How depleted oil inventories could prolong inflation risks

A scenario, not a prediction

In their November 2026 arXiv preprint, Fueling Inflation Risks: An Oil Shortage Scenario, Christiane Baumeister, Todd E. Clark and Florian Huber investigate how a prolonged disruption to shipments through the Strait of Hormuz could affect oil prices and US inflation.

Their central question is conditional: if petroleum inventories fall to critically low levels and stay there, how severe could the consequences become? The study does not estimate the probability of that inventory collapse occurring. Nor do the authors present its results as an ordinary forecast of what will happen. It is a hypothetical stress analysis intended to inform central-bank risk management. The supplied source identifies it as a preprint and provides no journal publication reference.

Making the shortage explicit

The researchers prescribe a path for petroleum inventories rather than letting the model determine it. Stocks decline from their February 2026 level to a historical minimum by June, remain there for a year, and then partially recover over six months. An operational minimum means the lowest stock level compatible with functioning pipelines, storage tanks and export terminals—not literally no oil remaining.

The scenario assumes that disrupted shipping, constrained production and slow replenishment keep supplies tight even after passage through the strait resumes. These are inputs to the analysis, not independently established outcomes of the model.

Although the narrative concerns global shortages, the model’s inventory measure uses petroleum stocks in Organisation for Economic Co-operation and Development countries, with historical extensions described in the paper.

From inventories to inflation

The analysis has three stages. First, the assumed inventory path enters an oil-market model covering prices, fuel consumption, economic conditions, industrial commodity prices, refining margins and drilling activity.

The model combines conventional relationships with Bayesian additive regression trees, or BART. These trees divide historical observations through sequences of questions about economic conditions. Combining many trees allows relationships to change with circumstances instead of imposing a single fixed response. The oil model also allows uncertainty to vary over time.

Second, each simulated oil price is translated into a US gasoline price using a historical rule of thumb: a one-dollar increase per barrel corresponds to about 2.5 cents more per gallon. Local starting prices produce different projected price levels across locations.

Third, separate nonlinear Phillips-curve models estimate inflation in energy, food and other consumer spending. Here, a Phillips curve relates inflation to economic conditions and expectations, alongside the scenario’s oil-price changes. The components are combined into headline personal consumption expenditures, or PCE, inflation. Core PCE excludes food and energy.

Instead of producing only one path, the framework generates distributions of possible outcomes. This lets the authors calculate probabilities of crossing price or inflation thresholds under the specified scenario.

What the simulations show

The simulations suggest that inventory releases initially restrain prices, but sustained depletion is followed by substantial increases. The median inflation-adjusted Brent price reaches about $166 per barrel in mid-2027, expressed in March 2026 dollars. By the end of 2027, the conditional probability of exceeding $200 reaches 28%.

For national gasoline prices, the model gives a better-than-even chance of exceeding $5 per gallon about a year after the scenario begins. Those gasoline projections are also expressed in March 2026 dollars. Higher-price locations, particularly on the West Coast, face greater risks of crossing specified thresholds.

Median headline PCE inflation peaks near 4.8% in December 2026, while core inflation peaks around 4%. At the February 2028 horizon, the model assigns an 86% probability to headline inflation remaining above the Federal Reserve’s 2% objective and a 97% probability for core inflation.

Projected average household expectations for year-ahead inflation rise from 6.2% to roughly 9% in early 2027. These are modeled survey expectations, not measured future inflation. Responses differ across demographic groups; lower-income households start with higher expectations but show a smaller additional increase.

Reading the limits

All these probabilities depend on the imposed shortage path and estimated historical relationships. They do not establish how likely the disruption itself is. The fixed gasoline conversion also carries existing geographical price differences forward rather than modeling every local adjustment.

The analysis largely uses information available in February 2026, although a footnote identifies one later input: the oil price observed on the March emergency-release announcement day. Historical forecasting tests support aspects of the oil model, but do not validate this particular severe scenario.

The contribution is therefore a conditional risk assessment: it shows how depleted inventories could change both projected prices and the range of damaging outcomes, without establishing that those outcomes will occur.