Invention's Other Half, Part 2: The Flexibility Tax
A person is offered a better job in another state, and they turn it down. Not from fear. From arithmetic. They have a 3% mortgage. The new house is at least 7%. On the same $300,000 loan, that is about $1,265 a month against $1,995, an extra $730 for the identical house. To hold the old payment they would have to trade 2,500 square feet for something closer to 1,600.
They can’t afford to leave the loan. Their net worth has never been higher. Yet, their freedom has never been lower. And every month, their loan payment is recorded somewhere as a strong, performing asset.
Let’s be clear, they’re not a victim; they’re optimizing. The cheap mortgage is a valuable thing; a rational person doesn’t throw it away. The choice is privately sensible and collectively immobilizing. The economists Fonseca and Liu measured it. The wider the gap between a homeowner’s locked-in rate and the market rate, the less likely they are to move. In the recent rate spike, that was as much as nine to sixteen percent. The cost is the move that never happens, the reallocation frozen in place.
GDP and the balance sheet book the debt as strength. The loan sits in the asset column, the payer reads as a strong borrower, the house as growing wealth. Nothing anywhere records that the person can’t move. The account holds the debt on the asset’s side, so every month that pins them in place makes a country look richer. It will never count the job they couldn’t take.
When debt is an on-ramp, and when it’s a toll
Most debt is ordinary, and a lot of it bought good things. The HVAC expert down the street has a mortgage, a truck loan, and a balance left from their kid’s tuition, and none of those were mistakes. The line between an on-ramp and a toll runs by what the debt does, not by how it happened to turn out.
Debt is an on-ramp when it buys an appreciating thing whose returns you keep. It’s a toll when it services a sunk or shrinking position with a claim that sits ahead of your own freedom. Run the repair worker’s three through it.
For example, the truck loan is an on-ramp, since it widens the radius of jobs the expert can reach. The mortgage depends on the location, an on-ramp where the locale is growing and dead weight where it isn’t. The tuition balance is the toll, if their kid’s field just got automated: a claim on the family’s freedom against a position the machine already shrank.
A claim on the family’s income against a job the machine already shrank. It takes the wheel.
An on-ramp in a growing town, dead weight in a shrinking one. The same loan, two fates.
Buys an asset he keeps and widens the radius of jobs he can reach.
The most mortgaged generation in American history was the postwar one, which pushed homeownership from around 44 percent to 62. It was also among the most upwardly mobile. If debt were simply the flexibility tax, that should be impossible. It worked then because the payments were small against fast-rising incomes, and the mortgage bought an appreciating asset the family got to keep. The toll arrives later, when income stops outrunning the obligation and the borrowing shifts toward the sunk and the depreciating.
Credit builds the good half too, the on-ramp the first part was full of. The charging grid was poured on borrowed money. The power lines ran out to farms the market had written off.
The finance literature has a name for where the curve turns. Total private credit, households and firms together, lifts growth until it reaches somewhere around 80 to 100 percent of GDP, and past that point it starts to drag.
The United States is well past it.
GDP hears the ding, not the later
A sixty-dollar dinner goes on a card. The drawer opens, and sixty dollars of GDP exist that instant. The consumer is “strong.” What nobody records is the second number. They’ll pay seventy-eight dollars over the next three years. They came in tonight precisely because the card let them not feel the difference. The ding is what the country hears. The later is what only they hear, when the statement comes.
National accounts record spending no matter how it’s financed. Income, savings, or borrowing, it all counts the same when you buy the dinner or the firm buys the server. The debt that paid for it shows up only in the financial accounts, which nobody steers by. So borrowing to spend props up the number we do steer by, and the obligation it created is invisible to that number.
GDP isn’t a measure of welfare, but it’s the number we optimize. It defines “growth,” and it anchors policy and markets. But it cannot tell you whether the growth was bought with equity or with debt. So in its own terms, a debt-financed boom reads exactly like the real thing. The flattery runs one way: it favors debt. Simon Kuznets built American national-income accounting. In 1934 he warned Congress that “the welfare of a nation can scarcely be inferred from a measurement of national income.” We kept the number and lost the warning.
The ding is what the country hears. The later is what only he hears.
The “debt tax” is real, not a metaphor
What debt degrades is measurable: a society’s capacity to reallocate, to move to the job, switch trades, start the firm, absorb the new. A new tool or trade spreads only as fast as people can take it up, and a household pinned by a fixed payment takes up nothing. A tool you can finance: buy it, keep it, done. A trade you can only reach by surviving the year you are nobody, and a fixed payment is the month that says you can’t be.
Di Maggio, Kalda, and Yao found a natural experiment. A servicer could no longer prove it owned a pile of private student loans, so courts discharged them, on grounds unrelated to who the borrowers were. The people whose debt vanished moved more, changed jobs more, and earned about three thousand dollars more over the next three years. Same people. Only the debt removed. This is the distressed tail, borrowers already in collections, which is exactly what made the experiment clean.
The same pattern turns up wherever a fixed obligation meets someone who needs to move. Negative home equity cuts household labor supply two to six percent. Rate lock-in holds people in the house from the top of this essay. And a standard-deviation rise in student debt tracks about fourteen percent fewer small-business starts.
Credit frees people sometimes, then drags past a point. The borrower and the lender do different things with a freed dollar. The constrained borrower has a high propensity to act on it, to move, to switch, to start something. The pension fund holding the note reallocates no labor at all. Debt concentrates fixed obligations on precisely the people most likely to need to move. Mian and Sufi found the macro version: high household debt amplifies downturns and predicts slower growth on the far side. The dollars are conserved across a loan; meanwhile, the reallocation is not.
You can put a figure on the standing weight of it. Before an American household decides anything in a given month, about one dollar in nine of its after-tax income is already committed to debt service . The total stock tops eighteen trillion dollars, a record in raw dollars.
The average isn’t what we should watch
By international standards American household leverage is moderate, below its 2007 peak against income, because households deleveraged hard after 2008. That is why the average misleads. The flexibility tax doesn’t live in the mean. It lives in who carries it: the student and medical debt loaded onto the specific frozen workers the first part was about. It is the share of a country pinned in place, and no scoreboard reports it yet.
Deep DiveWhere the average lies: household debt, country by country
America’s total looks moderate because the mean hides the shape. The IMF measures household debt as loans and securities against GDP. By that measure the United States sits below several peers: 126 percent in Switzerland, 110 in Australia, 102 in Canada, and about 73 in the US. That is well under the 90-plus it carried in 2010. To an accountant that reads as the measured, prudent economy. To the person at the window it reads as a country that spread its leverage thin and still froze, because the total was never the problem. Look again at Switzerland. It owes far more than America against its output, and is nowhere near as stuck. The number on the balance sheet says nothing about whether a household could take a hit. The burden was never the size of the debt. It is whether a shock ends you, and that lands hardest on the people least able to carry it, the frozen workers of the first part. The mean hides who holds the note, and the amount hides what it does to them.
Same instrument, opposite effects, one blind scoreboard
The first part’s $380 billion-plus build-out, the machines, is increasingly borrowed. Hyperscalers issued more than a hundred billion dollars of bonds in 2025. Bank research reckons another hundred and twenty billion or so has been pushed off their balance sheets. It went into special-purpose vehicles that own the data centers, while the tech company signs a long lease. Meta’s largest such deal, around thirty billion dollars, ran through a vehicle named “Beignet Investor.” The biggest private-credit data-center financing in history, named after a doughnut. So the machine the first part celebrated is financed by the very mechanism this part is naming. The on-ramp and the toll run on the same borrowing, and GDP counts them identically, both as strength. It cannot see the debt, its fragility , or which kind it is. The Bank for International Settlements has started warning about exactly this “shadow borrowing.”
One person’s debt is another’s asset; in aggregate it nets to zero. The claim that locks the borrower in place is also somebody’s income: a private-credit fund on the corporate side, a retiree’s bond fund on the household side. And that retiree needs the certainty more than the mobility.
Net worth won’t save the scoreboard
We do keep balance sheets. The Federal Reserve puts American household net worth around 184 trillion dollars. It doesn’t help. Net worth nets debt against asset prices. So a debt-fueled bubble flatters the books at the very moment fragility is rising. And it measures wealth, not reallocation capacity. You can be rich on paper and frozen in place. That is the person at the window.
The worry Richard Koo named isn’t imaginary. When everyone deleverages at once, the adjustment itself freezes the economy, the way Japan’s did for a decade. That is an ex-post warning, and Koo’s own prescription is not to force the write-down early but to absorb it over time on the public balance sheet. The argument here is the ex-ante half of that same coin. See the overhang forming, before the panic. Then you never have to choose between forcing it and absorbing it. There is already a number that points the right way, the household debt-service ratio; watch it beside GDP and you see what GDP hides. But a monthly payment is only the visible edge of the tax. The deeper question is whether a household could take a surprise hit and keep moving, or whether one bad month ends the crossing . Two countries can get the same tools into the same hands. The one whose people can absorb the shock is the more adaptable, even if its GDP reads lower, and even if it owes more. The finale puts that on one chart.
Next: The Share That Can’t Move, the gauge I built, and which measure of debt actually predicts who moves.
