Luxembourg · What moved my fuel price?
Petrol and diesel prices in Luxembourg
What a litre costs at the pump, what it costs before tax, and what moved between two weeks. Prices are published weekly; a week the publisher skipped is left as a gap.
What moved the price across this window
The change in the pump price, split into the parts that make it up. Each bar is the movement of one component between the first and last week we hold in the window.
Between 2021-09-27 and 2026-09-21 (260 weeks apart). These are the weeks we actually hold, which may differ from the dates requested.
| Component | Change (EUR/L) | Share of pump move | Type | What the value is |
|---|---|---|---|---|
| VAT | +0.06931 | 14.5% | Calculated | VAT extracted from the observed pump price at the published rate, differenced between the two endpoint weeks. |
| Excise duty | +0.01007 | 2.1% | Calculated | Excise per litre from the tax table, differenced between the two endpoint weeks. |
| Other specific taxes | 0.00000 | 0.0% | Calculated | Other specific per-litre levies from the tax table, differenced between the two endpoint weeks. |
| Pre-tax price | +0.39762 | 83.4% | Calculated | Published pre-tax price, differenced between the two endpoint weeks. Broken down further below. |
| Pump price change | +0.47700 | 100.0% | Calculated | Reported by the API as the total move. It is not the sum of the bars above; where the two differ, the difference is drawn as its own bar. |
| Component | Change (EUR/L) | Share of pump move | Type | What the value is |
|---|---|---|---|---|
| Input-cost proxy | +0.20929 | 43.9% | Calculated | Brent converted to euro per litre at the reference rate of each quote day, differenced between the two endpoint weeks. Refinery yield is not modelled. |
| UNMODELLED | +0.18833 | 39.5% | Calculated | The part of the pre-tax move the proxy does not account for. It is differenced like any other component and is never distributed across them. Nothing here says what is inside it. |
| Pre-tax price change | +0.39762 | 83.4% | Calculated | Reported by the API as the total move. It is not the sum of the bars above; where the two differ, the difference is drawn as its own bar. |
- Pump identity check
- 0.00000 EUR/L
- Pre-tax identity check
- 0.00000 EUR/L
Pump change minus the four components above. Returned by the API so it can be checked rather than trusted.
Pre-tax change minus the proxy and the residual.
Plain differences between the two endpoint weeks. The residual is reported as it moved and is not distributed across the other components; nothing here says what caused any of it.
Shares are arithmetic on two stored differences. They can exceed 100% or go negative when components move against each other, and they are not an attribution of cause.
Petrol (Euro-super 95) - with tax and without tax
Both lines are published figures. The gap between them is the tax; it is broken down below.
- Pump price, with tax(official)
- Pump price, without tax(official)
- Brent (EUR per litre)(calculated)
The pre-tax price is the part that moves with crude, refining and transport costs. Excise duty is a fixed amount per litre, so it makes up a larger share when fuel is cheap.
Brent is plotted in euro per litre on the same axis as the pump prices, never on a second scale: two y-axes would invent a relationship by choosing where they line up.
What the latest week's price is made of
The most recent week in this window, split into tax and pre-tax, and then into the part the input cost explains and the part it does not.
- Pre-tax priceOfficial
- 1.0540 EUR/L57.0% of the pump price
- Excise dutyOfficial
- 0.5264 EUR/L28.5% of the pump price
- Other specific taxesOfficial
- 0.0000 EUR/L0.0% of the pump price
- VATOfficial
- 0.2687 EUR/L14.5% of the pump price
Published by the source as the price without taxes.
Excise per litre, from the tax table row in force for this week.
Other specific per-litre levies, from the tax table row in force for this week.
Extracted from the observed pump price at the published rate.
Inside the pre-tax price1.0540 EUR/L = input-cost proxy + unmodelled
- Input-cost proxyCalculated
- 0.6331 EUR/L34.2% of the pump price
- UNMODELLEDCalculated
- 0.4208 EUR/L22.8% of the pump price
Brent (EIA) USD/bbl, ECB EUR/USD reference rate of the same day, divided by 158.987294928 L/bbl (definitional, no density). Mean of 2 quote day(s) in 2026-09-21 to 2026-09-22. Refinery yield factor 1.0 and NOT modelled.
The part of the pre-tax price the proxy does not account for. Nothing here says what is inside it.
How the input-cost proxy was built: Brent (EIA) USD/bbl, ECB EUR/USD reference rate of the same day, divided by 158.987294928 L/bbl (definitional, no density). Mean of 2 quote day(s) in 2026-09-21 to 2026-09-22. Refinery yield factor 1.0 and NOT modelled.
How long Brent takes to reach the pump
The delay at which changes in Brent line up best with changes in this country's pre-tax price, measured over the whole sample we hold.
Historically, changes in Brent have been reflected in Luxembourg's pre-tax petrol (euro-super 95) price with a median lag of 1 week (90% CI 1 week, r = 0.61, n = 1,084 weeks).
How this was measuredCalculated
Provenance
CalculatedOfficial- Dataset id
- Weekly Oil Bulletin - retail fuel prices with and without taxesWOBec_weekly_oil_bulletin
- Period
- not recorded
- Retrieved
- 2026-09-26 06:54 UTC
- Type
- Calculated
- Transformation
- lag_analysis-1.0
- Licence
- CC BY 4.0 (Creative Commons Attribution 4.0 International), implemented by Commission Decision 2011/833/EU
- A lag and a correlation, not a causal estimate (docs/PRICE_MODEL.md §3).
- bootstrap_seed is fixed, so a published interval reproduces exactly.
- Read max_correlation before lag_weeks_median: a precise lag on a weak correlation is still a weak result.
- PRICE_MODEL.md v1.0 §3. Two weekly series aligned on the retail publication week, both first-differenced (dx_t = x_t - x_{t-1}); cross-correlation r(k) = corr(d_to_t, d_from_(t-k)) for k = 0..8 weeks; k_hat = argmax_k |r(k)|. CI by stationary block bootstrap over the jointly-differenced series with geometric block lengths of mean n^(1/3) and 2000 resamples, seeded at 20260922; the reported lag is the median k_hat across resamples and the CI is the 5th/95th percentiles. Minimum sample 104 weekly observations, below which status = INSUFFICIENT_DATA and every value is NULL. A lag is never assumed, never carried over from another country and never carried over from another period.
- docs/PRICE_MODEL.md §3 step 1: Brent in EUR/L averaged over the price week. Taken from analytics.price_decomposition.input_cost_proxy_eur_per_litre, which tax_split builds as the mean of the daily quotes in the bulletin week, each converted at the ECB EUR/USD rate of its own day (assumptions A1/A5/A6). No as-of rule is used.
- Both series come from analytics.price_decomposition (same row, so the weeks cannot drift).
This says the two series moved together with this much delay over this sample. It does not say that one caused the other. Correlation at each lag from 0 weeks: 0.102, 0.609, 0.200, 0.009, 0.075, -0.012, 0.027, 0.102, 0.023. The bootstrap seed is fixed (20260922), so this interval reproduces exactly.
How the pump price follows crude: rises vs falls
How much of a 1 c/L move in the crude input cost had reached the pre-tax pump price after a given number of weeks, estimated separately for moves up and moves down, over two samples.
Pass-through is measured separately for weeks when the crude input cost rose and weeks when it fell, over two samples. Both are shown: the recent window and the whole history we hold are different samples and can give different answers. The figures describe how the petrol (euro-super 95) price and the input cost moved together. They say nothing about why.
Latest three-year window
2023-09-25 to 2026-09-14
Pass-through from the crude input cost to Luxembourg's pre-tax petrol price cannot be estimated (window 2023-09-25 to 2026-09-14): the pre-tax price and the crude input cost are not cointegrated over this window, so no long-run relationship between them is reported and no short-run model is fitted on top of one; no fallback model is substituted.
Cointegration test statistic -1.785 against a critical value of -3.376 (the statistic has to be below the critical value to clear it), on 156 weekly observations.
Full sample
2005-01-03 to 2026-09-14
- After a rise in the input cost
- After a fall in the input cost
- Shaded band: the 90% interval published with each estimate.
| Move in the input cost | Pre-tax price | Interval |
|---|---|---|
| 1 c/L rise | 0.90 c/L | 90% CI 0.71 to 1.09 c/L |
| 1 c/L fall | 0.92 c/L | 90% CI 0.73 to 1.10 c/L |
p = 0.931 against alpha 0.05, does not survive false-discovery correction (q = 0.931 against 0.10).
What the statement engine wrote
Pass-through from the crude input cost to Luxembourg's pre-tax petrol price cannot be estimated (window 2023-09-25 to 2026-09-14): the pre-tax price and the crude input cost are not cointegrated over this window, so no long-run relationship between them is reported and no short-run model is fitted on top of one; no fallback model is substituted.
Latest three-year windowUnknown
In Luxembourg, a 1 c/L rise in the crude input cost was followed within 2 weeks by a 0.90 c/L rise in the pre-tax petrol price (90% CI 0.71 to 1.09 c/L); a 1 c/L fall by a 0.92 c/L fall (90% CI 0.73 to 1.10 c/L). The difference is not statistically distinguishable at the published 5% per-country threshold (p = 0.931): no asymmetry is detected at that level in this sample, which is not the same as finding the response symmetric. Estimated over 2005-01-03 to 2026-09-14, 1084 weekly observations.
Full sampleCalculated
Each sentence is rendered from a fixed template against the stored estimate. None of it is written by a language model, and none of it states a cause.
The estimate says how the two series moved together over each window. It is not a causal estimate, and a difference between the two directions is reported only where the stored test found one. Where a window has no estimate, the reason the check stopped is printed and nothing is substituted for the missing coefficients.
Actual vs crude-only
What the pre-tax price actually did across this window, next to what it would have done had it followed only the crude input cost at the speed and strength the stored pass-through estimate measures. The shaded difference is not explained by crude.
- Actual pre-tax change(official prices)
- What crude alone would have done(calculated, with range)
- Shaded: not explained by crude(not attributed to any cause)
Changes are measured from the bulletin week of 2021-10-04. Crude response taken from the stored pass-through estimate over the full sample (2005-01-03 to 2026-09-14). The light band is a conservative range, not a 90% interval. The shaded area is not explained by crude: refining margins, freight, retail margins and model error are all inside it, and none is named as its cause.
- Over the 259 weeks after the bulletin week of 2021-10-04, crude costs alone would have moved Luxembourg's pre-tax petrol price by +18.9 c/L (range -70.7–+108.6); it actually moved +37.5 c/L, leaving +18.6 c/L not explained by crude.
The expected path uses only the measured short-run response to crude. It is a description of how the two series usually move together, fitted on a sample that may include this window - not a forecast and not a causal estimate. Where no pass-through estimate exists the path is UNKNOWN and nothing is substituted.
US gasoline crack spread (proxy)
A US product spot price minus Brent. Shown because European product benchmarks are not licensed for redistribution; it is a proxy for the Atlantic-basin product market and not a European price or margin.
US proxy — not a European price
Tax rates in force
The excise, carbon and VAT rates the publisher recorded for this fuel, with the dates they apply from.
| From | To | Excise (EUR/L) | Carbon (EUR/L) | Other (EUR/L) | VAT | Type |
|---|---|---|---|---|---|---|
| 2026-07-01 | open | 0.5264 | not published | not published | not published | Official |
| 2026-01-01 | 2026-06-30 | 0.5691 | not published | not published | not published | Official |
| 2025-01-01 | 2025-12-31 | 0.5591 | not published | not published | not published | Official |
| 2024-01-01 | open | 0.5486 | not published | not published | 17.0% | Official |
| 2023-01-01 | 2023-12-31 | 0.5380 | not published | not published | 16.0% | Official |
| 2022-09-01 | 2022-12-31 | 0.5272 | not published | not published | not published | Official |
| 2022-04-13 | 2022-08-31 | 0.4631 | not published | not published | not published | Official |
| 2022-01-01 | 2022-04-12 | 0.5272 | not published | not published | not published | Official |
| 2021-01-01 | 2021-12-31 | 0.5163 | not published | not published | not published | Official |
| 2019-05-01 | 2020-12-31 | 0.4721 | not published | not published | not published | Official |
| 2015-01-01 | 2022-12-31 | not published | not published | not published | 17.0% | Official |
| 2008-09-25 | open | not published | not published | 0.0000 | not published | Official |
| 2007-01-01 | 2019-04-30 | 0.4621 | not published | not published | not published | Official |
| 2005-01-01 | 2014-12-31 | not published | not published | not published | 15.0% | Official |
| 2004-01-01 | 2006-12-31 | not published | not published | not published | not published | Official |
| 2002-01-01 | 2003-12-31 | not published | not published | not published | not published | Official |
| 1999-03-12 | 2001-12-31 | not published | not published | not published | not published | Official |
| 1997-01-01 | 1999-03-11 | not published | not published | not published | not published | Official |
| 1994-10-10 | 1996-12-31 | not published | not published | not published | not published | Official |
| 1994-07-01 | 1994-10-09 | not published | not published | not published | not published | Official |
| 1993-07-19 | 1994-06-30 | not published | not published | not published | not published | Official |
| 1993-01-01 | 1993-07-18 | not published | not published | not published | not published | Official |
| 1992-01-20 | 1992-12-31 | not published | not published | not published | not published | Official |
| 1992-01-01 | 1992-01-19 | not published | not published | not published | not published | Official |
What this page does not show
- No cause is attributed to any movement. The page shows what moved and in what order, not why.
- The residual is the part the input-cost proxy does not account for. Refinery yield, wholesale margin, freight and inventory are inside it and are not modelled separately, which is why it is labelled UNMODELLED rather than named.
- Pass-through above describes how the pre-tax price and the input cost moved together over each window. It is not a causal estimate, and a window with no estimate is a window where the test did not clear - not a finding that the relationship is absent.