Over-75s Up 38%, Under-35s Down 23% · Morning Gist
Wednesday, October 144 stories · 11 min
Over-75s Up 38%, Under-35s Down 23%
What you own matters as much as what you save: an age-wealth gap, a loosening housing market, drifting portfolios and debt math riding on productivity.
Picked from 46 new articles across 16 sources · arrow keys for other days
Today in 30 seconds
30s left
Median wealth since 2022 rose 38% for households 75+ and fell 23% for under-35s; homes, stocks and cars explain most. story 1 ↓
September existing-home sales: 3.99 million pace, down 2.2% on the year; listings up 4.6%, prices up just 1.3%. story 2 ↓
A 60/40 portfolio untouched since 2020 is now about 77/23; Ben Carlson says 5%+ bond yields make rebalancing worthwhile. story 3 ↓
KfW estimates an extra 0.4% annual productivity growth would cut projected 2056 US debt from 185% to 113% of GDP. story 4 ↓
This year's economics Nobel rewards work on sticky prices, which helps explain why recessions do so much damage.
1/4Of Dollars and Data· Nick Maggiulli· Oct 13
Fed survey: households over 75 got richer since 2022, under-35s got poorer
Nick Maggiulli looked at the Fed's 2025 household wealth survey. He found the age gap came from what each group owns: older households held homes and stocks, which rose, while young households held cars, which fell.
Why it matters · It shows the asset mix of your balance sheet, not just how much you save, decides how a three-year market move treats you. It also makes a practical case for giving to heirs earlier.
Know this first
Context from the editor, not the article · 29 sec
1 · The Survey of Consumer Finances
Every three years the Federal Reserve runs this survey of a few thousand US households, recording their assets, debts and income in detail. It is the main source for how household wealth is spread by age and wealth level.
2 · Median and percentiles
The median is the household in the exact middle. The 20th percentile household is wealthier than 20% of households. Each survey interviews different households, so a group is compared with the same age group three years earlier, not with the same people.
3 · Real (inflation-adjusted) figures
2/4Calculated Risk· CalculatedRisk by Bill McBride· Oct 13
US existing-home sales rose to a 3.99 million annual pace in September
The National Association of Realtors reports that September home resales ran at a 3.99 million annual pace, still near four-year lows. Inventory keeps growing faster than sales, and the median price rose only 1.3%.
Why it matters · If you're buying, selling or watching home prices, the balance of supply against sales is the best early sign of which way prices are heading.
Know this first
Context from the editor, not the article · 25 sec
1 · What existing-home sales measure
The National Association of Realtors (NAR) counts resales of previously owned homes each month. Sales are recorded at closing, so they mostly reflect contracts signed one to two months earlier.
2 · Annualized and seasonally adjusted
Monthly sales are reported as a seasonally adjusted annual rate (SAAR): one month's pace, corrected for normal seasonal swings, then scaled to a full year. That makes months comparable with each other.
3 · Months of supply
Months of supply is current listings divided by the monthly sales pace, or how long it would take to sell everything listed. Higher numbers usually mean buyers have more leverage and price growth slows.
3/4A Wealth of Common Sense· Ben Carlson· Oct 13
A 60/40 portfolio left alone since 2020 is now roughly 77/23
Ben Carlson notes that the stock boom has quietly turned untouched 60% stock, 40% bond portfolios into 77/23. With bonds now paying over 5%, he argues it is time to sell some stocks and rebalance.
Why it matters · If you have a target stock-bond mix, a long bull market has probably pushed you into more stock risk than you signed up for.
Know this first
Context from the editor, not the article · 25 sec
1 · The 60/40 portfolio
A classic balanced mix: 60% stocks for growth, 40% high-quality bonds for income and stability. Because the two often move differently, losses in one can be cushioned by the other.
2 · Drift and rebalancing
As prices move, a portfolio's weights drift away from their targets. Rebalancing means selling some of whatever grew and buying whatever lagged, so the mix returns to the risk level you chose.
3 · Low yields, 2010s to 2021
For most of the 2010s and into 2021, bond yields were near record lows, so bonds offered little income. Rates rose sharply from 2022, which pushed bond prices down and yields up.
The stock return a 60/40 portfolio needs to reach 7% a year at different bond yields.1 / 3
4/4Klement on Investing· Joachim Klement· Oct 13
How much AI productivity it would take to slow US debt growth
A study from KfW, Germany's state development bank, models AI productivity gains of 0.2–0.4% a year. They would cut projected US debt in 2056 from 185% of GDP to between 143% and 113%.
Why it matters · Long-term Treasury yields, mortgage rates and the safety of bond portfolios all depend on whether US debt stabilises. This shows how much depends on one uncertain productivity number.
Know this first
Context from the editor, not the article · 24 sec
1 · Primary deficit vs. total deficit
The primary deficit is government spending minus tax revenue, leaving out interest on existing debt. The total deficit adds interest. Debt grows by the total deficit, so higher interest costs push debt up further.
2 · What drives debt-to-GDP
Debt-to-GDP compares what a government owes with the size of its economy. It falls when deficits shrink, when the economy grows faster, or when interest rates drop below the growth rate.
3 · The productivity paradox
Economist Robert Solow remarked in 1987 that computers showed up everywhere except in the productivity statistics. Big technologies often take years before measured output per worker rises.
All dollar amounts are converted to 2025 dollars, so a gain means more buying power, not just bigger numbers. Inflation was high in 2022–2023, which makes any real gain harder to achieve.
Change in median net worth by age, 2022–2025: big moves only at the youngest and oldest ends.1 / 4
The Federal Reserve's 2025 1, a detailed check of American household wealth taken every three years, shows a sharp split by age. After inflation, 2 for households 75 and older rose 38% from 2022 to 2025, to $504,600. For households under 35 it fell 23%, to $32,820. No age group in between moved more than 6%. Nick Maggiulli of Of Dollars and Data looked at why. He cautions that the SCF samples only a few thousand households, so the direction of each change is more reliable than the exact percentage.
Age, not just wealth
Across all households the pattern looks "3": the 20th 4 lost about 11% while the 70th percentile and above gained about 11% or more. Split by age, that pattern breaks down. Young households lost ground at almost every wealth level, and older ones gained at almost every level.
Compare two households with similar wealth. A young household at the 90th percentile and an older one at the 40th percentile each had about $351,000 in 2025. The young one lost 14% over three years, while the older one gained 38%. Some of the older group's gain comes from wealthier boomers aging into the 75+ bracket. Following the same 5 instead, the gain is 21%.
What they own
The middle half of each age group held different assets:
Homes. Young homeownership fell from 38.5% to 35.6%. The median young owner is worth about $184,000, against $12,000 for the median young renter. Young renters' wealth is actually at a record high, but fewer young people became owners.
Stocks. Stocks, funds and retirement accounts made up 9% of young households' assets, compared with 16% for older ones. US stocks rose about 48% after inflation.
Cars. Vehicles made up 20% of young households' assets, compared with under 6% for older ones. Used-car prices fell back after the pandemic spike, and young households' vehicle values dropped 21%.
Context: a record jump first
From 2019 to 2022, young households' median wealth rose 140%, from about $18,000 to $43,000, the biggest rise on record. Stimulus checks, paused student-loan payments and rising asset prices likely helped. Some of the latest drop may be a return to normal.
Maggiulli still worries about Gen Z, given a record-weak job market for new graduates and the least affordable housing in 40 years. He urges older relatives who can afford it to give while recipients are young rather than leave inheritances, which usually arrive when heirs are in their 50s or 60s. Options include help with a down payment, rent-free housing, or annual gifts under the $19,000 6 ($38,000 for married couples).
The Federal Reserve has published its 2025 1, a detailed snapshot of American household balance sheets taken every three years. The headline is a split by age. After inflation, 2 for households 75 and older rose 38% between 2022 and 2025. For households under 35 it fell 23%. Nick Maggiulli of Of Dollars and Data looked at what caused the gap. His answer is that the assets each group happens to own explain it better than how wealthy each group is.
The numbers by age
Median net worth in 2025 dollars:
Age
2022
2025
Change
Under 35
$42,725
$32,820
−23%
35–44
$148,069
$156,000
+5%
45–54
$269,983
$254,670
−6%
55–64
$398,649
$411,340
+3%
65–74
$448,695
$431,900
−4%
75+
$366,288
$504,600
+38%
None of the middle age groups moved more than 6%. Maggiulli warns that the SCF samples only a few thousand households, so single age brackets are noisy. Readers should trust the direction of the changes more than the exact percentages.
Not simply "the rich got richer"
Across all households, the data looks like a "3" economy, where the top rises and the bottom falls. The 20th 4 lost about 11% of its real net worth. The 70th percentile and above gained about 11% or more. Split by age, though, that pattern breaks down.
Among under-35s, net worth fell at nearly every percentile. The one exception was the 80th, which Maggiulli suspects is a sampling quirk. Among households 75 and older, net worth rose everywhere except the bottom 10%.
He uses a matched comparison to show that wealth level alone isn't the driver. In 2025, a young household at the 90th percentile had about $351,000, roughly the same as an older household at the 40th percentile. The young one lost 14% over three years. The older one gained 38%. Less wealthy older households gained even more: 26% at the 20th percentile and 45% at the 30th. If the size of a household's wealth were what mattered, households with similar wealth would have moved together. They didn't.
Some of the older group's gain comes from wealthier baby boomers moving into the 75+ bracket. Maggiulli followed the same 5, people 72 and older in 2022, so 75 and older by 2025. Their median rose from $418,720 to $504,600, or 21%. That is smaller than 38% but still large.
What drove it: three asset classes
Maggiulli compared the average change in each asset and debt for the middle 50% of each age group. Older households gained on their primary residence, retirement accounts and other assets. Younger households lost on housing, vehicles and retirement accounts.
Housing. The two groups own stocks, retirement accounts and cars at similar rates, but not homes. Homeownership among under-35s fell from 38.5% in 2022 to 35.6% in 2025, so about 64% of young households now rent. The gap between owning and renting is huge. The median young homeowner is worth about $184,000 and the median young renter about $12,000, roughly 15 times less. So a drop of a few points in ownership pulls the typical young household down a lot.
Young renters did not get poorer. Their $12,000 median is the highest in SCF data going back to 1989. The decline comes from fewer young people crossing into ownership. An illustration (not from the article): if 3 of every 100 young households move from the "worth ~$184k" group to the "worth ~$12k" group, the median household shifts toward the renter end, even though no one in either group lost money.
Financial assets. In 2022, stocks, funds and retirement accounts made up about 9% of total assets for the middle half of under-35s. For the middle half of 75+ households the share was 16%. US stocks rose about 48% after inflation from 2022 to 2025, so older households had nearly twice as much of their balance sheet in the rally.
Cars. Vehicles were about 20% of total assets for the young middle half, compared with under 6% for the old. Used-car prices spiked during the pandemic and then came back down. The median value of young households' vehicles fell 21%, while older households' vehicles barely changed (−1%).
The asset mix explains the result. Homes and stocks went up, cars lost value, and each age group held a different mix of the three.
A step back after a record jump
Going back to 2019 changes the picture. Under-35 median net worth climbed from about $18,000 in 2019 to nearly $43,000 in 2022. That 140% rise was the largest on record. Over the same years, the 75+ group gained a modest 13%. Maggiulli says the cause of the young households' jump isn't known for certain. Likely contributors were pandemic stimulus checks, the pause on student-loan payments and rising asset prices. Those supports have ended, so the 2025 figure may partly be a return to normal. Young households are still well above their 2019 level in real terms.
Maggiulli's argument: give now, not later
In 2021 Maggiulli argued that millennials were not poorer than earlier generations, and he says the data later bore that out. He is less confident about today's youngest adults, many of them Gen Z. He points to the weakest job market for recent graduates on record, which he links partly to AI, and the least affordable housing in 40 years. That threatens the two main routes to building wealth: rising income and owning a home. Older households, he notes, gained largely because their assets went up in value, not because they did anything differently.
His suggestion is to pass money down earlier. Most heirs inherit in their 50s or 60s, when the money matters far less than it would at 30. He suggests:
helping with a down payment
letting adult children live at home rent-free
using the annual 6: $19,000 per recipient, or $38,000 from a married couple
offering job contacts and mentoring
What stays open
The SCF's small sample makes subgroup percentages unreliable. The cause of the 2019–2022 surge for young households is unknown. The cohort adjustment shows that some of the older group's gain comes from wealthier people aging into the bracket. The asset breakdown describes only the middle half of each group. Whether young households' lower homeownership is temporary, caused by high rates and prices, or lasting, the article doesn't say.
Remember
Similar-wealth households moved in opposite directions by age, so asset mix, not wealth level, drove the split.
Young renters reached record wealth, but falling homeownership dragged the young median down.
Young households' 2022 peak followed a record 140% pandemic jump, so part of the decline is a return to normal.
Existing-home sales at an annual pace since 1994, showing four years near multi-decade lows.1 / 4
Sales are still near four-year lows
The 1, the trade group that tracks resales of previously owned homes, says existing-home sales rose 1.5% in September 2026. That put them at a 3.99 million2: one month's pace, corrected for normal seasonal swings and scaled to a full year. Sales were 2.2% lower than a year earlier. The median price was $417,500, up just 1.3%.
The monthly gain looks better than it is. August was revised down from 3.98 million to 3.93 million, and the 1.5% is measured from that lower number. Against August's first reported figure, September is essentially flat. As Bill McBride of Calculated Risk puts it, "Sales have been very low for the last 4 years." For scale (general background, not from the post), resales ran at about 5.3 million in 2019.
Inventory keeps building
The number of homes listed for sale rose to 1.60 million. That figure isn't seasonally adjusted: listings normally peak in late summer and bottom out in winter. So the useful comparison is with the same month last year, and on that basis inventory was up 4.6%. McBride notes that this 3 growth has started speeding up again.
4 tells the story
Months of supply is the number of listings divided by the monthly sales pace: how long it would take to sell everything on the market at today's pace. In September, 3.99 million ÷ 12 is about 333,000 sales a month, and 1.60 million listings ÷ 333,000 ≈ 4.8 months. That was unchanged from August and above the 4.0 months of September 2019.
That's McBride's key point. Listings are still below pre-pandemic levels, but sales have fallen even more, so relative to demand there is now more supply than before the pandemic. Slow price growth fits that picture. A common rule of thumb treats about six months of supply as balanced, so the market still leans toward sellers, but much less than it did.
What the post doesn't cover
The free portion doesn't say why sales have stayed low. The usual explanation is that owners with cheap mortgages won't sell, while high rates and prices keep buyers out. McBride says the year-over-year drop matched early local-market data. His comparison of 2025 and 2026 monthly sales before seasonal adjustment is paywalled.
The September numbers
The 1, the trade group that tracks resales of previously owned homes, reported that existing-home sales rose 1.5% in September 2026. That put them at a 2 of 3.99 million. SAAR takes one month's sales pace, corrects it for normal seasonal swings, and scales it to a full year. Sales were 2.2% lower than in September 2025. The median sale price was $417,500, up 1.3% from a year earlier.
Bill McBride of Calculated Risk sums up the bigger picture: "Sales have been very low for the last 4 years." For scale (general background, not from the post), resales ran at about 5.3 million in 2019, so the current pace is roughly a quarter lower.
The revision matters
August sales were revised down from 3.98 million to 3.93 million. The 1.5% monthly gain is measured from the lower, revised figure. Against the 3.98 million first reported for August, September's 3.99 million is up only about 0.25%. Read the headline as "sales are flat at a low level," not "sales are picking up." Sales also count at closing, so September's figure mostly reflects contracts signed in July and August.
Inventory: read it year over year
The number of homes listed for sale rose to 1.60 million, up 1.9% from an upwardly revised 1.57 million in August. McBride notes that this inventory figure is not seasonally adjusted. Listings normally peak in mid-to-late summer and hit their lows in December and January, so month-to-month changes are mostly seasonal noise. Comparing with the same month a year earlier removes that pattern. On that basis inventory was up 4.6% from September 2025, and McBride notes that 3 inventory growth "has started increasing again recently."
4: more slack than before the pandemic
Months of supply divides current listings by the monthly sales pace. It answers one question: at today's pace, how long would it take to sell everything on the market? A worked example with September's figures:
Monthly sales pace: 3.99 million ÷ 12 ≈ 333,000 homes
Listings: 1.60 million
1.60 million ÷ 333,000 ≈ 4.8 months
That was unchanged from August. In September 2019 the figure was 4.0 months. McBride's point is that there is now more supply relative to demand than before the pandemic, even though the raw number of listings is still below 2019. Sales have fallen further than inventory has, which pushes the ratio up.
One quirk: the calculation divides inventory that is not seasonally adjusted by sales that are. That is another reason to compare like months across years, such as September with September.
What it means for prices
The article stops short of a price forecast, but the numbers are consistent with each other. A median price up just 1.3% year over year is a slow gain for housing, and it comes while months of supply sits above its pre-pandemic level. A common rule of thumb (general convention, not from the article) treats about six months of supply as a balanced market. Below that favors sellers, above it favors buyers. At 4.8 months the market still leans toward sellers, but much less than it did when supply was very tight.
The article does not explain why sales have been so low for four years. The usual explanation, outside this post, is that many owners locked in very low mortgage rates and don't want to sell and take on a higher rate, while high prices and rates keep buyers out. That would also explain why inventory recovers only slowly.
What's left open
McBride adds that sales fell year over year in September "as expected using local data," meaning early reports from local markets had already signaled the drop. His not-seasonally-adjusted comparison of 2025 and 2026, and the analysis that followed, are behind a paywall, so this write-up covers only the free portion. Things to watch next: whether year-over-year inventory growth keeps speeding up, whether months of supply climbs past 5, and whether price growth turns negative in some regions.
Remember
Against August's first reported figure, September sales were flat; the 1.5% gain comes from a downward revision to August.
Because inventory isn't seasonally adjusted, compare it year over year. It's up 4.6% and that growth is speeding up.
Months of supply is 4.8, above 2019's 4.0. Supply is looser relative to demand than before the pandemic.
Ben Carlson, a wealth manager who writes the blog A Wealth of Common Sense, has a simple message for balanced investors: the stock rally has probably left you holding more risk than you meant to, and now that bonds pay a real yield, it's time to fix that.
Why many people moved into stocks
When interest rates were very low in the 2010s and early 2020s, bonds paid almost nothing. If you wanted a decent return from a 1 (60% stocks, 40% bonds), stocks had to do most of the work. Carlson says many investors who would have preferred 60/40 shifted to 80/20, and some put everything into stocks.
His team's chart shows the arithmetic. Here's an illustrative version: for a 60/40 mix to earn 7% a year when bonds earn 1%, stocks have to return about 11%. If bonds earn 5%, stocks only need about 8.3%.
Diversification worked, just backwards
Bonds have done badly this decade, but stocks more than made up for it. Carlson points to a 2 fund that holds a 60/40 mix and has returned nearly 8% a year in the 2020s. In his view, the 60/40 portfolio "did not die." This time stocks were the part that rescued bonds. Next time it could be the other way around.
The case for rebalancing now
Carlson says high-quality bonds now yield more than 5%, for the first time in almost 20 years. That means a balanced portfolio no longer needs a huge stock market to earn a reasonable return. He argues that leaning heavily into stocks made sense when yields were tiny, but now investors should go back to their target weights.
Drift matters too. A 60/40 portfolio set up at the start of the 2020s and never touched is now about 77/23. Its risk looks much more like an aggressive stock portfolio than a balanced one.
Not a market call
Carlson doesn't predict a crash. The spending boom on artificial intelligence could keep the bull market going for five more years, or it could end this year. His point is that nobody knows, and that uncertainty is the reason to diversify in the first place. An 3, meaning your chosen mix of investments, only works if you have a 4 to go with it.
He adds some caveats. You don't need to rebalance constantly, you should let winners run to some degree, and you should think about taxes before selling. Rebalancing "will never feel right during a bull market," he writes, because it means selling what has done best. He also admits he doesn't know when it will pay off. His argument is that today's best performers will eventually become the worst ones.
Ben Carlson, a wealth manager who writes the blog A Wealth of Common Sense, argues that investors with a balanced portfolio should rebalance now. After years of strong stock returns, a portfolio that started as 60% stocks and 40% bonds and was never adjusted now holds about 77% stocks. Bonds also pay meaningful income again. Together, he says, those two facts mean the case for leaning heavily into stocks has faded.
How low rates pushed investors into stocks
For most of the 2010s and into the early 2020s, interest rates sat near historic lows. Carlson says investors who wanted a balanced portfolio had two choices: accept lower returns from their bonds, or take more risk to make up for it. He says many chose more risk. Investors who would naturally have held 60/40 moved to 80/20, some went entirely into stocks, and others moved into strategies riskier than bonds.
His research team made a chart showing the stock return a 1 needs to reach a 7% annual return, at different bond yields. The math behind it is a weighted average. Here's an illustrative version that treats a bond's return as roughly equal to its yield:
To get 7% with bonds at 1%, stocks must return (7 − 0.4) ÷ 0.6 ≈ 11% a year.
With bonds at 5%, stocks need only (7 − 2) ÷ 0.6 ≈ 8.3%.
In early 2020, at the height of the pandemic, bond yields were at the far low end of that chart. Hitting 7% required a very strong stock market.
Diversification worked, in an unexpected direction
The stock market did deliver. Bonds have done badly this decade, because rising rates push down the prices of existing bonds. Stocks more than made up for it. Carlson cites a 2 fund, a simple off-the-shelf 60/40 portfolio, that has returned nearly 8% a year in the 2020s.
He disputes the popular claim that the 60/40 portfolio is dead. Diversification worked, he argues, but this time stocks were the asset that bailed out the other holding. Usually people expect bonds to cushion a stock slump. He suggests that next time the cushion may come from bonds.
Why the math has changed
Carlson says high-quality bonds now yield more than 5%, for the first time in almost 20 years. He notes that a whole generation of newer investors has never invested with yields this high, and that investors with larger, more mature portfolios are looking to reduce risk.
Using the arithmetic above, a 5% bond yield means stocks no longer have to "knock it out of the park" for a balanced portfolio to earn a decent return. "Bonds are finally earning their keep," he writes. He calls tilting your portfolio beyond its target toward stocks "3." In his view, that was reasonable early in the decade when yields were tiny. Now, he argues, it probably makes sense to go back to target weights.
What 4 looks like
The second problem is drift. Carlson's chart follows a hypothetical 60/40 portfolio set up at the start of the 2020s and left alone. It now sits at about 77/23. An illustrative way to get there: $60 in stocks grows to about $134 while $40 in bonds stays roughly flat. Stocks are then $134 of $174, or 77%. Getting back to 60/40 would mean selling about $30 of stocks and buying bonds with it.
The risk shift matters more than the exact figure. A 77/23 portfolio behaves much more like an aggressive stock portfolio in a downturn than the 60/40 its owner chose.
Rebalancing as risk management, not a forecast
Carlson makes no prediction about timing. He says the bull market has already run longer than almost anyone expected. The spending boom on artificial intelligence could extend it another five years or more. It could also end this year, give way to a recession, or be derailed by something nobody sees coming.
That uncertainty is the core of his argument. You diversify because you can't know what comes next. If you could, you wouldn't need to. He describes a chain of three steps:
Diversification happens by setting an 5: target weights across asset classes, regions and strategies.
That allocation should reflect your risk tolerance, your time horizon, and how you react emotionally to losses. Carlson says no allocation is perfect. The goal is to keep regret low by balancing short-term spending needs against long-term goals.
An allocation is useless without a 6, a rule for when and how you bring weights back to target.
Caveats he acknowledges
Carlson doesn't call for mechanical, constant trading. You don't need to rebalance all the time. Letting winners run is part of investing. In taxable accounts, selling winners can trigger capital-gains tax, so where and how you rebalance matters. He doesn't go into specific methods, such as rebalancing on a calendar, rebalancing when weights drift past a set threshold, or steering new contributions toward the underweight asset. Those are common ways investors handle exactly the tension he describes.
He's also candid about how it feels. Rebalancing "will never feel right during a bull market," because it means selling your best performers. His reasoning is that leadership rotates: eventually today's best holdings become the worst and vice versa. He admits he doesn't know when rebalancing will pay off, only that at some point it will.
What's left open
This is a short opinion piece, not a study. It doesn't state the exact assumptions behind the drift or required-return charts. It doesn't compare rebalancing rules or estimate what rebalancing has historically added or cost. It also doesn't address the obvious counterpoint: anyone who rebalanced diligently since 2020 gave up some stock gains. His case rests on risk control rather than return-maximizing. You're meant to rebalance because the target mix is the portfolio you decided you could live with.
Remember
When bonds yield 5% rather than 1%, a 60/40 portfolio needs about 3 points less from stocks to earn 7%.
An untouched 60/40 portfolio from 2020 now carries about 77% stocks, far more risk than intended.
Rebalancing is risk control, not a market forecast; Carlson admits he can't say when it will pay off.
Range of published forecasts for AI's annual productivity boost, from cautious to bullish.1 / 4
A short study from 1, Germany's state development bank, asks what AI-driven productivity growth could do for US government debt. Investment writer Joachim Klement calls it the clearest treatment of the topic he has seen.
Nobody agrees on the size of the boost
Forecasts of AI's effect on productivity (output per hour worked) range from MIT economist Daron Acemoglu's very low estimates to Goldman Sachs's much higher ones. Klement sides with Acemoglu and the 2, Congress's nonpartisan budget office. He expects about 0.1% a year.
Timing is the other question. The study's author, Stephan Bales, shows that the steam engine, electricity and the computer each went through years or decades of hype with little productivity growth. Gains came only after a "3" emerged. Klement thinks AI may be in that hype phase now, with a breakthrough perhaps 5 to 20 years away.
From productivity to the budget
Bales assumes the gains start immediately and last. The chain then runs like this:
Higher productivity raises GDP growth.
Faster growth brings in more tax revenue and reduces welfare spending.
That shrinks the 4, the budget gap before interest payments.
Less borrowing means less interest owed, and probably lower bond yields too.
Debt also shrinks relative to a GDP that is now growing faster.
The numbers
Scenario
Primary deficit, 2056
Debt-to-GDP, 2056
No AI boost
2.8% of GDP
185%
AI adds 0.2%/yr
1.5%
143%
AI adds 0.4%/yr
roughly zero
113%
So 0.4 percentage points of extra annual growth, compounded over 30 years, removes more than 70 points of 5.
The bond-market bet
Klement argues that anyone buying 30-year US 6 today is implicitly betting that AI will substantially raise growth. If it doesn't, debt keeps rising, and yields could drift higher to compensate lenders. Rising yields mean falling prices for existing long bonds.
Caveats
Klement's own 0.1% forecast is below even the smaller scenario, so his expected fiscal benefit is modest. The scenarios also assume immediate gains, while the study's own history suggests a long delay. The piece doesn't explain how the study models interest rates, and it summarises the KfW work without giving its full method.
Joachim Klement, an investment strategist who writes the newsletter Klement on Investing, points readers to a short study he calls the clearest treatment of its subject he has seen. It comes from Stephan Bales at 1, Germany's state-owned development bank. The question is what AI-driven productivity growth would do to the US government's debt over the next 30 years. Bales's answer is that it would do a lot, provided the gains are real and arrive soon.
The forecasts disagree widely
Productivity here means output per hour worked. Its long-run growth rate sets how fast an economy can grow without causing inflation. Estimates of AI's contribution vary hugely. At the low end is MIT economist Daron Acemoglu, whose widely cited 2024 estimate works out to well under 0.1 percentage points a year. At the high end is Goldman Sachs, which in 2023 suggested AI could eventually add about 1.5 percentage points a year. Bales's study collects these forecasts in a single chart.
Klement places himself in the cautious camp, alongside Acemoglu and the 2 (Congressional Budget Office, Congress's nonpartisan budget forecaster). He expects about 0.1% a year.
Timing may matter more than size
The study also looks at history. The steam engine, electricity and the computer each went through years, sometimes decades, of excitement while productivity growth stayed at or near zero. Growth took off only after someone found the "3", the use that reorganised how work was done. Electricity is the textbook case: factories had to be redesigned around individual electric motors before the gains appeared.
Klement suggests AI may be in that early phase now, with plenty of hype and little measurable payoff. In five, ten or twenty years, someone might find the breakthrough use that raises AI's productivity effect "by orders of magnitude."
How productivity reaches the budget
For the fiscal exercise, Bales assumes the optimistic timing: AI raises productivity right away and keeps doing so. The chain from there works like this:
Faster productivity growth means faster GDP growth.
Faster GDP growth raises tax revenue, because incomes and profits are higher, and lowers welfare spending, because fewer people need support.
Both effects shrink the 4, the gap between spending and revenue before interest payments.
Smaller deficits mean less new borrowing. Less debt means less interest to pay. Klement adds that bond yields would probably fall too, cutting interest costs further.
Meanwhile GDP, the bottom of the 5 ratio, grows faster.
The effects compound. A purely illustrative example, not from the study: suppose debt is 100% of GDP and the deficit is held fixed. If GDP grows 0.4 percentage points a year faster, that alone leaves the ratio about 11% lower after 30 years. Add smaller deficits and lower interest payments and the gap widens much further.
The numbers
Under the baseline projection, the US primary deficit in 2056 is 2.8% of GDP.
Scenario
Primary deficit, 2056
Debt-to-GDP, 2056
No AI boost
2.8%
185%
AI adds 0.2%/yr
1.5%
143%
AI adds 0.4%/yr
roughly eliminated
113%
An extra 0.4 percentage points of annual productivity growth would cut the projected 2056 debt burden by more than 70 points of GDP. That is a lot for what sounds like a small number.
Klement's bond-market argument
From this, Klement draws an investment conclusion. Anyone buying 30-year US 6 today, he argues, is implicitly betting that AI will raise productivity and GDP growth significantly. If it doesn't, debt keeps climbing, and Treasury yields "may drift much higher" to compensate lenders for the growing burden. When yields rise, the prices of existing long bonds fall. Long-dated bonds lose the most, because their payments stretch furthest into the future.
What to keep in mind
Klement's own forecast sits below both scenarios. He expects about 0.1% a year, half the smaller case modelled. By his own numbers, the likely fiscal benefit is modest and debt still heads well above 143% of GDP.
The scenarios assume the gains start now. The history section argues against that. If the payoff takes a decade or two to arrive, the compounding has less time to work.
The interest-rate assumptions aren't spelled out. The piece says yields would "likely" fall but doesn't say how the study models them. In debt projections, the rate assumption often matters as much as the growth assumption.
It's a summary of someone else's work. Klement describes the KfW note rather than reproducing its methods, so readers can't check the baseline, the tax and spending responses, or whether growth would also push up some spending, such as wage-linked benefits.
The broader point is that the long-run path of US debt is very sensitive to productivity growth. Fractions of a percentage point, compounded over 30 years, separate a debt load near today's level from one near twice the size of the economy.
Remember
Fractions of a percentage point of productivity growth, compounded over 30 years, change projected US debt by tens of points of GDP.
Past general-purpose technologies took decades to raise productivity, which weakens scenarios that assume gains start now.
Long-dated government bonds are exposed to growth outcomes as well as inflation and Federal Reserve policy.
