Wednesday, December 23, 2009

Lack of Consensus on Iranian Oil Production

Iranian oil either:
  • Declined sharply by 500kbd when OPEC cut back last fall, and has remained there since (JODI - Iranian self reports),
  • Barely declined, and in any case quickly recovered to almost the same level (EIA), or
  • Began a steady decline in 2006 long before the OPEC cuts, that more or less has continued since (Oil and Gas Journal)

Welcome to the wonderful world of oil statistics analysis...  Half a million barrels/day - 12% - in discrepancy.


Morgan Downey said...

True Iranian actual production is close to the EIA number.

Notice that the divergence occurs at the end of 2008 when the big recent round OPEC cuts were announced. The JODI numbers submitted by OPEC members will always tend to reflect their OPEC quota regardless of what cheating level they are producing. This will always make self reported JODI numbers for OPEC members unreliable.

As with all OPEC members, Iran will state that they are in full compliance in their JODI disclosure regardless of their actual cheating higher production.

A table of OPEC members quotas is here.

I have no opinion on O&GJ production data as I don't know how they calculate their numbers.

Great blog Stuart. You have a great eye for pattern recognition. Enjoy the warm weather over the holidays.

Stuart Staniford said...

Thanks Morgan - and happy holidays to you also! I've been enjoying your blog too.

When I only had put the EIA and JODI numbers in, my initial reaction was the same as yours - the JODI numbers are self-serving, and the EIA more accurate. However, putting in the O&GJ has made me question that story. Guess at some point I'll have to add in the OPEC and IEA numbers and see how they stack up.

As far as I can tell, none of these series are really derived in a transparent way that allows one to cross-check, or reason about which is more likely to be more true than another. So my default approach has always been to only rely on features of the data which are present across data sources.