a vibes relevant (unrepresentative) subsample

120

John Voorheis @johnvoorheis.bsky.social · 01/10/2026

I do wonder if "you can do whatever remodelling you want to any federal building in exchange for purging the current cabinet and stopping everything you did the first two years" might actually work as a bargain in divided govt.

040

John Voorheis @johnvoorheis.bsky.social · 30/09/2026

I think the distinction is not solely about the illegibility of jobs, but how that illegibility is approached. The bossism-Bullshit jobs axis about being confidently wrong about the illegible jobs and not having the curiosity to be less wrong vs. being humble about understanding a hard thing.

150

John Voorheis @johnvoorheis.bsky.social · 30/09/2026

Knowing ball is not a time-invariant characteristic is the thing But it does make it even more pathetic when someone who should know ball and has in the past known ball very much does not know ball currently.

030

John Voorheis @johnvoorheis.bsky.social · 30/09/2026

There were dozens of men's college basketball games that day though -- including some quite good matchups (e.g. Tennessee vs Kentucky, who ended up #2 and #3 in the final AP poll). It's just a comprehensive failure at knowing ball all around

230

John Voorheis @johnvoorheis.bsky.social · 30/09/2026

It's just multiple levels of not knowing ball, kind of a masterpiece tbh

1110

John Voorheis @johnvoorheis.bsky.social · 29/09/2026

Jokes aside, the "going into the trades pays off" discourse, which represents a common view, is I think an example of people largely getting magnitudes wrong (about distributions and lifecycle earnings)!

2131

John Voorheis @johnvoorheis.bsky.social · 29/09/2026

I think people probably can get an ordinal ranking of occupations/college majors half decently, but people 100% get the magnitudes wrong (and it's the magnitudes that matter)

170

John Voorheis @johnvoorheis.bsky.social · 29/09/2026

Real Personal income per capita is ~ 60% higher than median real personal income. We absolutely could raise the standards of the bottom half of the income distribution to the median with redistributive room to spare

030

John Voorheis @johnvoorheis.bsky.social · 29/09/2026

In this country we only believe in two genders, treated and control

151

John Voorheis @johnvoorheis.bsky.social · 24/09/2026

Connecticut actually hasn't legally had counties since 1960, the Census made up county equivalents for statistical purposes in 1970 without really asking anyone and everyone went with it. CT government finally got around to suggesting a different definition in 2020

010

John Voorheis @johnvoorheis.bsky.social · 23/09/2026

Swing state vibe check: the Republican ads during the college football and jeopardy slots (the only times I watch live TV) have largely dropped the transphobia in the last two weeks

010

John Voorheis @johnvoorheis.bsky.social · 22/09/2026

I think they pretty much just rake to the cvap totals by e.g. race, age, educ etc. adjusted for their turnout assumptions

230

John Voorheis @johnvoorheis.bsky.social · 18/09/2026

Saw a 5 handle for non-discounted regular gas in SE Michigan

030

John Voorheis @johnvoorheis.bsky.social · 13/09/2026

What if another name for "jagged frontier" is "not a general purpose technology"?

083

John Voorheis @johnvoorheis.bsky.social · 04/09/2026

You lose oregon if they don't also make a vegan version though is my understanding

150

John Voorheis @johnvoorheis.bsky.social · 30/08/2026

It's also structurally the same insult as "Democrat Party", something that I guess every Republican elected is required to say now. So they have all the tools to understand the bit! They just don't know ball.

120

John Voorheis @johnvoorheis.bsky.social · 26/08/2026

(it only works as long as you bring in more workers, either via fertility or immigration)

020

John Voorheis @johnvoorheis.bsky.social · 26/08/2026

Every society that doesn't let its old folks die in ditches is a ponzi scheme that redistributes from workers to those no longer able to work

291

Reposted by John Voorheis

jamelle @jamellebouie.net · 23/08/2026

but like, it's not my fault that you have chosen to embrace one of the most dogshit ideas the human mind has ever conjured into being!

263186192

John Voorheis @johnvoorheis.bsky.social · 13/08/2026

In expectation, papers that referees ding because "we already know this" are more socially valuable than the typical cool top 5 paper

1303

John Voorheis @johnvoorheis.bsky.social · 13/08/2026

A meta problem with academic journals (e.g. referees and editors) -- obsession with novelty means that unique, first of a kind papers are prized, and drive narratives/policy/future science, but also, because they are unique, may be wrong in a way that a bunch of similar papers are unlikely to be

4578

John Voorheis @johnvoorheis.bsky.social · 09/08/2026

The answer is actually that they apply weights to adjust for the fact that you didn't respond (so other sampled basement gooners who were taking a break and answered the door get upweighted)

0100

John Voorheis @johnvoorheis.bsky.social · 09/08/2026

essentially all of social media is just people who should know better backfilling non-idiotic reasons for why idiots say and do idiot things

210

John Voorheis @johnvoorheis.bsky.social · 08/08/2026

As someone who has written papers grappling with how we can improve CPS weighting, sampling, imputation, etc I'm all ears if you have anything concrete that could be implemented to improve the survey

130