The Right Autopsy, the Wrong Corpse

Bill Nelson has written a careful, well-sourced obituary for M2 as a guide to inflation. Every word of it is correct.

And it settles nothing, because the aggregate he buries is not the one anyone serious is asking the Fed to watch — and the mechanism he says monetarists never supply has been sitting in plain sight the whole time.

Bill Nelson is one of the most careful writers on Fed plumbing alive, and his latest essay — prompted by M2’s cameo in the Monetary Policy Report and Warsh’s “I think money matters” — is a small masterpiece of demolition. It should be read. It is also, in the way of the most elegant demolitions, aimed at a building nobody is defending.

Let me concede the whole case first. Simple-sum M2 is a broken gauge, and has been since the early 1980s. Greenspan said so in 1993 — the historical relationships between money and income had “largely broken down.”

The money-growth ranges in the old reports were a vestigial ritual by the late 1990s. The money multiplier is dead, as Carpenter and Demiralp showed, and deader still now that reserve requirements are gone.

Reserves are not in M2 and do not drive the money in public hands; shrinking the balance sheet does not mechanically shrink inflation. And the 2021 M2 surge was, in large part, exactly what Nelson says — frightened households and firms parking transfer checks in deposits at zero interest, a rise in savings that inflated the aggregate for a reason that had little to do with the classic “too much money chasing too few goods.”

On every one of these, Nelson is right, and anyone still targeting simple-sum M2 deserves the beating.

But watch the move at the center of the essay, because it is the whole game. Nelson demolishes simple-sum M2, and then, in a parenthesis, waves the coup de grâce at everything else: “I guess Divisia monetary aggregates are next up,” with a link to the Hudson Bay note and a shrug.

That parenthesis is where the argument quietly cheats. He has performed a meticulous autopsy on one corpse and asked us to accept it as the autopsy of a body that is still walking around.

Why simple-sum M2 was always going to fail

Here is the thing Nelson knows and does not say. Simple-sum M2 fails for a reason, and the reason tells you what to do instead. Adding up currency, checking deposits, savings deposits, small time deposits, and retail money funds — and giving every dollar of each the same weight — assumes that a dollar locked in a savings account provides the same monetary service as a dollar in your checking account ready to be spent. It does not.

One is money; the other is a short-term asset that happens to be denominated in money. Sum them with equal weights and you have built an index that moves for reasons that have nothing to do with spending — like a rush into savings during a pandemic. Which is precisely the artifact Nelson correctly identifies in 2021.

The fix for that flaw is not to give up on money. It is to weight each component by the monetary service it actually provides — to count the checking dollar fully and the locked-up time-deposit dollar barely at all. That is what a Divisia aggregate does.

It is not an exotic new candidate in the doomed parade of M1, M2, M3, P-star. It is the correction of the specific measurement error that killed all of them. Barnett showed in the 1980s that the simple sum is a theoretically indefensible index number — you would never add up quantities of different goods without weighting them by price, and money is no different.

The reason M2 kept breaking is that it was the wrong arithmetic. Nelson has spent an essay proving the wrong arithmetic gives wrong answers and treating it as proof that arithmetic is hopeless.

And the empirical literature he skips past is not a fringe. The credit-card-augmented Divisia work of William A Barnett and co-authors, the Serletis studies, Barrette and Paquet’s 2025 paper running the aggregates from 1967 through 2023 — these are recent, peer-reviewed, and they keep finding that properly weighted money carries information about output and inflation that the simple sum destroys. You can argue with them. You cannot dismiss them with a parenthetical “I guess.”

The mechanism he says doesn’t exist

Nelson’s strongest-sounding point is the oldest one, and he attributes it to Tobin: monetarists never describe the precise mechanism linking money to inflation. He asks, pointedly, how — in the United States, right now — money and inflation are actually linked.

It is a fair question, and it has an answer, and the answer is one Nelson himself writes down without noticing. Read his own account of 2021 again. Deposits swelled, he says, and the link to future inflation ran “through the traditional Keynesian impact of savings on future consumption and aggregate demand.” Yes. Exactly.

That is the mechanism. Money matters because it becomes spending — because those balances are latent nominal demand that gets released when confidence returns and rates rise. Nelson thinks he is describing an alternative to the monetary story. He is describing its transmission.

This is where his framing does the most damage to his own case. He treats “money → inflation” and “aggregate demand → inflation” as rival explanations, one monetarist and discredited, one Keynesian and sound.

But they are the same chain viewed at two distances. Money is not inflationary because of the letter M on a Fed table. It is inflationary when it is spent — when nominal expenditure rises faster than the economy can produce goods. The aggregate is a leading, imperfect proxy for that spending; the spending is the thing. Nelson has, in effect, conceded the mechanism and denied the word.

The aggregate was never the target

Which brings us to the deepest problem with the essay, the one that makes its careful demolition beside the point. Nelson writes as though the only two options are (a) revive a monetary aggregate as a guide to policy, or (b) admit money is irrelevant and move on. He spends the piece foreclosing option (a), and treats option (b) as the residue.

But the serious position is neither. Money is not the target. Nominal spending is — nominal GDP relative to a stated level path. The right monetary aggregate, weighted properly, is a useful confirming indicator of where nominal demand is heading before the GDP data arrive, which is exactly the modest claim the Hudson Bay note Nelson links actually makes: not that we should target aggregates, but that they carry information that should not be thrown away.

Nelson beats the maximal monetarist position — target M2, watch the balance sheet, mind the multiplier — and leaves the minimal one untouched, because he never engages it. He argues against 1979 and declares victory over 2026.

And there is an irony in his conclusion he does not seem to feel. He ends — rightly — by skewering Warsh for saying “inflation is a choice” and “the Fed will take ownership” while refusing to specify what the Fed would actually do.

Resolve is not a plan; a commitment to nothing is a commitment to nothing. On this Nelson and I agree completely. But he has spent the preceding pages removing, one by one, the candidates for what a plan might be — money, the multiplier, the balance sheet — without putting anything in the empty space he is now demanding Warsh fill.

If not money, and not the balance sheet, then what is the quantity the Fed commits to? Nelson leaves the question hanging as an accusation against Warsh. It is equally an accusation against his own essay.

The answer, of course, is the one neither man will say. The plan Warsh lacks and Nelson gestures toward is a commitment to a path for nominal spending — the level of NGDP, which the Fed can steer with all its tools and be held to in public every quarter.

Watch a properly weighted monetary aggregate to see the pressure building; commit to the spending path to do something about it. Nelson is right that M2 is a dead guide and right that resolve is not a plan. He has simply written the second half of a two-part argument and called it complete. The corpse on his table is real. The patient is in the next room, and its name is nominal GDP.

A Note: If you have been reading Money Fetish for a while and find it worth your time, I would ask you to consider whether it is also worth five dollars a month — roughly the price of a coffee you will have forgotten by tomorrow.

Every paid subscription provides the means for constant improvement: more time for research, deeper engagement with the ideas, and the independence to follow the argument wherever it leads without looking over my shoulder at anyone’s preferences but the reader’s.

Paid subscriptions are available at $5 a month or $50 a year. If you are already a paid subscriber, thank you. If you are not, the option is always there when you are ready.

Either way — thank you for reading.