FuelWatch – ACCC Econometrics
Posted by Possum Comitatus on May 29, 2008
When the ACCC releases their econometric modelling of FW, we’ll update and whack it in here.
But as a bit of a warm up we’ll run a basic test to to show that there was price change when FuelWatch was introduced.
The series we’ll be using is Price Difference which equals the difference between the Perth price and the 3 State average (Melbourne, Adelaide and Sydney – again, Brissy is excluded because of the rebate we get up here) price for unleaded petrol. We’ll use the quarterly data from the Australian Automobile Association.
We’ll run a very simple regression. We’ll regress the Price Difference on a constant and a dummy variable that has the value of zero for the period before FuelWatch in WA and a value of 1 after FW was introduced.
Yes, serial correlation I know , see comments.
The constant tells us that before FW in WA, the Perth price was, on average, 1.38 cents per litre more expensive than the 3 State average. It also tells us that after FW, the price difference reduced by an average of 1.57 cents per litre – leading Perth petrol being, on average, 0.2 cents per litre cheaper than the 3 State average price (1.38-1.57). That’s around what our non regression results told us earlier which it should be.
That’s a statistically significant structural break. potentially statistically significant structural break – see comments regarding serial correlation.
This is part of the weekly series the ACCC uses rather than our quarterly data:
What we’ll be looking for in ACCC work is talk of the volatility changes that resulted from FuelWatch in WA since we already know that there was clearly a structural change in the price.
The ACCC has now released a paltry little document over here:
The first thing to note is they did the same simple test that we did earlier, apparently in the previous report, but on weekly and monthly measurements. Their results showed the existence of a structural break in the price at the implementation of FW.
They found reductions in the price of petrol in Perth compared to the eastern capital city average of:
1.92 cents per litre as a weekly average
1.86 cents per litre as a monthly average
0.9 cents per litre as a weekly average based on the lowest price point from week to week.
The important thing was always going to be about the volatility, in terms of whether the price on the cheapest day of the week in Perth increased or decreased relative to what they were pre-FuelWatch
Importantly here, the ACCC measured the week to week price changes from the lowest price point of the week, the highest price point of the week and the remaining five days. They found:
prices decreased an average of 3.5cpl for the highest price day of the week prices decreased an average of 0.7 cpl for the lowest price day of the week prices decreased an average of 1.8 cpl for the remaining middle five days of the week.
So the volatility post FW still enables motorists which buy fuel on the cheapest days to benefit from those cheapest days, but it’s had the effect of slightly compressing the bottom end of the variance of Perth petrol prices.
While Perth petrol consumers that buy on the cheapest day still save money compared to pre-FuelWatch – around 0.7cents per litre, the real area of pricing change is at the top end of the weekly price cycle, with that top reducing by an average of 3.5cpl with the other 5 days being in the middle at 1.8cpl.
The ACCC also performed some more advanced tests on whether FuelWatch induced a structural break in the price of Perth fuel and found that it had.
We can conclude that there is an extremely high probability that FW reduced overall prices in Perth that the biggest price reduction occurred at the top of the weekly cycle (making the most expensive days of the week cheaper), that the smallest price reduction occurred at the bottom of the weekly cycle (making the cheapest day of the week slightly cheaper) and the other 5 days fell in the middle.
And we can also conclude that FuelWatch, based on the WA experience, had zero evidence of increasing prices anywhere, at anytime in the weekly price cycle.
It would have been nice to see the modelling itself – and the data.
Sinclair Davidson crunches numbers on the original ACCC modelling HERE:
It confirms the importance of the ACCC releasing the data and all of their modelling.
At the moment, the more you look at the ACCC work, the more uncertainty it seems to contain. Without all of their work and data, it’s almost impossible to make a definitive independent conclusion.
Competition or Fuelwatch? – that’s the big question when it comes to determining what reduced the difference in petrol prices between Perth and the eastern capitals.
Sinclair reckons it was competition – and it’s hard to argue with, the evidence is compelling.
To show how compelling we’ll do three regressions; the first one again (but this time adjusting for serial correlation in the errors since we have Buckleys and none of getting the actual ACCC data ), then we’ll replace the Fuelwatch dummy variable with a Coles dummy variable (which has a value of 1 in 2004 onwards after Caltex and Coles Express moved into the Perth market, and zero at all other times), and we’ll finally run a regression with both of them.
This final regression is compelling – it would suggest that, using quarterly data, the extra competition that the introduction of Caltex and Coles Express had on lowering the difference in prices between Perth and the eastern capitals was more powerful than Fuelwatch.
So much so that Fuelwatch became statistically insignificant when the increase in competition was controlled for.
Even if we limit the time period to 1992 onwards, nothing much changes.
If FuelWatch really did decrease prices – we’ll need a lot more evidence than the ACCC has thus far provided. At best we could say that FuelWatch combined with higher levels of competition reduced the price difference between Perth and the eastern States, but looking at Sinclair’s data (which uses a higher resolution of data, monthly rather than quarterly) – even that might be a bit of a stretch.