A New Keystone XL Paper is Probably Wrong

I’ve been trying to avoid two things lately: Keystone XL and picking on shaky scientific papers. But a new paper on Keystone XL in Nature Climate Change has been generating a lot of  buzz and requests for comment, so a post on it seems worthwhile.

The paper claims to show that the State Department has underestimated the emissions impact of the pipeline by as much as 83 million tons of carbon dioxide equivalent annually. The authors say that the State Department “did not account for global oil market effects” that would lead to greater world oil consumption – and therefore emissions – as a result of higher Canadian oil sands production. They claim that, by now including those effects, they have produced the correct emissions number.

The authors are on reasonable grounds to argue that State should have been less confident in assuming no impact of higher oil sands production on world oil consumption – an issue that scores of analysts (myself included – see, for example,  this 3+ year old blog post or my 2013 book) have long understood is real. But their estimate of that impact is thoroughly unpersuasive and almost certainly too high. Let me explain why in three pieces.

The authors assume that, climate change aside, the world oil market is a perfect market.

No manipulation, no politics, nothing. Just textbook economics. If you’re skeptical of this, you’re right.

The authors’ basic methodology is straightforward. Assume that Keystone would add to world oil supplies. This would push world oil prices down. The result of that would be to encourage increased oil consumption and deter some now-uneconomic oil production. The system would come to equilibrium at a lower oil price and a higher level of oil consumption. That new price and consumption level can be estimated using observed or calculated elasticities of oil supply and demand.

Here is the central problem: It is hugely controversial to claim that that price changes affect oil production only through altered project economics. That’s how a normal market works. (If, say, I’m selling milk, and the price of milk drops so far that I can’t profitably do my business, I’ll cut back or shut down.) But there are oodles of serious people (including  big environmental groups and analysts that produce the  regulatory impact assessments that justify things like fuel economy rules) who reasonably think that that’s not all that’s going on when it comes to oil. Instead they believe that one or more big players in the world oil market – think Saudi Arabia or OPEC – manipulate their production to maximize revenue or target a particular oil price. What that means in practice is they believe that a production increase in, say, the United States or Canada would be met by production restraint in, say, Saudi Arabia. The net result is to blunt the impact of Keystone XL and its ilk on supply, prices, and consumption. Indeed it is precisely this phenomenon that environmental advocates rely on when they claim that increased U.S. oil production would do almost nothing to benefit U.S. consumers at the pump.

I don’t know how Saudi Arabia and other low cost producers might react to increased Canadian oil production. The State Department analysis implicitly assumes that their reaction would fully offset any Canadian increase; I find that implausible. But the new paper assumes without any serious attempt at an argument that they don’t change anything at all; I find that even more difficult to believe.

In essence, both the State Department and the Nature Climate Change paper probably have the politics wrong. Assuming that each of them otherwise have the economics correct, the State Department number becomes a lower bound on the impact of Keystone XL, and the Nature Climate Change paper gives an upper bound.

Which brings us to the second and third problems with the Nature Climate Change paper.

The authors’ numbers for oil supply elasticity are shaky.

A low elasticity of oil supply means that lower prices caused by a rise in Canadian oil production will do relatively little to prompt other suppliers to cut back on their own production. The lower an oil supply elasticity one uses, the greater the impact of Keystone XL appears to be. The headline number that the Nature Climate Change paper reports (a net addition of 0.6 barrels to world oil production for each extra barrel of Canadian crude) assumes a supply elasticity of 0.13. The paper reports a sensitivity test where the supply elasticity is 0.6; that cuts its estimated impact by more than a factor of two.

So where does the 0.13 figure come from? It’s read off the following chart from Rystad, a consultancy.

The steeper the supply curve, the lower the supply elasticity – and the authors have read their supply elasticity off the almost-steepest part of the curve (at 96.62 million barrels a day, an EIA projection for 2020 oil consumption, to be precise). That might be fine if this particular supply curve were gospel, but it’s not – there’s enormous uncertainty around the cost of future supplies. (Incidentally, if anyone can explain to me why the right hand side of the supply curve is a mile above the estimates shown for cost of each source of supply, I’d really appreciate it.) The authors also have no reason to be confident that, excluding Keystone XL, world oil consumption will be 96.62 million barrels a day in 2020 – yet a quick glance at the curve shows that even if you accept the Rystad supply estimates, the point on the curve at which you estimate supply elasticity can have a very large impact on the result. The basic implication of all this is that a lot of pieces need to line up in order for the paper’s central conclusion to hold up (again, we’re setting aside the political assumptions for now), but there’s no compelling reason given for believing that they actually do. Indeed I’ve left one more out so far.

The authors’ assumptions are in far deeper conflict with the State Department analysis than they acknowledge.

The authors understand that State Department analysis already concluded that only if oil prices were somewhere around $65-$75 would a lack of pipeline capacity tilt the balance against oil sands, since otherwise, rail would be able to pick up the slack. They write:

“The overall GHG emissions impact of Keystone XL is determined… by the extent to which Keystone XL leads to an increase in oil sands production. Here, the State Department concludes that owing to availability of other pipelines... or rail for transporting oil sands crude, the rate of Canadian oil sands extraction would most likely be the same with or without Keystone XL.... The State Department also suggests a case in which the oil sands production could increase by Keystone’s full capacity. If future oil prices are lower than expected, specifically $65–$75 per barrel, ‘higher transportation costs (due to pipeline constraints) could have a substantial impact on oil sands production levels, possibly in excess of the capacity of the proposed Project’.”

The authors then go on to elaborate reasons that oil prices might indeed end up in that range as a way to motivate the possibility that blocking Keystone would actually constrain oil sands production.

But take another look at the supply curve. If you really believe that curve (which the authors say you should), and you’re now looking at a case where oil prices are $65-$75, that means you ought not be looking at the point on the curve where production is 96.62 million barrels a day – instead, you should be looking well to the left, at a part of the supply curve that’s far flatter and hence less supportive of the paper’s conclusions. A quick read of the supply curve suggests an elasticity of about 1 in the $65-$75 price range – well above the number they use, and even outside the range of their sensitivity analysis.

Despite all this, I wouldn’t go so far as to say with 100 percent confidence that the authors’ emissions estimate is too high.

Odds are that the numbers in the Nature Climate Change paper are too high and that those in the State Department report are too low. (That doesn’t necessarily mean that State has overestimated the likely emissions impact -- given that, in the case it considers most likely, Keystone has no impact on Canadian oil production, and hence no impact on emissions. What’s too low is State’s upper bound.) If you forced me to bet, I’d put the real number a lot closer to the State Department one than to the Nature Climate Change result. But there are all sorts of uncertainties involved in analyzing world oil markets beyond the ones I’ve just discussed. There are undoubtedly more uncertainties than the new Nature Climate Change paper acknowledges. One nice thing for policy analysis is that many of those uncertainties (such as in the elasticity of oil supply) affect not only the environmental costs of Keystone XL but also the economic benefits – so that, in a proper cost-benefit analysis, they often cancel out.

One last thought: Cross-over papers that take climate change and fossil fuel markets both seriously are important and too rare. It’s good to see people trying to bridge that gap. (Even the Bob Howarth paper on methane leaks, for all its extraordinary flaws, was commendable for trying to grapple with both worlds). But you’re rarely going to find two peer reviewers who can credibly tell an editor whether such a paper is fit for publication – the range of expertise required is almost inevitably too large. Either journal editors need to solicit a larger numbers of reviews (covering a wide range of expertise) than typical for such papers, or journalists need to lay off treating them as much more than preliminary ideas until they’ve withstood sustained public scrutiny. I doubt the latter will happen, but one can hope that journal editors take new steps to get these sorts of papers right.