Does anyone have a history (i.e., stats, graph) on the accuracy of DoE vs. API 'actuals'?

It looks like the Bloomberg analyst survey is a rough average of the two.

I have this data going back to the start of 2004. The standard deviation of the DOE vs Bloomberg estimate is 2819k over that period (222 weeks). The standard deviation of the API vs Bloomberg estimate is 3949k.

So the estimate is significantly closer to the DOE number over the long run.

Interesting the DOE number is closer to the estimate than even the average of the DOE and the API. If you take the average number each week, then the standard error is 3012k.

Hope this helps.