Are cuts in diesel and petrol prices passed on at the pump more slowly than rises? Our analysis of nearly 10,000 filling stations in France, covering 18 months of data, shows a real but limited effect. In a typical station, 90% of a wholesale price rise is passed on within 0 to 3 working days, compared with 3 to 6 days for a cut. However, after three weeks, cuts are passed on at least as fully as rises: our data show no “confiscated” price cuts.
Where does the expression “rockets and feathers” come from?
The expression rockets and feathers was popularised by the economist Robert Bacon in 1991, based on a study of the British petrol market. The image is simple: pump prices supposedly shoot up like a rocket when oil prices rise, and drift down slowly like a feather when they fall. In 1997, the economists Severin Borenstein, A. Colin Cameron and Richard Gilbert formalised a method to measure this phenomenon on the US market. This is the method we applied to French data.
The suspicion resurfaces every time oil prices fall. In July 2026, Michel-Édouard Leclerc, head of the E.Leclerc retail chain, summed up the situation to justify a sale at cost price: wholesale prices are easing, but pump prices are falling much more slowly. We reported on it in our article on Leclerc’s cost-price operation (in French).
Our method
For each working day, we compared two price series from 20 March 2025 to 25 September 2026:
- the wholesale price in Rotterdam, in euros per litre excluding taxes, published by UFIP Énergies et Mobilités, the French oil industry association;
- the pump price, including all taxes, station by station, taken from the prices filling stations are required to declare.
For each station, we measured the share of a change in the Rotterdam price that reaches the pump, day after day, over three weeks, separately for rises and for cuts. Full pass-through corresponds to about €1.20 at the pump for each €1 change in Rotterdam, because of the 20% VAT. The excise duty, fixed per litre, does not vary with the wholesale price.
The analysis covers 9,735 stations for diesel, 7,533 for SP95-E10 (petrol with up to 10% ethanol) and 3,085 for SP95 (E5 petrol). The results below are medians by station type: half of the stations do better, the other half worse.
Rises arrive in 1 day, cuts in a week
The following table shows the delay, in working days, after which 90% of the final pass-through is reached:
| Fuel | Station type | Stations | Rise: 90% by | Cut: 90% by |
|---|---|---|---|---|
| Diesel | Supermarkets | 4,744 | D+1 | D+5 |
| Diesel | Oil companies | 3,612 | D+0 | D+5 |
| Diesel | Independents | 1,295 | D+1 | D+5 |
| SP95-E10 | Supermarkets | 3,861 | D+1 | D+3 |
| SP95-E10 | Oil companies | 2,731 | D+0 | D+5 |
| SP95-E10 | Independents | 885 | D+1 | D+5 |
| SP95 | Supermarkets | 1,931 | D+2 | D+4 |
| SP95 | Oil companies | 612 | D+2 | D+6 |
| SP95 | Independents | 527 | D+3 | D+6 |
The result is remarkably consistent: in every case, the cut arrives later than the rise, with a lag of 2 to 5 working days. For diesel, whatever the type of station, a rise is almost fully passed on the next day, whereas it takes about a week for a cut to be 90% passed on. For diesel and SP95-E10, oil company stations are the fastest to pass on rises (the same day, as a median), without being slower than the others on diesel cuts.
This lag of a few days has a cost for motorists, but it remains limited. It bears no comparison with the image of a price cut taking weeks to reach the pump.
After three weeks, cuts are fully passed on
The second question is one of magnitude: is a cut eventually passed on in full? For diesel, the answer is yes. Three weeks after a change, the median pass-through reaches:
| Diesel | Rise | Cut | Cut / rise ratio |
|---|---|---|---|
| Supermarkets | 1.14 | 1.22 | 1.07 |
| Oil companies | 0.85 | 1.03 | 1.20 |
| Independents | 1.08 | 1.16 | 1.09 |
In supermarket stations, a 1-cent cut in Rotterdam thus translates, three weeks later, into a 1.22-cent cut at the pump, which is full pass-through including VAT. For all station types and all three fuels, the median ratio between cuts and rises is above 1: cuts are not passed on less than rises, only slightly later.
One exception appears for petrol at oil company stations: pass-through after three weeks remains partial, at 0.46 for SP95-E10 rises and 0.59 for SP95 rises, instead of 1.2. Cuts are also only partly passed on (0.85 and 0.51). This result coincides with the price cap applied in 2026 by a major network, which mechanically limits the pass-through of rises. Our calculations do not, however, allow us to isolate this effect.
At national level, a gap hard to distinguish from noise
Measured on the national average price rather than station by station, the asymmetry is not statistically significant on a daily basis. For diesel, the gap between the pass-through of rises and cuts stays within the margin of uncertainty at all horizons, both over the whole period and since the start of the crisis on 28 February 2026.
The weekly data published by UFIP, which distinguish the price excluding all taxes from the price including taxes, provide two insights:
- the only statistically significant asymmetry goes in the opposite direction to the “rockets and feathers” effect: two weeks after a change, diesel cuts are passed on more than rises;
- the results are identical, in proportion, for the price excluding taxes and the price including taxes: the speed of pass-through is determined by the pre-tax price, which includes the cost of the fuel and the retailers’ margin. Taxes only amplify the changes, through VAT.
To understand how the price of a litre breaks down between product, distribution and taxes, see our article on the breakdown of fuel prices (in French).
What this means for you
When oil prices fall, the cut does reach the pump, but a few days late. Comparing stations during this period is therefore particularly useful: not all of them pass on the cut at the same pace. As price changes are concentrated in the morning, between 6 and 10 a.m., it is also better to compare prices in the late morning, as shown by our analysis of the times when prices change (in French).
Limitations of the analysis
These results are based on a statistical model with several limitations:
- delays are expressed in working days, and station results as medians, which do not describe each individual station;
- the responses estimated station by station are irregular from one day to the next: only overall trends are interpreted here;
- SP95-E10 is compared with the SP95 price in Rotterdam, as it has no quotation of its own;
- the period studied, from March 2025 to September 2026, was marked by a crisis of exceptional scale, with sharp rises followed by sharp falls;
- margins of uncertainty are calculated using a standard method, which may underestimate them.
Sources
- Do Gasoline Prices Respond Asymmetrically to Crude Oil Price Changes? – Borenstein, Cameron and Gilbert, The Quarterly Journal of Economics, 1997
- Petroleum product values and quotations – UFIP Énergies et Mobilités
- Fuel prices in France – public data – French Ministry of the Economy