The Average American Doesn’t Exist

The Arithmetic Is Already Written

*On the two-track economy, and what it will cost the party that ignores it*

---

Margaret Thatcher's most durable predictions were not moral arguments. They were arithmetic. When she told the Commons in 1990 that a single European currency without a single European treasury would end in crisis, she was not expressing a preference. She was pointing at a contradiction and following it to its necessary conclusion: bind twelve fiscal policies to one monetary policy, and eventually one of them will break in a direction the others cannot absorb. It took eighteen years and a Greek balance sheet to prove her right, and by then almost no one remembered that she had said it first.

The method is worth borrowing even by people who disagreed with everything she did with it. Find the structural contradiction. Follow it. Name what would prove you wrong. Put a date on it.

Here is one.

## The contradiction

There are two American economies now, and the aggregate statistics are averaging them into a single number that describes neither.

In the first quarter of 2026, the Mortgage Bankers Association reported that delinquency on FHA-insured mortgages reached 11.88 percent — roughly one borrower in nine behind on payments. VA loans hit 4.99 percent. Over the same quarter, conventional mortgage delinquency *fell* fourteen basis points. The MBA's own analyst called it the widest spread since 2021. FHA foreclosure inventory is at its highest level since 2018; VA's since 2017.

That is not a downturn. A downturn moves everything in the same direction. This is a divergence, and divergence is what a two-track economy looks like from inside the data.

The tracks sort by entry point, not by income. Conventional borrowers bought earlier, with larger down payments, and locked in rates below four percent. FHA and VA borrowers bought later, with less cushion, at rates twice that, into a housing market where the ratio of median price to median income sits at levels not seen since before the war. One group is insulated by a decision made in 2019 and 2020. The other is exposed by having been three years too young to make it.

The aggregate number — total mortgage delinquency of 4.44 percent — describes an economy that no one actually lives in.

## Why this is political

The insulated group and the exposed group do not sort by party. That is the entire point, and it is the thing both parties are currently getting wrong.

FHA and VA borrowers are disproportionately rural, exurban, younger, and — in the VA case — veterans. These are the demographic foundations of the current Republican coalition. The people falling behind are, in substantial part, the people who elected the government now telling them the economy is strong.

The political consequence of that arrangement is not persuasion. It is dissolution. A voter who cannot make a mortgage payment while equity indices set records does not typically switch parties. He stops believing that either party is describing the world he lives in, and then he either stays home or he waits for someone who talks about what he can see out the window.

We already have a measurement of what that looks like. In the March 2026 Texas primaries, Democratic turnout more than doubled its 2022 and 2024 levels, reaching roughly 2.3 million against 2.16 million Republican ballots — the first time since 2020 that Democrats out-turned Republicans in a Texas primary. The detail that matters is not the margin. It is that the marquee contested race, the three-way brawl between an incumbent senator, the attorney general, and a sitting congressman, was on the *Republican* ballot. Republicans had the reason to show up. They did not.

Primary turnout is a poor predictor of general elections and anyone who tells you otherwise is selling something. But it is an excellent measure of enthusiasm, and enthusiasm asymmetry is the mechanism by which structurally safe seats become competitive ones.

## The prediction

Republicans spent 2025 and 2026 redrawing congressional maps in Texas, Florida, Ohio, Missouri, North Carolina, and four states since. The goal was to convert a two-seat House majority into a durable one. The method was to take districts that were safe by fifteen points and make them safe by six, freeing up voters to flip Democratic seats elsewhere.

This is leverage. Leverage amplifies whatever direction the underlying asset moves. Drawn against a stable electorate, those maps hold comfortably. Drawn against an electorate in which one in nine entry-level homeowners is behind on a mortgage, they are a bet that nothing changes.

So: **Democrats take the House in November 2026, most likely by twelve to twenty seats.** The Senate is closer to a coin flip than anyone in either party will admit publicly, turning almost entirely on Ohio and Texas, both of which currently sit inside a single point.

That is the small prediction. Here is the larger one.

**The gerrymanders of 2025 and 2026 will be remembered as a strategic error, but not for the reason their critics expect.** They will not produce a spectacular backfire in a single cycle. They will produce something slower and worse for the party that drew them: a decade in which two dozen Republican members have no margin for a bad year, in a period that is likely to contain several.

Because the second wave is coming, and it is larger. If artificial intelligence substitutes for cognitive labor at anything like the pace its own developers project, the displacement will land on precisely the constituency that has never before been economically expendable: credentialed, high-turnout, mortgage-holding professionals. Paralegals. Analysts. Junior developers. Radiologists. The people who did everything they were told to do.

When mechanization did this to agricultural labor between 1870 and 1900, the political result was the Populist movement — a program that lost every presidential election it contested and then had nearly every one of its planks written into law within twenty-five years. Income tax. Direct election of senators. The secret ballot. The eight-hour day. The movement was destroyed by absorption. The agenda won completely.

That is the shape of what is coming, and the timing is the only genuine uncertainty.

## What would prove this wrong

An honest prediction names its own failure conditions. Four would falsify this one:

**Housing affordability recovers as rates normalize.** If the price-to-income ratio corrects when monetary policy loosens, then this was a rate cycle and not a structural break, and everything above is overwritten.

**FHA and VA delinquency converges back toward conventional by mid-2027.** The MBA has attributed part of the current spike to the expiration of pandemic-era relief and to trial payment plans that count as delinquent until permanent workouts land. If the divergence closes as those wash through, the two-track thesis was an accounting artifact.

**Real wage growth outpaces cost of living through 2027.** Material improvement dissolves political realignment faster than any argument. It has done so repeatedly.

**AI displacement follows the pattern of computerization, containerization, and the ATM** — genuine disruption, absorbed labor market, no aggregate employment collapse. This is the most likely of the four, and the one this argument is most exposed to.

Predictions about technological displacement have an excellent record on direction and a dismal one on timing. Mine may be no different.

## The part that is not a prediction

If the analysis holds, the political opening it creates will be seized by whoever has a program drafted when the window opens — and the window after a shock is roughly eighteen months.

The Populists did not win because the crisis arrived. They lost the crisis. Their agenda won because it had been written, argued, and refined for fifteen years by the time a major party needed something to say. Roosevelt did not invent the New Deal in 1933. He picked it up off a shelf where it had been sitting since Omaha.

The lesson is not to wait for the collapse. It is to have the answer finished before anyone asks the question.

---

*The four falsification conditions above are all measurable, and three of them will report within eighteen months. Check them.*