Here is a single sixty-year-old in America, buying their own health insurance. At an income of sixty-two thousand dollars a year, they pay about five hundred dollars a month. Now let them earn two thousand dollars more — sixty-four thousand. Their premium jumps to over twelve hundred a month. That is more than eight thousand dollars a year, gone, for earning two thousand more.

They did not get sicker. Their plan did not change. They crossed a line — and on the far side of that line, the help does not shrink a little. It disappears, all at once. This is called a cliff. And the important thing about a cliff is that it is not a fact of nature. Someone drew it. The question is not whether the government should help. It is why, when it decided to help, it built the help as a cliff you can fall off by earning a single dollar.

For five years, from 2021 through 2025, the US ran a more generous version of its health subsidies. No matter how much you earned, your premium for a standard benchmark plan was capped at eight and a half percent of income. Earn more, and the subsidy shrank smoothly. There was no edge to fall off.

That enhanced version expired on the last day of 2025. For 2026 the old rule came back — a hard line at four hundred percent of the poverty level, about sixty-two thousand dollars for a single person. Below it you get help; at exactly the line you still do; one dollar over, and the subsidy is zero. And because the 2026 rules also scrapped the limit on paying it back, an older couple who took the help in advance and earned a dollar too much can be billed for the whole thing at tax time — around twenty thousand dollars.

The fight you will hear is the usual one. One side says these subsidies are too expensive and should stay expired. The other says they are a lifeline and should come back. That is a real fight about how much a country should spend. But it is only about the size of the help.

Look instead at the shape. Whatever you think the government should spend, notice that it chose to end the help with a cliff — a single income line where everything stops at once — rather than letting it fade out gently. That shape is a choice, made separately from the size. The very same subsidy, for the very same money, can be built as a cliff or as a ramp. For five years it was a ramp.

Economists have a plain way to measure a cliff. They ask: for each extra dollar you earn, how much do you actually keep, after taxes and after any help you lose? Most of the time that number is under a hundred percent. At a cliff it goes insane — you earn one dollar and lose thousands. Put the other way round, as a tax on that one dollar: not forty percent, not even ninety, but thousands of percent.

Why build that on purpose? Usually nobody chose the wall. The wall is just the laziest way to aim help at the people who need it most — one clean line, cheap to write, easy to explain. The brutality only shows up at the edge. And we know it does not have to be this way, because for five years it was not: the enhanced version capped premiums at a share of income and let the help fade with no line at all. Same subsidy, just a ramp instead of a wall. The cliff came back because the smoother version was allowed to expire.

This shape is everywhere in how governments help people. Food assistance, the child tax credit, health help, a credit for low earners — each has to fade out as income rises, and they tend to fade around the same stretch of income, so they pile up. A family climbing out of low income can hit all of them shrinking at once, losing most of each new dollar — sometimes more than the whole dollar. Working harder, ending up with less.

But there is an honest catch, and it is the whole reason this is hard. A ramp is not free either. Any help aimed at lower incomes has to shrink as income rises — and shrinking help, spread over a range, is still a tax on earning more. It is just a slope instead of a wall. The clearest proof it can be done well is a program both parties like: the credit for low-wage workers. It is built as a ramp on purpose — it grows as you earn more at the bottom, holds, then eases down. It rewards work instead of punishing it. The tools exist. The cliff is a choice not to use them.

So if a country wants to help people afford something — health coverage, food, rent — without a wall to fall off, what can it do? The first turns the wall into a ramp: cap what people pay at a share of income and let help fade above it, exactly as the expired version did. The price is money and reach — a ramp that never fully stops climbs higher up the income ladder, so it costs more; you have not removed the trade-off, only spread it out. But nobody falls off an edge.

The second gives help to everyone and claws it back through the income tax, which already tapers smoothly, so there is no single cliff. The price is that paying everyone up front is expensive and reads as money for the rich — and the clawback is still a tax on earning more, just hidden in the tax code. The third merges programs so their fade-outs don't stack. The price is complexity, and a single slope set too steep still bites. None of these says stop helping. Each asks: how do you bring help down to zero without a wall someone can fall off by earning a dollar?

So come back to that sixty-year-old — five hundred dollars a month on one side of a line, twelve hundred on the other. The headline will be about Obamacare, and whether Congress should bring the bigger subsidies back. That is a real question about how much to spend. But underneath it is something more durable, and it lives in almost every program that helps people: a decision about the shape of help — whether it fades out on a slope you can walk down, or ends at a wall you fall off.

There can be good reasons to aim help at the people who need it most, and every such program has to end somewhere. But ending it with a cliff is a choice, not a requirement — and a cliff quietly punishes the exact thing we say we want: people earning their way up. So the next time a benefit is debated as simply "too generous" or "not generous enough," that is only the argument about size. The more useful question is about shape: when we build help, why do we so often build a cliff — and who decided that one extra dollar should cost you twenty thousand?