Glad you linked to Lyman Stone's piece. I am not sure that "stay at home" is the best alternative term to "wage earner." I know this demographic and they are not "at home" but in sitting on community, school, and foundation boards; in churches; raising money; leading cultural organizations, political organizations; managing staff, shopping, influencing -- all in all, super active, as well as raising the children.
Hmmm, more likely the way that sufficient affluent wives have always been -- delegating the housekeeping, cooking, cleaning, and child care to servants and using their time on the various external engagement activities you mention.
Just bonkers to me that while the leading AI companies are forecasting “complete automation” in the next decade, the data doesn’t support ANY automation at all yet for the occupations most exposed to it
I'm glad you highlighted the Sanders chart. My son showed it to me and we've nominated it for the most misleading chart of the year, perhaps the decade. I wish there was something equivalent to fact check ––a chart check!
#1 - I'm getting a little tired of reading lots of stories on AI and how it will revolutionize the things people do to earn a living. I would like to see a post by Noah about ALL the jobs that AI will have a minimal impact on. A recent experience of a close friend who suffered a major ankle injury brought this home to me. AI will not replace orthopedic surgeons, nurses, physical & occupational therapists. Most rehab facilities are woefully understaffed particularly at night and on weekends. Many such as the one where the friend is are owned by private equity who want to squeeze out profit at the expense of patients.
AI will not replace construction workers, plumbers, electricians, yard care workers, many different agricultural workers whose tasks cannot be automated.
No doubt Bernie Sanders's chart is silly. But I've often heard that the ratio of house price to income ratio roughly doubled since 1980. Claude suggests an increase from 2.5 in 1980 to 4.4 in 2023: an increase of 76%. Not quite a doubling, but considerable. Can this be reconciled with your figures for the time it takes to pay for a home increasing by 25%?
Well, the Fred graph in the column says the ratio has increased from 8 to 10. That suggests that what Fred and Claude are looking at is decidedly different. But as BronxZooCobra notes, if the mortgage rates go down, the same *payment* supports a larger *principal* and thus a larger house price.
In any case, the argument is pointless: If the average person can't afford the average house, its price will come down because the current owner will have to cut the price they are demanding -- prices for houses are determined directly by what people can pay for them.
If the ratio of house price to individual income nearly doubled, but the number of individual incomes per household went up by 30%, then the ratio of house price to household income only goes up by about 30% as well.
I don’t know if this is precisely what the numbers you are describing are measuring, but this is the sort of thing where it really matters whether you are looking at individual incomes or household income, and median or mean!
I would like to see these "weekly earnings vs. home prices" charts disaggregated by region. America is a big country. We know there are lots of cheap houses in deep rural areas, but the structural problem we face is a lack of well-paying jobs where the houses are cheap. It's *possible* that these charts smooth everything out by blending nationwide wages and separately blending nationwide housing prices. But I'm concerned that this specific blending process is equally likely to be masking the real story.
Maybe the people with affordability concerns are just overreacting, but can we possibly find some metric that looks at these regions more specifically?
I think the "trad wife" characterization is a little off-point, given the popular associations with
Trad Wife. One reason highly educated, high-earning women are very useful participants in a power couple is that they bring tremendous working and networking skills to the couple, which can now be put into service for the couple's agenda and can free the male partner to pursue a single-minded focus in his working life. These are not, typically, "cookie moms," and they themselves may be supported by domestic help.
After underwhelming Chat GPT5 and growing environmental opposition to power and water requirements, what if the AI future is never getting here outside of some coding help and image and video generation fun?
The case for AI Agents replacing jobs seems further away with every new AI model release. As long as the hallucination rate remains at 5% then any complex workflow will never be able to be replaced by this technology due to compounding errors. Suppose you have a 10 step process, then the success rate is a mere 60%, which is unacceptable to any company.
People in the jobs with AI risk will probably end up using it to become more efficient rather than being replaced by it, unless this can be solved. My company has been trying to leverage it and I've noticed that the leadership has gone from full hype 12 months ago to noting some of these problems now as they get into the implementation weeds.
60% is unacceptable for some tasks but great for others! If you’ve got something that takes a lot of effort to produce but very little effort to check, then having an AI produce 5 of it and a human check them to find the three working ones is probably a lot easier than having the human generate one working one. (I trust you don’t need any examples of the kind of task where 60% success rate is no good.)
Even at its best, a medium which relies on the continuous partial attention of its audience is no place for a chart with two lines going upwards at slightly different gradients. And now that the social media 'zone' is full of Steve Bannon's 'shit', a politician would be daft to try to campaign so prosaically. Neil Postman was right all along. But we are where we are.
Regarding AI automation and the lack of impact on jobs, here are a few hypotheses.
Adoption among companies is still pretty uneven, not that many companies have made it to production yet with AI workflows, and for the ones that have, there is still a human in the loop. So let's say you're having AI check the documentation for a car purchase agreement. Someone is still looking at all the work to confirm the AI didn't make a mistake. All the other tasks for that job still exist, too. And to the extent your process is now more productive, maybe you're trying to complete it faster or better and take market share from competitors rather than just banking cost savings. Once your competitors all adopt AI, there won't be as much benefit from being faster, and then maybe cost reduction becomes more of a focus. When the AI version of the workflow becomes sufficiently reliable, or you can introduce another AI to verify the work, that may be another inflection point where the effect on employment becomes noticeable.
AI adoption is also generating a lot of work in terms of integrating the technology, reengineering the business processes in question, and building a data foundation that AI can use to get accurate information. So early adopters of AI may actually be hiring more people in many cases.
Overall, we are definitely in the phase where you have more bank tellers even though there are now ATMs, but I still think at some point a lot of work gets replaced wholesale. I could be wrong, and on top of that, the timing is wildly unpredictable.
I would also quibble with the usage of “overwhelming” in Lyman Stone’s charts. Especially since they imply that despite a significantly higher fraction of women staying home when their financial situation is more conducive to that, most still work!
I suspect another effect is that the amount of inherited wealth in the middle classes is increasing, so the size of small inheritances the middle classes can use for down payments is going up.
<Erdman blames this on overcautious prudential regulators who overlearned the lessons of under cautious regulations before 2008.>
It's easy to call the prudential regulators "overcautious" when times are good. Next thing you know, a housing bubble bursts and everybody and their mother starts asking "why didn't you do anything to stop it?" Bureaucrats are gonna do what they've always done: interpret the law in the most cautious way possible to avoid litigation and scrutiny.
On the spousal income chart, it might be helpful to chart median spousal income percentile on the y axis rather than average spousal income. It would also be useful to see if this is average spousal income of all spouses (including non-working ones) or if that chart only shows working spouses, and if it shows mean spousal income before marriage or mean spousal income at present. There are several different stories these numbers might be telling with slightly different versions of this!
Glad you linked to Lyman Stone's piece. I am not sure that "stay at home" is the best alternative term to "wage earner." I know this demographic and they are not "at home" but in sitting on community, school, and foundation boards; in churches; raising money; leading cultural organizations, political organizations; managing staff, shopping, influencing -- all in all, super active, as well as raising the children.
Needs a cute acronym, like we have for NEETs, but meaning the opposite. I propose MESSI: Married Embracing Service, Social Impact.
Hmmm, more likely the way that sufficient affluent wives have always been -- delegating the housekeeping, cooking, cleaning, and child care to servants and using their time on the various external engagement activities you mention.
Just bonkers to me that while the leading AI companies are forecasting “complete automation” in the next decade, the data doesn’t support ANY automation at all yet for the occupations most exposed to it
I'm glad you highlighted the Sanders chart. My son showed it to me and we've nominated it for the most misleading chart of the year, perhaps the decade. I wish there was something equivalent to fact check ––a chart check!
#1 - I'm getting a little tired of reading lots of stories on AI and how it will revolutionize the things people do to earn a living. I would like to see a post by Noah about ALL the jobs that AI will have a minimal impact on. A recent experience of a close friend who suffered a major ankle injury brought this home to me. AI will not replace orthopedic surgeons, nurses, physical & occupational therapists. Most rehab facilities are woefully understaffed particularly at night and on weekends. Many such as the one where the friend is are owned by private equity who want to squeeze out profit at the expense of patients.
AI will not replace construction workers, plumbers, electricians, yard care workers, many different agricultural workers whose tasks cannot be automated.
Why isn't someone writing about this?
No doubt Bernie Sanders's chart is silly. But I've often heard that the ratio of house price to income ratio roughly doubled since 1980. Claude suggests an increase from 2.5 in 1980 to 4.4 in 2023: an increase of 76%. Not quite a doubling, but considerable. Can this be reconciled with your figures for the time it takes to pay for a home increasing by 25%?
Of course - interest rates for 30 year fixed rate mortgage in 1980 were 13.74%. That nearly doubles the payment compared to today's 6.6%.
You did the math :).
Thanks. That explains it.
Well, the Fred graph in the column says the ratio has increased from 8 to 10. That suggests that what Fred and Claude are looking at is decidedly different. But as BronxZooCobra notes, if the mortgage rates go down, the same *payment* supports a larger *principal* and thus a larger house price.
In any case, the argument is pointless: If the average person can't afford the average house, its price will come down because the current owner will have to cut the price they are demanding -- prices for houses are determined directly by what people can pay for them.
If the ratio of house price to individual income nearly doubled, but the number of individual incomes per household went up by 30%, then the ratio of house price to household income only goes up by about 30% as well.
I don’t know if this is precisely what the numbers you are describing are measuring, but this is the sort of thing where it really matters whether you are looking at individual incomes or household income, and median or mean!
I would like to see these "weekly earnings vs. home prices" charts disaggregated by region. America is a big country. We know there are lots of cheap houses in deep rural areas, but the structural problem we face is a lack of well-paying jobs where the houses are cheap. It's *possible* that these charts smooth everything out by blending nationwide wages and separately blending nationwide housing prices. But I'm concerned that this specific blending process is equally likely to be masking the real story.
Maybe the people with affordability concerns are just overreacting, but can we possibly find some metric that looks at these regions more specifically?
I think the "trad wife" characterization is a little off-point, given the popular associations with
Trad Wife. One reason highly educated, high-earning women are very useful participants in a power couple is that they bring tremendous working and networking skills to the couple, which can now be put into service for the couple's agenda and can free the male partner to pursue a single-minded focus in his working life. These are not, typically, "cookie moms," and they themselves may be supported by domestic help.
After underwhelming Chat GPT5 and growing environmental opposition to power and water requirements, what if the AI future is never getting here outside of some coding help and image and video generation fun?
That would be one of the better case scenarios, actually. If the AI future ever gets here, it probably won't have people in it.
https://ifanyonebuildsit.com/
The case for AI Agents replacing jobs seems further away with every new AI model release. As long as the hallucination rate remains at 5% then any complex workflow will never be able to be replaced by this technology due to compounding errors. Suppose you have a 10 step process, then the success rate is a mere 60%, which is unacceptable to any company.
People in the jobs with AI risk will probably end up using it to become more efficient rather than being replaced by it, unless this can be solved. My company has been trying to leverage it and I've noticed that the leadership has gone from full hype 12 months ago to noting some of these problems now as they get into the implementation weeds.
60% is unacceptable for some tasks but great for others! If you’ve got something that takes a lot of effort to produce but very little effort to check, then having an AI produce 5 of it and a human check them to find the three working ones is probably a lot easier than having the human generate one working one. (I trust you don’t need any examples of the kind of task where 60% success rate is no good.)
Even at its best, a medium which relies on the continuous partial attention of its audience is no place for a chart with two lines going upwards at slightly different gradients. And now that the social media 'zone' is full of Steve Bannon's 'shit', a politician would be daft to try to campaign so prosaically. Neil Postman was right all along. But we are where we are.
Have to admire that Bernie's chart also compared real wages to nominal home prices.
(Statistics never lie, and...)
Regarding AI automation and the lack of impact on jobs, here are a few hypotheses.
Adoption among companies is still pretty uneven, not that many companies have made it to production yet with AI workflows, and for the ones that have, there is still a human in the loop. So let's say you're having AI check the documentation for a car purchase agreement. Someone is still looking at all the work to confirm the AI didn't make a mistake. All the other tasks for that job still exist, too. And to the extent your process is now more productive, maybe you're trying to complete it faster or better and take market share from competitors rather than just banking cost savings. Once your competitors all adopt AI, there won't be as much benefit from being faster, and then maybe cost reduction becomes more of a focus. When the AI version of the workflow becomes sufficiently reliable, or you can introduce another AI to verify the work, that may be another inflection point where the effect on employment becomes noticeable.
AI adoption is also generating a lot of work in terms of integrating the technology, reengineering the business processes in question, and building a data foundation that AI can use to get accurate information. So early adopters of AI may actually be hiring more people in many cases.
Overall, we are definitely in the phase where you have more bank tellers even though there are now ATMs, but I still think at some point a lot of work gets replaced wholesale. I could be wrong, and on top of that, the timing is wildly unpredictable.
I would also quibble with the usage of “overwhelming” in Lyman Stone’s charts. Especially since they imply that despite a significantly higher fraction of women staying home when their financial situation is more conducive to that, most still work!
"but down payments have gotten less affordable"
Erdman blames this on overcautious prudential regulators who overlearned the lessons of under cautious regulations before 2008.
I suspect another effect is that the amount of inherited wealth in the middle classes is increasing, so the size of small inheritances the middle classes can use for down payments is going up.
<Erdman blames this on overcautious prudential regulators who overlearned the lessons of under cautious regulations before 2008.>
It's easy to call the prudential regulators "overcautious" when times are good. Next thing you know, a housing bubble bursts and everybody and their mother starts asking "why didn't you do anything to stop it?" Bureaucrats are gonna do what they've always done: interpret the law in the most cautious way possible to avoid litigation and scrutiny.
On the spousal income chart, it might be helpful to chart median spousal income percentile on the y axis rather than average spousal income. It would also be useful to see if this is average spousal income of all spouses (including non-working ones) or if that chart only shows working spouses, and if it shows mean spousal income before marriage or mean spousal income at present. There are several different stories these numbers might be telling with slightly different versions of this!
"Democrat Party" or "Democratic Party" -- time for a fierce debate!