The apparent skew upward of income is based in false assumptions and weightings of the real cost of middle-class life. The trend lines presented are not a reflection of the rise of net wealth of the American middle-class. Instead, they reflect an undercounting of inflation of core costs of what was once considered to be the real measures of being middle-class in America (mortgaged home ownership, overall costs of one or more automobiles, family food costs, and a college degree). All of these things are far more expensive in real terms than they were in the 1970s. Middle-class consumer debt loads are far, far higher today. Economists need to reweigh cost of living indicators and net wealth measures, otherwise the policy decisions based upon them will continue to be based in a false sense that things are up, up, up for the American middle-class!
RealTime Inequality, from U.C. Berkeley economists, shows these figures, 1976 to 2023:
The top 1% income increases by 333% or $1.5 million (from$452K to $1.9 million)
The 90% to 99% top incomes increase by 132% or $160K (from $123K to $283k)
The middle 50% to 90% income increases by 78% or $43K (from $60k to $102K)
The lower 50% increases its income by 28% or $6K (from $20K to $26K) roughly.
This is factor (market) income for adults 20 to 65 years.
Post-tax incomes of course look better, but not a lot in my opinion.
RealTime also shows a shift in national income distribution, about 15.4%. The lower 90% was earning 63%, now it earns 48%. RAND Corporation also shows similar shift, "Income Trends 1975 to 2018, with a more recent update to 2021. Both posit the counterfactual that incomes would be far higher for the majority if the distribution ratios of 1975 were reestablished. The difference is shocking. RAND says that the median income would not be $50,000 by $92,000.
I discovered that the median sales price of a house in 1973 was 4.7 times the median yearly income of the nonsupervisory worker. In 2025 the house was 7.6 times. -- From Fed's FRED graphs). That's 62% higher. Housing cost is the biggest section of a family budget. Far bigger than the cost of a computer, which negates the point of your comparison. Chinese imports have also lowered the cost of many household goods. Taxes are also down.
David Madland produced a report showing this, in maybe 2018, showing increased cost of housing, healthcare, higher education and childcare. Sounds right.
Between January 1973 and January 2026 "Real Disposable Income per capita" has grown by 154% (Fed FRED). The Real GDP per capita also showed about 140% growth.
In the same period, per the EPI graph you show, middle income growth has been 23%.
High income has been 40%, and low income 65%.
The United Way's report ALICE states that 42% of adults live with hardship or poverty. A Brookings Inst. report says 43%. The Supplemental Poverty Measure, U.S. Census, says 41.5% live with incomes below 200% OPL. "Making Ends Meet" report from the CFPB, 2024, says that 52% cannot afford a $2,000 emergency expense. Looks like poverty to me.
Wonderful post. But the thought of 215,000 people selling their plasma daily to make ends meet reminded me of "Whatever you do to the least of these, you do to me."
I remember being really shocked coming from Indiana for a week in San Francisco, and hearing my San Francisco well-paid friends discuss how their engineering jobs were insufficient for their lifestyle which drove them to be Uber drivers in the evening.
The New York Times also had a piece on how the current tax season is affecting people. I was struck by the fact that one household that needs a new car and earns $48k per year went from a $4k+ return to about $1800 this year. And another family bringing in just over $500k got such a big windfall they are buying a car for their 16-year old daughter. The first family has to give up necessities to address the transportation issue. Equally striking — a retired couple that is getting a bigger windfall claims to be embarrassed by it. They said the money should be going to help those less fortunate. I expected them to say they are donating it to a charity — but no, they are taking a trip to Ireland.
I would be much more interested in reading your columns if you would follow the rules found in Strunk and White’s “The Elements of Style.” Having to stop and puzzle over abbreviations like “ftr” distracts from what you are trying to communicate. Why shoot yourself in the foot?
But the emerging AI revolution will drive a much bigger share of income to the 1% while slamming the college educated middle class into a problem similar to blue collar workers in the Rust Belt.
On cost-of-living, on the healthcare component, around the recent effects of the ACA expanded subsidies that happened Jan 1, 2026.
1) I note that whatever increase there is in what people have to pay for health insurance (it seems, on average, maybe 4% or 5% of income for the 24 million people affected), this increase is NOT captured, at all, in the CPI. (Because the ACA subsidies are a means-tested benefit, which is excluded from CPI.)
2) Since the public use files for the 2026 open enrollment period have recently become available, I have done a little computing myself, looking to see enrollment-decreases in the particular case, people who are now over the returned-400%-of-Federal-Poverty-Level "subsidy cliff".
(Motivated by that group, facing complete loss of subsidy, has increases in what they have to pay for health insurance (of after taxes money) commonly around 20% to 40% of (pre-tax) income.)
This approach, for looking at the effects of the expiration of the ACA expanded subsidies, seems to be new. Much of the reporting so far has been that about 5.1% of the 24 million people dropped coverage so far, with some proportion of them also downgrading their plan level, say from silver to bronze.
However, I did get much bigger numbers dropping coverage on the just-over-the-returned-cliff-group.
I got about 38.6% dropping coverage in the 400%-500% FPL reported income group!
(Somehow, I seem to have beat Charles Gaba to the calculation! Perhaps he even missed seeing the aspect, in his oodles of posts he did using the same data!)
If anyone's interested, the info, with by-state breakdowns, and little notes on interpretation and other finer details, is here:
(Addionally, if anyone is trying to be really up on the issue of what happened due to the expired ACA expanded subsidies, note that the currently-available numbers are from the end of open enrollment, in January. They do not generally reflect that some people, many of them counted as enrolled because they were auto-reentrolled by the exchange, will see that their premium has become too high to afford, and will not pay it. Others are expected, over the coming months or years, to find the premium becomes too much of a financial burden after some time of paying it. So, in short, all numbers from the current CMS datasets, which reflect status immediately after open enrollment, are expected to show further losses of coverage over time.)
I've probably commented before in your substack, I think even if you go into fine distributions of "real wages" and "real wealth" (say listing at at each percentile), the basic problem is that "real" relies on a Consumer Price Index. CPI is simply an attempt to get into a single 1-number time-series something that is much, much, too complex.
You have touched on that issue with:
"(In the context of this discussion, the comparison invokes the key issue of relative price trends: some very cool, living-standard-enhancing goods have fallen in (quality-adjusted) price, while some old, boring necessities, including housing, childcare, healthcare, electricity, have gone up faster than average (this was Annie Lowrey’s key insight years ago, in the early affordability takes)."
Frankly, sharp critical thinking would make one see at the outset that something like a Consumer Price Index, though of some use, is bound to fail at describing reality well over time. There is too much changing of needs, and available products, going on.
(I suspect that many potential young economists in training with a natural disposition towards very careful thought, and superior critical thinking skills, are in unbelief about such oversimplifications as CPI, and get dis-motivated, and flunk out of economics and switch to pure math. (Playful jibe!) )
I have seen Dean Baker as well, with similar "what can it really mean?" problems with the CPI.
(On top of the issues with that, as time passes over decades, there is a changing selection of available goods, and changing needs at each fixed period of time of a person's life, I also note that healthcare advances, say since 1950, cause both more expenses to be needed by individuals on health care, as well as longer lives. The longer lives necessitate more savings needed for retirement. I believe this is nowhere to be accounted for in the CPI. Is the magnitude of unaccounted-for cost-of-living increase from this healthcare-consequences-effect, since 1950, 30%?)
There is a problem with the statement “everyone is doing better”. Yes, you can create a graph where low, middle, and high income groups all see their average income rising, but we must remember that roughly half the people in each of those groups will find their incomes rising less than the mean.
I’m in the high income group, but I haven’t had a raise in literally 10 years. My wife has had her compensation cut in half. So although we’re still doing OK in absolute terms we’re finding it harder to keep up. Each year less goes into savings as expenses rise and we approach retirement.
Every graph you see of compensation increases over the past year has a big spike at zero. Those are people that are being squeezed hard by inflation even if average wage increases exceed the inflation rate. I haven’t a buffer. A lot of people don’t.
Would not the chart showing the price indices for computers vs. homes give a more relatable comparison if the scales on the left and right showed the same % increase per vertical gradation?
The apparent skew upward of income is based in false assumptions and weightings of the real cost of middle-class life. The trend lines presented are not a reflection of the rise of net wealth of the American middle-class. Instead, they reflect an undercounting of inflation of core costs of what was once considered to be the real measures of being middle-class in America (mortgaged home ownership, overall costs of one or more automobiles, family food costs, and a college degree). All of these things are far more expensive in real terms than they were in the 1970s. Middle-class consumer debt loads are far, far higher today. Economists need to reweigh cost of living indicators and net wealth measures, otherwise the policy decisions based upon them will continue to be based in a false sense that things are up, up, up for the American middle-class!
RealTime Inequality, from U.C. Berkeley economists, shows these figures, 1976 to 2023:
The top 1% income increases by 333% or $1.5 million (from$452K to $1.9 million)
The 90% to 99% top incomes increase by 132% or $160K (from $123K to $283k)
The middle 50% to 90% income increases by 78% or $43K (from $60k to $102K)
The lower 50% increases its income by 28% or $6K (from $20K to $26K) roughly.
This is factor (market) income for adults 20 to 65 years.
Post-tax incomes of course look better, but not a lot in my opinion.
RealTime also shows a shift in national income distribution, about 15.4%. The lower 90% was earning 63%, now it earns 48%. RAND Corporation also shows similar shift, "Income Trends 1975 to 2018, with a more recent update to 2021. Both posit the counterfactual that incomes would be far higher for the majority if the distribution ratios of 1975 were reestablished. The difference is shocking. RAND says that the median income would not be $50,000 by $92,000.
I discovered that the median sales price of a house in 1973 was 4.7 times the median yearly income of the nonsupervisory worker. In 2025 the house was 7.6 times. -- From Fed's FRED graphs). That's 62% higher. Housing cost is the biggest section of a family budget. Far bigger than the cost of a computer, which negates the point of your comparison. Chinese imports have also lowered the cost of many household goods. Taxes are also down.
David Madland produced a report showing this, in maybe 2018, showing increased cost of housing, healthcare, higher education and childcare. Sounds right.
Between January 1973 and January 2026 "Real Disposable Income per capita" has grown by 154% (Fed FRED). The Real GDP per capita also showed about 140% growth.
In the same period, per the EPI graph you show, middle income growth has been 23%.
High income has been 40%, and low income 65%.
The United Way's report ALICE states that 42% of adults live with hardship or poverty. A Brookings Inst. report says 43%. The Supplemental Poverty Measure, U.S. Census, says 41.5% live with incomes below 200% OPL. "Making Ends Meet" report from the CFPB, 2024, says that 52% cannot afford a $2,000 emergency expense. Looks like poverty to me.
My blog -- http://benL88.blogspot.com -- more details like this.
Wonderful post. But the thought of 215,000 people selling their plasma daily to make ends meet reminded me of "Whatever you do to the least of these, you do to me."
I remember being really shocked coming from Indiana for a week in San Francisco, and hearing my San Francisco well-paid friends discuss how their engineering jobs were insufficient for their lifestyle which drove them to be Uber drivers in the evening.
The New York Times also had a piece on how the current tax season is affecting people. I was struck by the fact that one household that needs a new car and earns $48k per year went from a $4k+ return to about $1800 this year. And another family bringing in just over $500k got such a big windfall they are buying a car for their 16-year old daughter. The first family has to give up necessities to address the transportation issue. Equally striking — a retired couple that is getting a bigger windfall claims to be embarrassed by it. They said the money should be going to help those less fortunate. I expected them to say they are donating it to a charity — but no, they are taking a trip to Ireland.
I would be much more interested in reading your columns if you would follow the rules found in Strunk and White’s “The Elements of Style.” Having to stop and puzzle over abbreviations like “ftr” distracts from what you are trying to communicate. Why shoot yourself in the foot?
But the emerging AI revolution will drive a much bigger share of income to the 1% while slamming the college educated middle class into a problem similar to blue collar workers in the Rust Belt.
On cost-of-living, on the healthcare component, around the recent effects of the ACA expanded subsidies that happened Jan 1, 2026.
1) I note that whatever increase there is in what people have to pay for health insurance (it seems, on average, maybe 4% or 5% of income for the 24 million people affected), this increase is NOT captured, at all, in the CPI. (Because the ACA subsidies are a means-tested benefit, which is excluded from CPI.)
2) Since the public use files for the 2026 open enrollment period have recently become available, I have done a little computing myself, looking to see enrollment-decreases in the particular case, people who are now over the returned-400%-of-Federal-Poverty-Level "subsidy cliff".
(Motivated by that group, facing complete loss of subsidy, has increases in what they have to pay for health insurance (of after taxes money) commonly around 20% to 40% of (pre-tax) income.)
This approach, for looking at the effects of the expiration of the ACA expanded subsidies, seems to be new. Much of the reporting so far has been that about 5.1% of the 24 million people dropped coverage so far, with some proportion of them also downgrading their plan level, say from silver to bronze.
However, I did get much bigger numbers dropping coverage on the just-over-the-returned-cliff-group.
I got about 38.6% dropping coverage in the 400%-500% FPL reported income group!
(Somehow, I seem to have beat Charles Gaba to the calculation! Perhaps he even missed seeing the aspect, in his oodles of posts he did using the same data!)
If anyone's interested, the info, with by-state breakdowns, and little notes on interpretation and other finer details, is here:
https://normspier828307.substack.com/p/aca-2026-enrollment-after-expanded
(Addionally, if anyone is trying to be really up on the issue of what happened due to the expired ACA expanded subsidies, note that the currently-available numbers are from the end of open enrollment, in January. They do not generally reflect that some people, many of them counted as enrolled because they were auto-reentrolled by the exchange, will see that their premium has become too high to afford, and will not pay it. Others are expected, over the coming months or years, to find the premium becomes too much of a financial burden after some time of paying it. So, in short, all numbers from the current CMS datasets, which reflect status immediately after open enrollment, are expected to show further losses of coverage over time.)
I've probably commented before in your substack, I think even if you go into fine distributions of "real wages" and "real wealth" (say listing at at each percentile), the basic problem is that "real" relies on a Consumer Price Index. CPI is simply an attempt to get into a single 1-number time-series something that is much, much, too complex.
You have touched on that issue with:
"(In the context of this discussion, the comparison invokes the key issue of relative price trends: some very cool, living-standard-enhancing goods have fallen in (quality-adjusted) price, while some old, boring necessities, including housing, childcare, healthcare, electricity, have gone up faster than average (this was Annie Lowrey’s key insight years ago, in the early affordability takes)."
Frankly, sharp critical thinking would make one see at the outset that something like a Consumer Price Index, though of some use, is bound to fail at describing reality well over time. There is too much changing of needs, and available products, going on.
(I suspect that many potential young economists in training with a natural disposition towards very careful thought, and superior critical thinking skills, are in unbelief about such oversimplifications as CPI, and get dis-motivated, and flunk out of economics and switch to pure math. (Playful jibe!) )
I have seen Dean Baker as well, with similar "what can it really mean?" problems with the CPI.
(On top of the issues with that, as time passes over decades, there is a changing selection of available goods, and changing needs at each fixed period of time of a person's life, I also note that healthcare advances, say since 1950, cause both more expenses to be needed by individuals on health care, as well as longer lives. The longer lives necessitate more savings needed for retirement. I believe this is nowhere to be accounted for in the CPI. Is the magnitude of unaccounted-for cost-of-living increase from this healthcare-consequences-effect, since 1950, 30%?)
There is a problem with the statement “everyone is doing better”. Yes, you can create a graph where low, middle, and high income groups all see their average income rising, but we must remember that roughly half the people in each of those groups will find their incomes rising less than the mean.
I’m in the high income group, but I haven’t had a raise in literally 10 years. My wife has had her compensation cut in half. So although we’re still doing OK in absolute terms we’re finding it harder to keep up. Each year less goes into savings as expenses rise and we approach retirement.
Every graph you see of compensation increases over the past year has a big spike at zero. Those are people that are being squeezed hard by inflation even if average wage increases exceed the inflation rate. I haven’t a buffer. A lot of people don’t.
Would not the chart showing the price indices for computers vs. homes give a more relatable comparison if the scales on the left and right showed the same % increase per vertical gradation?
Yep. But point comes across either way and I thought this was simpler.