The Price of Rigidity — When a Small Shortage Makes a Huge Price

The most instructive number in the world economy this autumn is not an interest rate. It is the difference between the price of a barrel of crude oil and the price of the diesel refined from it — the crack spread — which spent this year doing something it had never done in its history, passing one hundred dollars a barrel and then going further.

The instinct is to read a price like that as a measure of how severe the shortage is. It is not. The volume of supply actually lost is modest set against the size of the oil market. What the price is measuring is something else entirely: the number of places in the system where an adjustment could have happened and didn’t. Refining capacity that takes years to build. Inventories already drawn down. A demand side with no substitute fuel. When every channel of adjustment is blocked at once, the price is the only variable left with any freedom to move, so it moves all the way.

That is a general principle, and the second half of this briefing applies it somewhere that looks completely different. Brazil has just voted, and will vote again on the twenty-fifth, in an election framed as a choice between economic models. The framing is real but it answers the wrong question. What determines whether any programme becomes an outcome is not which one wins. It is the same thing as in diesel: how much, and how fast, the system can actually adjust.

Start with what the market did, because the facts are unusually clean.

The US diesel crack spread reached an all-time high of around $102 a barrel in mid-August — the first time it had ever been above one hundred. It did not stop there: it closed at roughly $108 in early September, and the New York Harbor ultra-low sulphur contract reached about $110 the following day. For context, the broader 3-2-1 refining margin averaged its highest monthly value since 2006 in August. At the pump, the US average diesel price hit an all-time high of $6.53 a gallon in late September.

Two things about that chart matter more than the level itself.

The first is that this is a crack spread, not an oil price. It is the margin between crude and the refined product, which means the story is not “oil got expensive.” Crude is not the binding constraint here. The constraint is the plants that turn crude into diesel, and the distinction is the whole briefing.

The second is the scale mismatch. Estimates of the refined-product exports lost from the Middle East and Russia run to something like four million barrels a day, with a much larger share of global refining capacity offline in total. Against a crude market of roughly a hundred million barrels a day, a loss of that order is meaningful but not catastrophic. The price response has been catastrophic. That gap between a moderate quantity shock and an extreme price response is the thing to explain.

A shortage only becomes a price explosion when every available adjustment channel is blocked simultaneously. Normally at least one gives. This year, none did.

Refining capacity could not respond. US refiners were running at around ninety-eight per cent utilisation by late August — effectively flat out, with no idle plant to bring on. And the direction of travel has been the other way for years: a series of US closures and conversions has removed over a million barrels a day of crude processing capacity, more than six per cent of the operable total. New capacity is not a decision, it is a construction project measured in years.

Inventories could not cushion it. US distillate stocks fell to around 108 million barrels in mid-September, the lowest level for that point in the year since 1982. A buffer exists to be spent in exactly this situation; this one was already low going into the peak agricultural and heating season.

The product mix could not be redirected. The facilities knocked out by conflict and outage were disproportionately geared toward diesel and jet fuel rather than gasoline. A refinery has some flexibility in its output slate, but not unlimited — you cannot simply decide to make distillate instead.

And demand could not fall. This is the one that does most of the work. Diesel is the fuel of freight, farming, construction and heavy industry, and in the short run those users have no alternative. A haulier cannot switch fuel for the quarter; a farmer cannot delay the harvest until the price improves. Demand that cannot respond to price is demand that transmits the entire shortage into the price instead.

With supply fixed, inventories empty, the output slate constrained and demand unable to retreat, there is exactly one variable left in the system that is free to adjust. That variable is the price, and it did all of the adjusting on its own.

There is a reason this matters beyond energy traders, and it is a specific one rather than a general warning.

Diesel is an input to almost everything physical. It is not a consumer good that households can substitute away from — it sits underneath food, freight and construction, which means its cost propagates into prices that have nothing obviously to do with fuel. That makes it one of the few commodities whose move is genuinely an inflation event rather than a line item in one.

It is also precisely the kind of shock monetary policy handles worst. The cause is a missing refinery, and the instrument available is one that works by suppressing demand. Raising rates does not produce diesel. It can only reduce the amount of economic activity that wants to consume it, which is the same thing as accepting a slower economy as the price of a lower print. That is the uncomfortable trade sitting underneath this autumn’s data, and it is worth naming before it arrives dressed as a surprise.

The honest counterweight is that this works in both directions. A price driven by an acute supply loss unwinds when the supply returns, and the base effects then flatter the following year’s figures as hard as they hurt this one. The question is entirely how long the outage lasts — and that is a question about repair schedules and shipping routes, not about policy.

Which brings us to an election, and to why the frame above travels.

Brazil voted on 4 October. Neither candidate cleared fifty per cent — Flávio Bolsonaro took roughly 56 million votes against Lula’s 53.7 million with virtually all of the count in — so the presidency goes to a runoff on 25 October. Two of the moderate first-round candidates have said they will not endorse either side.

The coverage will frame this as a choice between economic models: more state direction, gradual fiscal consolidation, or deeper structural reform. That framing is accurate and it is also where most analysis stops, which is a mistake. The useful question is not which programme wins. It is which parts of any programme can actually be executed, and on what timetable.

Because the constraints are structural in exactly the sense the diesel market is. A Brazilian president does not command a legislature; fiscal rules and budget changes require congressional majorities that the presidency does not confer, and the composition of that congress is a separate result from the headline one. Privatisation is a legal process before it is an economic one, with timelines that routinely outlast the term that began them. And the macro variables everybody actually cares about — the debt path, the investment rate, the cost of capital — respond last of all, if they respond at all.

So the gap between a result on 25 October and a measurable change in Brazilian economic data is not months. On the structural items it is years, frequently spanning more than one term. The election determines direction. It barely touches speed, and speed is what decides whether a direction turns into an outcome.

Several things cut against the clean version.

The diesel squeeze is not purely structural, and it would be wrong to present it that way. A substantial part of the lost capacity is offline because of conflict and attack rather than because of a long-run investment shortfall, and that part can come back faster than a construction timetable implies. A ceasefire, a reopened shipping route, or a set of repairs completing would unwind a good deal of this in months rather than years. The structural story — closures, no spare capacity, thin buffers — is what made the system fragile; it is not by itself what produced the spike.

Elasticity is also a function of time horizon, not a fixed property. Diesel demand is close to unresponsive over a quarter. Over three or four years it is not: fleets get replaced, routes get reorganised, electrification proceeds at the margin, and genuinely high prices call forth capacity that did not previously pencil. “Inelastic” is a statement about the short run that gets quietly treated as permanent.

On Brazil, the implementation-constraint argument can be overdone in the opposite direction. Governments with narrow mandates have delivered consequential reforms, markets price expectations well before any legislation passes, and currency and rate moves can arrive on the night of a result rather than years later. The claim here is not that politics does not matter. It is narrower: that the distance between a mandate and a measurable change in the real economy is routinely underestimated, and that the second-order question — can this be passed, and how fast — carries more information than the first.

This is not a forecast of diesel prices, refining margins, or how long any outage lasts. It is not a prediction of the Brazilian runoff, nor an endorsement of any candidate or programme — the structural point applies identically whoever wins. It is not a claim that monetary policy is wrong to respond to a supply shock; it is an observation about what the instrument can and cannot reach. And it contains no trading view and no investment recommendation of any kind.

It is one structural observation: a price tells you how rigid a system is, not how large the shortage is, and the same question — what can actually adjust, and how fast — is the one worth asking of a government as well as of a market.

So the next time a price does something unprecedented, resist the first reading — that the underlying shortage must be unprecedented too — and ask the structural questions instead:

How much supply actually went missing, and how does that compare to the size of the move? And of all the places this could have been absorbed — spare capacity, inventories, substitution, demand falling away — how many were available, and why was none of them?

We are not here to tell you where diesel trades, or who governs Brazil in January. We are here to make sure that when a number does something it has never done, you look past the number to the thing it is actually measuring: not how bad it is, but how little in the system was free to move.