How far has the diesel pool moved from the certification fuel? A state-level specification-based atlas for the United States, 2011–2024

Abstract

Compression-ignition engines sold in the United States are certified on a test fuel fixed by 40 CFR 1065.703. The fuel they subsequently burn is whatever the market supplies, and the market has changed: fatty acid methyl ester biodiesel and hydrotreated renewable diesel now account for a growing share of transportation distillate, and that share is distributed across states extremely unevenly.

No federal programme publishes measurements of in-service diesel properties. This paper therefore asks a question that public data can answer: given what each state consumed, what do the governing fuel specifications imply about how far its diesel pool sits from the certification fuel?

We combine Energy Information Administration state-level consumption volumes for 2011–2024 (728 state-years, 51 jurisdictions) with published specification envelopes for the three pool components, and propagate the full envelopes rather than only their midpoints, so every estimate carries an interval that reflects what the specifications actually permit.

The headline results are asymmetric, and the asymmetry is the contribution. For states blending hydrotreated renewable diesel, the direction of divergence is established for any admissible property values: HVO’s density envelope (765–800 kg/m³) lies entirely below the certification fuel’s (838.9–864.6 kg/m³), and its cetane envelope (70–80) entirely above (40–50). California’s 2024 pool is 54.6% non-petroleum — 51% renewable diesel — and sits 34 kg/m³ lighter than the certification fuel, with a sensitivity interval of [−50.8, −17.4] that excludes zero.

For states blending only FAME biodiesel, the direction is not established. FAME’s density and cetane envelopes both overlap the certification fuel’s, so the sign of the deviation depends on where inside its permitted range the actual fuel sits. Across the panel, the density deviation is sign-robust in only 2.7% of state-years and the cetane deviation in 4.7% — essentially the HVO-blending states alone. Midpoint estimates do show biodiesel states heavier, and that may be true; the specifications do not establish it, and this paper does not claim it.

We also document a methodological trap in the source data. EIA changed its distillate accounting at data year 2021; the intuitive blend-share construction double-counts from that year forward while being correct before, inflating the national denominator by about 20,008 thousand barrels a year (1.726%) and manufacturing a trend break that looks like a change in the fuel supply. The pipeline detects the break from the data and aborts if it is not where it is expected.

Every state-year carries its sensitivity interval and two sign-robustness flags, so the distinction between what this dataset establishes and what it merely suggests is carried by the data rather than by a limitations section.