Pushing on the Price — What Intervention Can and Cannot Move

Two stories this summer look like they belong in different briefings. A finance ministry spent tens of billions of dollars buying its own currency, and the currency kept weakening. A warming Pacific started pushing up the price of coffee, cocoa, grain and fertiliser, and no government anywhere had a lever that reached it.

They are the same problem in two costumes. In both cases an authority can act on a price while the force actually setting that price stays exactly where it was. The intervention is real, the money is real, and the effect is real — for a while. What it buys is time, not direction, and the two get confused constantly, usually by the people who most want them to be the same thing.

This briefing is a structural map of that distinction: where intervention genuinely reaches, where it only appears to, what it costs to find out, and the single question worth asking before reading any announcement of decisive action.

Start with the cleaner case, because the arithmetic is visible.

Japan’s currency has been under sustained downward pressure, and the reason is not mysterious. Money flows toward yield, and the gap between US and Japanese policy rates has been wide — on the order of four percentage points. Borrow cheaply in yen, hold something that pays more elsewhere, pocket the difference: the carry trade is the oldest idea in currency markets and it works precisely as long as that gap stays open.

Against that, Japan’s Ministry of Finance — which holds sole legal authority to intervene, with the central bank acting only as its executing agent — has gone into the market repeatedly this year. The yen was bought when the currency pushed past one-sixty in the spring, the first such operation since 2024, and again at the end of July in an operation estimated at around eighty-five billion dollars.

The operations did what operations do: the currency moved sharply, immediately, and in the intended direction. Then it resumed weakening.

The reason is structural rather than a failure of execution. Spot purchases change who holds what today. They do not change the rate differential, which is what makes the trade profitable tomorrow. So the incentive survives the intervention intact — and, worse, the intervention improves the terms on which that incentive can be expressed. A carry trade that was entered at one level can be re-entered at a better one after the authorities have pushed the price up. Analysts have described recent operations as having turbo-charged the carry trade rather than broken it, and the mechanism in that phrase is exact: the defence subsidises the attack.

This is not an argument that intervention is pointless. Slowing a disorderly move has genuine value — it buys time for an adjustment to happen elsewhere, prevents a self-feeding panic, and signals that a line exists. But it is time, bought at a price, and the question that decides whether the purchase was worth it is always the same: what is supposed to change while the time is being bought? If the answer is “the rate gap closes,” the intervention is a bridge. If there is no answer, it is a subsidy.

Now the harder case, where the authority in question has no lever at all.

Forecasters moved to an El Niño watch in the spring and then to high confidence: a strong warming of the Pacific developing through the middle of the year, forecast to peak across autumn and winter, with a very high probability of conditions persisting into early 2027. Unusually for a macro variable, this one comes with a published calendar months in advance.

What it does on the way through is uneven and physical. Rainfall redistributes — drought in some growing regions, excess in others — and the damage lands inside specific planting and harvest windows that cannot be reopened later. South Asia, southern Africa and parts of East Asia carry the heaviest rice exposure. India has already seen a monsoon running well below normal in the early season, with cotton, soy, corn and rice sitting in the affected window. Brokers have cut cocoa surplus estimates on West African weather risk. Coffee, chocolate, palm oil and tropical fruit all sit directly in the transmission path.

The prices have already moved. Global food prices are at their highest since 2022, with corn up by roughly a third from its summer low and sugar up by about a quarter over a shorter window. Separately, and compounding it, fertiliser is forecast to rise around thirty per cent on average across this year, taking affordability back to its worst since 2022 — which matters because fertiliser is an input to next season, not this one.

Putting the two cases side by side makes the general shape visible.

The chain from a warm ocean to a headline inflation print has five links, and only the top two are reachable by anybody with a policy instrument. No finance ministry affects sea-surface temperature, rainfall distribution, or what a damaged crop yields. Those links are forcing, and they are already determined by the time anyone argues about them.

What is reachable is the buffer layer — strategic reserves, import policy, subsidies, export restrictions — and then the monetary response at the top, which arrives last and acts on an economy-wide average rather than on the thing that actually moved. This is why food-driven inflation is such an uncomfortable object for a central bank: the instrument available is a blunt one aimed at demand, while the cause was a supply shock that has already happened and will unwind on its own schedule regardless.

The currency case has the same architecture with different contents. The forcing layer is the rate differential. The reachable layer is spot intervention. Acting on the reachable layer while the forcing layer is untouched produces exactly what it produced: a large, immediate, temporary move.

Which gives the general test. Before treating any intervention as decisive, locate it on the chain. Is it acting on the force, or on the price the force is setting? Both are legitimate; they just have completely different half-lives.

Three things cut against the tidy version.

Climate forecasts are probabilistic, and their translation into crop outcomes is weaker than their translation into weather. A strong signal in the Pacific raises the odds of specific regional patterns; it does not guarantee them, and historical episodes have varied widely in agricultural impact for similar ocean readings. Anyone quoting a precise yield loss for a season that has not finished is extrapolating well past what the science supports.

Intervention also works more often than the cynical account allows. There are documented cases of operations that succeeded — generally where they were aligned with a turn in the underlying driver rather than against it, and where they were coordinated across more than one authority. “Intervention never works” is as wrong as “intervention fixes it.” What survives is the narrower claim: intervention against an unchanged structural driver does not persist.

And the food-price transmission into consumer inflation is heavily mediated by things that have nothing to do with the harvest. In wealthier economies, the raw commodity is a small fraction of retail food cost — processing, packaging, labour, energy and retail margin dominate — so a dramatic move in a futures price produces a muted move at the till. The exposure is radically uneven: the same shock that is an irritation in one country is a food-security event in another.

The practical use of this frame is that it tells you what to watch instead of what to react to.

In the currency case, the thing to watch is not the next intervention headline but the rate differential, because that is the variable whose turn would make an intervention stick. An operation announced into a closing gap is a very different object from the same operation announced into a stable one, even though the press release reads identically.

In the climate case, the useful signal is the calendar rather than the price. The forcing is already dated and published; the economic consequence runs one to three quarters behind it. That lag is the entire window in which anything is still cheap to do — adjusting import cover, securing fertiliser, planning around a known-bad season. Once it shows up in a CPI print, every remaining option is expensive.

In both, the same discipline applies: separate the force from the price, and ask which one the action in front of you is aimed at.

This is not a prediction about where any currency trades, nor a claim that any particular intervention will or will not succeed. It is not a forecast of crop yields or food prices — the honest position is that the climate signal is strong and its agricultural translation is uncertain. It is not a criticism of the authorities involved; both are using the instruments they actually have. And it contains no trading view and no investment recommendation of any kind.

It is one structural observation: an authority can act on a price while the force setting that price is untouched, and when that happens the result is always large, immediate and temporary.

So the next time a government announces decisive action — a currency defended, a price capped, a market stabilised — note the move, and then ask the question that actually determines whether it holds:

Is this acting on the force, or on the price the force is setting? And if it is the price: what is supposed to change while the time is being bought, and is anyone actually doing it?

We are not here to tell you whether the next intervention works. We are here to make sure that when something large and immediate happens, you can tell within a minute whether you are watching a bridge or a subsidy.