An excellent post. Thank you.

Perhaps the next step in this analysis would be to adjust energy usage based on import/export. That is, if a country is a net importer, such as the US, that country should, logically, be assigned the energy used to produce and transport those imports. Likewise, countries that are net exporters should receive an adjustment downward to reflect that the fact that some portion of their energy use is going to goods and services that they are not actually using.

I have seen this done for carbon emissions and, personally, I think it presents a more accurate picture of usage based on actual demand. It doesn't allow the industrial net importers to offload their impact/demand/consumption when they offshore their work force, etc. In short, it generates a consumption-based distribution... "costs" are assigned to the consumer of whatever.

Not sure what the best way to do this would be because I haven't thought about it at all. A simple method would probably be to adjust energy use by the percentage of import/export relative to GDP. If country X is running a 5% trade deficit relative to its GDP, you would increase its energy use by 5%. While not terribly accurate, this would tend towards reasonably scaled adjustment.

A more accurate (and much more complicated) method would be to determine imports/exports on a per-source country basis, and then calculate energy used based upon the relative value of that trade to the exporting country. For example, if the US imports $X of goods from China, determine what % of China's GDP is represented by X, then take that % of China's energy use and shift it to the US. Note that this works both ways. Even though the US is a net importer from China, its own energy use does need to be decreased by the amount of energy used to generate exports to China. This same exercise would be done for every country relative to every trading partner. Obviously, this is much more complex as it requires tracking import/export breakdowns on a per-country per-partner basis. But it would be more accurate. Probably. ;-)

I'm sure their are other better ideas, but those are what came to mind reading the above comments.

Brian