Actionable insights — Warsh is Wrong About Inflation

1. Model inflation as a ratio, not as a rate

1. Stop treating "inflation" as an output of the policy rate. Write it as a ratio of two growth rates: the growth of the money supply over the growth of the stock of goods and services. Prices rise when the numerator outruns the denominator.
2. Chart the numerator directly — money-supply growth year-over-year — against CPI year-over-year, and look at whether the two lines actually track. If they do, that correlation, not the funds rate, is your forecasting instrument.
3. Sanity-check the model by running it to both extremes. Double the money overnight with the goods fixed: nobody sells the old pizza for the old price, so prices roughly double. Double the goods overnight with the money fixed: abundance destroys scarcity value and prices collapse. If your model doesn't produce both results, it isn't a model of prices — it's a model of one policy lever.
4. Then judge any anti-inflation policy by a single question: which side of the ratio does it move, and in which direction? A tool that leaves both sides untouched cannot deliver the target no matter how forcefully it is applied.

• Money-supply growth re-accelerating while the policy rate is still described as restrictive — the divergence is the tell that the rate is not the binding variable.
• Balance-sheet run-off being slowed, paused or reversed while short rates stay high: the numerator loosening under cover of a tight-sounding rate.
• Supply-side data on the denominator — capex, capacity additions, productivity, business formation — as the half of the ratio nobody quotes on inflation day.
• Any 2% forecast that is justified purely by the level of the funds rate, with no claim about money growth.

2. Ask what the cure does to supply, not just to demand

1. For any tightening policy, trace its effect on both sides of the ratio above, and do the supply side second so it isn't forgotten.
2. Follow the cost of capital into the real economy: dearer debt raises the cost of financing production, which suppresses hiring, borrowing, R&D and new productive capacity. Each of those shrinks the future stock of goods and services.
3. Net the two effects. If the policy restrains demand but restrains supply by as much or more — and does nothing to money creation — the ratio can worsen. The intended cure becomes a cause.
4. Convert this into a portfolio question rather than a debating point: if the tightening path is real and money growth persists, favour scarce real assets and companies that don't need capital markets to fund production, and discount the businesses whose growth is financed rather than earned.

• Capex guidance being cut with financing cost named as the reason — the supply channel closing in real time.
• A widening gap between the rate the Fed sets and the rate at which real production is actually financed (high-yield and bank-loan spreads, not just Treasuries).
• Deregulation and spending restraint as the variables he says actually matter: their absence is the reason to disbelieve a disinflation forecast, whatever the funds rate does.
• The rhetorical warning sign: a policymaker naming one tool and describing the others as crisis-only, which fixes the policy mix before the diagnosis is settled.

3. Removing forward guidance moves volatility; it does not remove it

1. Identify who the largest single price-setter in a market is. If one participant controls a price that is an input to every other price — here, the cost of money, "literally half of every transaction" — then every other participant is obliged to watch them. Reflexivity is a structural fact, not a communication choice.
2. When that participant announces they will stop pre-announcing, do not model it as "less Fed influence." Model it as the same total repricing, delivered later and in fewer, larger increments: the market cannot price in advance what it is not told, so it prices on the day.
3. Re-plan around the new distribution rather than the new average: expect quieter drift between events and sharper gaps at events — meetings, minutes, and the data releases the reaction function keys off.
4. Sanity check with the price-control frame: if you would call a committee setting the price of bread a disaster, apply the identical reasoning to a committee setting the price of capital, and expect the same misallocation and the same loss of the scarcity signal.

• Bigger single-day moves in the front end and in 10-year yields on FOMC days and CPI/payrolls prints, with calmer stretches between — the signature of repricing concentrated into a shorter window.
• Implied vol in rates staying elevated into events even when realized vol between events falls.
• Positioning and stop placement built on an assumed pre-announced path — the thing that breaks first when guidance is withdrawn.
• Whether the Fed's actual behaviour matches the stated posture: a "silent player" that still moves the bond market when it speaks has not changed its role, only its schedule.

4. Read the labour market from both sides — openings alongside unemployment

1. Never read the unemployment rate alone. Pair it with the JOLTS job-openings count, and treat the two together as a 2×2: high unemployment with no openings is a demand collapse; unemployment alongside millions of unfilled openings is a matching or willingness problem, not an absence of work.
2. When openings are abundant, ask why specific openings go unfilled — pay below the worker's alternative income, or a skill the worker will not acquire — because the answer tells you whether the slack is cyclical (fixable by demand) or structural (not).
3. Stress-test the number before leaning on it: assume a share of postings are ghost listings a company has no intention of filling, and check whether your conclusion still holds at the haircut level. If it does, the conclusion is robust to the data's worst-known flaw.
4. Use the pair to judge the economy's tolerance for tightening. A labour market with a large openings buffer can absorb tighter conditions without the unemployment rate spiking; that buffer, not the unemployment rate itself, is the shock absorber to monitor.

• The openings-to-unemployed ratio falling toward 1 — the buffer emptying is what turns tightening into a rising unemployment rate.
• Openings falling while the unemployment rate is still low: the sequence that precedes labour-market cracks, and the one that would invalidate the concession above.
• Quits and hires within JOLTS, which separate genuine demand from stale postings better than the headline openings count.
• Wage trend versus wage level: check whether the growth rate is steeper than its pre-shock trend or merely a spike that reverted — the distinction he draws on average hourly earnings 2:01, and the one that decides whether "wages never kept up" is true.

5. Score a policymaker point by point, and record the concessions

1. Decompose a speech or statement into its separable claims — here: the economy is resilient; it can survive tightening; the Fed should stop driving the bond market; short rates are the inflation tool.
2. Grade each claim independently, and force yourself to state where the policymaker is right. A verdict that agrees with nothing is a prior, not an analysis.
3. For each disagreement, name the specific mechanism you think is wrong rather than the conclusion — a claim about reflexivity, or about which variable drives prices — because only a mechanism generates a testable forward signal.
4. Keep the score. Revisit it at the next speech: the claims conceded are the ones whose reversal would be genuinely new information, and the mechanisms disputed are your dated, falsifiable bets.

• The next Warsh appearance: whether the balance sheet stays crisis-only in practice, which is the claim his whole disagreement rests on.
• A disinflation that arrives without money-supply growth slowing — the cleanest disconfirmation of the ratio model.
• Commentators whose verdict on a speech is uniformly negative: with no conceded points there is nothing in the analysis that could have come out the other way.