## Friday, November 28, 2008

### More on price gas hysteresis

We redo the hysteresis analysis we conducted on gas prices, now for a longer period of time and using much more data than before:
• Average weekly retail prices without taxes of 95 octane gasoline and gasoil in Spain have been obtained from the European Commision Oil Bulletin (thanks to Wonka for the pointer).
• The US Energy Information Administration provides daily spot prices of Brent oil in dollars per barrel.
• Historical data on euro to dollar exchange rates can be retrieved from FXHistory.

Using this information we have constructed the following figure showing the evolution of Brent prices and gasoline and gasoil Spanish retail prices without taxes, all in c€ per liter, between January 2006 and October 2008.

And this is the scatter plot of Δ(gasoline price before taxes) against Δ(Brent price):

The linear regressions of this plot for the semiplanes Δ(Brent price) ≥ 0 and ≤ 0 are:

ΔBrent ≥ 0 → y = f+(x) = b+ + m+x = 0.1854 + 0.1691x,
ΔBrent ≤ 0 → y = f(x) = b + mx = 0.1273 + 0.3636x.

On one hand, m is more than twice as large as m+, which indicates that gasoline price follows oil price reductions much more strongly than oil price increases when these variations are large; when oil price variations are small, it is the fact that both b+ and b+ are positive that dominates the situation: gasoline prices have a very clear bias towards increasing then. The "consumer unfavorable" region in which f+(x) > −f(−x) is

−1.61 c€/l < ΔBrent < 1.61 c€/l.

As for gasoil, the plot look like this:

with regression lines

ΔBrent ≥ 0 → y = f+(x) = b+ + m+x = 0.2669 + 0.1269x,
ΔBrent ≤ 0 → y = f(x) = b + mx = 0.2391 + 0.4134x.

Again, m is much greater than m+ (~3.25 times), but this is masked for small variations of Brent price by the constants b+ and b+ being greater than zero. The consumer unfavorable region is

−1.77 c€/l < ΔBrent < 1.77 c€/l,

which is very similar to the consumer unfavorable region for gasoline.

So, we have a somewhat mixed scenario: when oil prices do not vary much, there is a tendency for retail gas prices to increase, whereas when oil price variations are high, reductions are tracked much more strongly than increases. We propose two explanations, both highly speculative, for this phenomenon:

• Small variations of oil prices do not reach public attention, which allows retailers to silently increase gas prices; high variations in oil price, however, generally make it to the news, so that retailers are under greater public scrutiny when transferring these variations to the pump.
• So as to soften the public negative reaction to price increments (and associated contractions in consumption), retailers do not reflect high increases in oil price as abruptly as oil price reductions, but opt to distribute the increment over a longer period of time; this would account both for the fact that m+ < m and for the positive bias in small price variations.