Slide 10 of 11
Notes:
- Using the results of this research, EIA has been able to create a model that takes observed changes in spot prices over the previous weeks, and forecasts what this week’s retail price change will be. As you can see from this chart, we’ve been fairly successful.
- This chart shows that the model is quite accurate at forecasting one week ahead. In fact, in the first 28 weeks of this year, our model correctly forecasted the direction of the retail price change 26 times, for an accuracy rate of 93 percent . Additionally, as you can see, most weeks it did a very good job of forecasting the relative magnitude of the increase or decrease in retail prices. We’re still refining this model, and customizing it for each region, so we’ve got quite a bit of work left to do.
- So what does this show us? To the extent that we can show a consistent relationship between wholesale and retail price movements, and even duplicate that behavior via a model, it would appear that no significant anti-competitive behavior is occurring in the marketing chain, at least beyond the spot price level, which is essentially the “refinery gate” or point of import.
- Of course, national average prices do not show what may be happening at a local level. But as we further develop this model, we will be able to state more definitively whether the relationship between spot and retail gasoline prices is as consistent on a regional basis. We also hope to be better able to understand what happens in a price spike, which is the most difficult phenomenon to explain statistically.
