The Seed Corn Problem

There is a real case for cutting federal research money. At first glance the numbers back it.

Since the Apollo years, federal research spending has collapsed as a share of the economy, falling by roughly two thirds. And over exactly the same period, patents went up sevenfold and total American research spending hit records.

So the obvious conclusion is that private money stepped in and nothing was lost. Both sets of numbers are real. And the way they fit together is the entire subject of this piece.

This framework’s job is to ask whether the abundance is real. Here it is not.

Those three lines look like a story about substitution: private money quietly replacing public money, nothing lost. They are actually a story about what kind of research got bought. And that difference decides everything.

What actually happened to the money

Three things happened underneath those lines, and none of them shows up in the totals.

First, the government stopped being the funder. In 1964 Washington paid for about 67 percent of all American research and development. By 2022 it paid for roughly 18 percent. Meanwhile the national total climbed to $937 billion in 2023. More money, from someone else.

Second, the money moved toward the safe end. Of that $937 billion, only $138 billion went to basic research, the kind where nobody knows what they will find. $174 billion went to applied research. And $625 billion, roughly two thirds of everything, went to development. That is engineering a known thing into a sellable product.

Which is exactly what you would expect a company to buy. It is not what anybody else was buying.

Third, and this is the one that matters, the money for the first mile is going away. At the turn of the century the federal government paid for about 60 percent of American basic research. That share has fallen steadily, and business is now close to overtaking it there too.

Which sounds fine until you ask why the government was ever doing it. Basic research is the part with no owner and no timetable. You cannot patent a question. No shareholder wants to fund something that pays off in thirty years, to somebody else.

The patent line proves almost nothing

Patents are the most quoted evidence that the cuts are harmless. They are also the weakest. Three problems.

They arrive decades late. The gap between a basic discovery and a patented product routinely runs fifteen to thirty years. So the patents granted in 2024 are largely harvesting science funded in the 1970s, 1980s and 1990s, when the federal share was two to three times what it is now. We are counting a harvest and calling it a forecast.

The unit itself got cheaper. Patent counts inflated after a 1982 court change. Then again when software and business methods became patentable in the 1990s. Then again with defensive patenting, where companies file to build a legal shield rather than because they invented anything. A patent in 2024 is not the same object as a patent in 1964.

And most of them are not American. This is the most direct evidence available and it is almost never mentioned. Foreign owners held 53 percent of US utility patents in 2022, up from 44 percent in 2000. The line labelled US patents is increasingly a record of other countries’ research, filed here because this is where the market is.

The yield is falling

If record research spending and record patent counts were producing record results, you would see it in productivity. That is the only measure that captures whether new knowledge is actually making an economy more capable. You do not see it.

So here is the picture. Research spending at an all-time high. Patents at an all-time high. And the rate at which the economy is actually getting better at things running roughly a third lower than it did back when federal research money was two to three times larger.

The honest caveat is that productivity has many parents and nobody has proved a single cause. Maybe we used up the easy gains from cars and electricity. Maybe the 1990s computing boom simply faded. Maybe we measure a service economy badly.

All of that is fair. But the composition shift is documented, the timing lines up, and the mechanism is not mysterious.

What is being cut, and what that does

The 2025 and 2026 cuts are not a trim around the edges. The administration proposed taking the National Science Foundation from $8.8 billion to $3.9 billion, a cut of more than half, alongside roughly $18 billion from the National Institutes of Health. At one point in the 2025 financial year, new NSF grants were running at about 20 percent of the pace of each of the previous four years.

That last number is the one this framework weighs most heavily, and it is the same argument as the public health piece. A budget can be restored in a single year. A cohort cannot.

A doctoral student whose funding vanishes in 2026 does not wait around. They take a job. They go abroad. They do something else with their life. And the specific person who would have spent forty years becoming the world expert in something nobody has named yet simply never becomes that.

The scarcinality reading

The treatise describes this in two stages. In the first, a money shortage is fake. The people and the equipment are all still there, and only the claim on them was withdrawn, so restoring the money restores the system. In the second stage the fake shortage has lasted long enough to destroy the thing it was only pretending to threaten. Then restoring the money restores nothing, because what it would have bought no longer exists.

Research is the purest example, because its output stays invisible for so long that destroying it costs nothing politically in the year you do it.

Cut the National Science Foundation today and nothing happens. No hospital closes. No bridge falls. Nobody loses a service they can name. The damage arrives in the 2050s, as an absence, in the form of things that simply never got discovered. And absences do not generate a headline.

And the deficit used to justify the cut is, in this framework’s terms, a shortage of money. It is real on its own level and worth taking seriously. But it is being relieved by destroying something physical that cannot be bought back later at any price, which is the exact trade the treatise says never works.

Education spending shows the same shape from another angle. It climbed steeply from about 2 percent of the economy in 1940 to a plateau near 5.4 to 5.7 percent by the mid 1970s, and it has essentially not moved in fifty years. What changed was who pays, as costs slid from states onto households.

The falsifiable claims

This framework cannot tell you what the discoveries would have been. Nobody can. That is the nature of research, and it is why the argument for funding it has always had to rest on structure rather than on prediction. What it can say is which layer is being spent to relieve which. A shortage of claims is being eased by consuming a stock of real capability, and that stock takes decades to rebuild.

Seed corn is the oldest metaphor in economics for exactly this, and it survives because it is precise. Eating it gives you a good year and a catastrophic decade. And the decisive part, the part that makes it happen again and again, is that the good year comes first.