PBF Energy Refining Margin Tracker

Regional weights: reported 2026Q2 throughput. LA/SF throughput is not separately reported; 50/50 is an assumption you can change. Current LA 50% / SF 50% of West Coast. History uses this fixed latest mix, not historical portfolio weights.

If the chart cannot load, use the recent daily observations below.

The index is a weighted gross price spread before costs, with all regional substitutions disclosed.

Measured tracking error · pbf-benchmark-check-v1

How close are the regional proxies?

2024Q3–2026Q2 · 8 completed quarters. We compare the mean of available, same-day EIA crack observations with the same-quarter benchmark published by PBF. These are benchmark tracking errors, not errors against realised margins or EPS forecasts.

Use East Coast and LA as level context; treat Chicago and SF more cautiously. SF is retained as a diagnostic trend proxy, not a dependable SF margin level. Chicago has a large historical outlier. The headline basket retains all five exposures: deleting a weak region would silently change what it represents. These raw comparisons remain unadjusted. A separate experimental regional adjustment is tested below.

The basket’s low error partly reflects cancellation. Its MAE is 0.83 $/bbl, versus 2.42 $/bbl for the weighted absolute regional errors on those same dates. Positive Chicago errors can offset negative Gulf Coast and SF errors. That cancellation may not persist.

For changes between consecutive quarters, the basket’s mean absolute tracking error is 1.31 $/bbl, versus 5.63 $/bbl if the published benchmark basket were simply assumed unchanged. This is a descriptive flat-change reference, not a forecast competition.

What this says about the next quarter: East Coast and LA have been closer level proxies in this sample. Chicago and SF need a wider allowance for error. This small quarterly sample cannot establish a reliable future error band, and none of these figures validates an earnings forecast. Daily tracking errors can be larger than errors in quarterly means.

StorageCurve model · experimental regional basis adjustment

Adjusted regional margin indicators

Locations stay fixed; their price differences do not. We correct persistent regional bias with a simple additive offset: raw proxy minus the median historical proxy error. A median limits the influence of a single extreme quarter. No fitted slope or extra market inputs are used. The offset combines grade, crude and location substitution effects; it is not an observed transport premium.

Regional adjustments help; the total basket did not improve. Gulf Coast, LA and SF get a separately labelled experimental adjustment. East Coast stays raw because the test benefit was small; Chicago stays raw because the publication-lag check worsened. The headline index, charts, regional contributions and reported company figures remain raw.

Does the correction help on later quarters?

Fit each region’s candidate offset on its first four quarters, freeze it, then test on later quarters. The current offset above is subsequently refitted on all available reviewed quarters; the scores below do not test that refit. The second check uses only prior-quarter report columns published before each test quarter began, with at least two training quarters.

Why the headline stays raw: on the same 4 later quarters, its MAE is 0.76 $/bbl raw versus 1.08 $/bbl with the three regional corrections combined. Better regional estimates can remove helpful cancellation in the aggregate.

Weekly · physical confirmation

Inventories, demand, utilization and exports

Stocks: million barrels and exactly-five-prior-ISO-year seasonal mean. Domestic demand: product supplied, four consecutive weeks, YoY against the same ISO week in the preceding ISO year. Exports: four-week average, US only. PADD utilization describes all regional refiners, not PBF plant utilization. Regional demand and exports are not inferred.

Weekly block collected 2026-11-02T06:17:54+00:00. Product supplied and exports are US-only; no regional allocation. Gasoline stocks include blending components; distillate includes heating oil. Exports of gasoline include blending components from June 2023, so the definition has a historical break.

Company disclosures · quarterly

Realized margin versus the published benchmark

Report ledger through 2026Q2, reviewed 2026-10-25. New filings require source review and import; this is separate from automatic price updates.

Reported gross refining margin excludes special items and is before operating expense. Benchmark values come from PBF reports, not the daily EIA proxies.

Compliance and crude slate context

EPA RIN prices and reported crude discounts

EPA observations through 2026-09-21 · source app updated 2026-10-15T11:00:35.499Z · collected 2026-11-02T04:22:12+00:00. Weekly transaction observations are published with a lag; they are not current dealer quotes or PBF costs.

EPA weekly separated-RIN transaction prices, weighted by traded RIN volume across QAP types within each transfer week, RIN vintage and D-code. Reuses EPA SUM(Price_INTERMEDIATE_PRICE)/SUM(Price_TOTAL_RINS). Different vintages are never spliced. Weekly transfer-date label is retained; publication may be monthly. Not a live quote or PBF expense.

Unavailable as live exact inputs: Chicago CBOB and ULSD; San Francisco CARBOB, CARB diesel and jet; Los Angeles jet; LLS spot; ANS spot; WCS spot; Maya spot; Licensed Dated Brent assessment; NYH RBOB spot. Quarterly reported differentials are not carried forward into daily crude prices.

StorageCurve earnings model · frozen October 26, 2026

Q3 2026: the central estimate is $10.8 per share

The central case points to about $1.31 billion of underlying, fully-converted earnings, or $10.8 a share. The cautious and strong cases run from $6.5 to $13.3. These are our model scenarios, not company guidance, analyst consensus or a statistical confidence interval.

The quarter’s prices are in. What remains uncertain is how much of that market margin reached PBF’s results. The central case carries forward Q2’s regional margin capture and operating costs, with Q3 throughput at the midpoint of management’s published range. West Coast matters: Q2 followed the Martinez restart, while the Q3 plan included a late-quarter hydrocracker turnaround.

A dated nowcast of an already completed quarter. Price observations cover July–September only. Company results used for calibration stop at Q2 2026. Q3 results were scheduled for October 29. This October 26 estimate is stored separately from the daily tracker and will not silently refit to the Q3 result. No actual Q3 result is included.

The assumptions that move the result

Capture means reported gross refining margin excluding special items divided by the matching published benchmark. The central case uses Q2’s ratio for each region. That is a carry-forward assumption, not a proven relationship. The downside cuts capture by eight percentage points, or twelve for West Coast; the strong case adds five points. Volume uses management’s low, midpoint and high guidance respectively. The price inputs stay fixed across all three cases.

Across the central case’s 85.56 million barrels, a $1/bbl change in realised GRM moves model EPS by about $0.52 after tax. A five-percentage-point change in capture across all regions moves it by about $1.71. That is why benchmark tracking accuracy alone cannot establish earnings accuracy.

Sources: PBF Q2 results and Q3 volume guidance · EIA spot prices · September financing. Forecast method pbf-q3-nowcast-v1, fixed as of October 26, 2026.

Sources and reproducible method

Market payload collected 2026-11-02T06:17:54+00:00. Source: U.S. Energy Information Administration API v2; EPA EMTS public RIN dashboard; PBF Energy earnings releases. Retrieval timestamps do not replace observation or release dates.

• EIA daily spot prices and source definitions
• EIA Weekly Petroleum Status Report
• EPA RIN trades and price methodology
• PBF January 2026 regional benchmark definitions, SEC exhibit
• PBF throughput and reported benchmark source

Sum of five proxy cracks × reported regional throughput weights. West Coast is split 50/50 between LA and SF as an explicit adjustable assumption, not reported plant throughput. Historical daily curves are reweighted to this fixed latest reviewed mix; this is not a point-in-time earnings backtest. Missing any active price leg makes the index unavailable; weights are never renormalized around missing prices. All five regional benchmarks are proxies, even though individual EIA prices are observed.

Generic cracks use gasoline:diesel ratios 1:1, 3:1 or 2:1, divided by 2, 4 or 3 input barrels. PBF regional proxies use the separate jet blends listed on their cards. Positive does not mean net profitability. All component units, series identifiers, observations, weights, reported results and source links are available in the normalized data.