Basically they showed they could predict or reproduce U.S. oil production levels from the 1930s through 1990s, using just 3 factors: the price of oil; the cost of oil production; and the allowed limits on oil production enforced by the Texas Railroad Commission, a de facto cartel operating up through the early 1970s. The resulting model predicts oil production far more accurately than does the Hubbert bell curve.

This is nice work as far as it goes, but at some level it is just statistics, as they don't go deeper to discuss what this means in terms of oil production in other areas and times. And further, this new paper is not necessarily inconsistent with Hubble's basic insight about geology eventually limiting production levels.

Consider for example the Texas RRC limit, which was steadily raised until it finally hit 100% in the early 70s. Yes, you can say this explains why oil production was rising during this time. Statistically the correlation is there - the allowable production levels were increasing, and sure enough, production was increasing too. But from the larger perspective, if we ask why the Commision was raising production limits, that was due to growing demand and supply factors. The Commision did not operate in a vacuum, it made its decisions based on economic and geologic forces. Well, that's implicitly how the Hubbert curve works too (at least how I interpret it) - the upswing is due to economic factors, investment and development and growth; and the peak and downturn is due to geologic limitations. So the operation of the Commission can be seen as part of what makes the Hubbert curve work.

Another similar effect is their use of cost of oil production as a component of the model. Obviously as we get towards the down side of the Hubbert curve, oil production is much more difficult and costly. So yes, if you include oil costs in your model, you again can explain why oil production declines after a while. This is not inconsistent with the Hubbert curve, but rather it is another way of looking at the same effect.

The point is then that just because you can get a good prediction from these three factors, that's not necessarily inconsistent with the basic soundness of the Hubbert curve as a somewhat crude model. Everyone knows that the Hubbert curve is not perfect, but the fact remains that it does pretty well considering what a simple model it is.