Estimating BP's Total Possible Liability - $66 billion dollars?!?

I've been curious about just how much BP will have to end up paying for the Deepwater Horizon Oil Spill.  Let's come up with a ball park guess: worst case estimates by a Purdue engineering team put the spill at 1.8 million gallons of oil per day.  At 50 days, that's 90 million gallons of oil spilled.  For comparison, according to the Wikipedia article, the Exxon Valdez 10.8 million gallons and they ended up paying $3,000 million total in cleanup and settlement of various fines and charges.  At a strict cost-per-gallon rate, that's $3,600 per gallon.  Factoring in a total inflation rate of 70% since 1990, that would mean BP could be on the hook for upward of $43 billion.  Note that I really didn't do anything fancy here beyond some back of the envelope calculations, and note just how close I am to Credit Suisse's estimate of $37 billion on June 2nd.  Perhaps they used the same methods I did?  In fact, when you take into account the fact that I did my computations 8 days later, our numbers are almost eerily similar.  Maybe being a high paid analyst isn't as hard as they make it seem?

However, I should point out that my estimates are fairly conservative, and further only take into account oil spilled up until now.  However, my estimates are also fairly dismal in the sense that the cost per gallon most likely does not scale linearly - I hate to say it, but there are almost large fixed costs in an oil spill, and certain economies of scale become possible.  Further, Exxon had a drunk driving their tanker; right now as far as we know, Hyundai (the people who made the oil rig that exploded after less than a decade in service) are actually the ones responsible.  Finally, nothing says that the spill rates aren't actually lower - CNN was reporting tonight that the Government's current best estimates are 10-19 thousand barrels per day, or roughly 0.6 million gallons.  As you can see, using that non-worst-case estimate cuts our estimate by nearly a third.  Of course, the spill may very well go on for another 2 months - say if the first relief oil well fails and we have to wait for the second - which would double my estimates.  As with all back of the envelope calculations, this one is a rough stab at orders of magnitude; unknown unknown random factors are just as likely to cut it in half as they are to double it.  Still, given my hope that Credit Suisse arrived at a number through more rigorous means (otherwise what are they getting paid for, anyway?) the fact that my alternative guesstimate confirms theirs is reassuring - chances are after all the unknown stochastic elements play out, this is probably an order of magnitude upper bound for the amount that BP will ultimately have to pay.

Another guesstimate at an upper bound would go like this: so far BP at day 50 has had to pay $1,250 million, and assuming that it takes another 2 months or 60 to contain the spill and that costs scale linearly with time, we'd be looking at $2,750 million spent to contain the spill.  Meanwhile, to get an estimate of the cleanup costs, this 1997 paper Estimating Cleanup Costs For Oil Spills says that the cost per tonne spilled in the U.S. is $73,156. Using a conversion of 7.14 barrels of oil per metric ton and 42 gallons in a barrel gives us a cost of $244 in cleanup costs per gallon spilled.  At 100 days and 1.8 million gallons per day, that's $44 billion dollars in estimated cleanup costs.  Holy frack!  Yet again, that's nearly the same amount I estimated earlier.  Of course, keep in mind that this number doesn't include civil and criminal liability costs like the previous estimate did.  In the Exxon Valdez case, Exxon had to pay roughly half of their cleanup costs to settle the issue in court, so using that rough metric would say that BP's legal liability could be another $22 billion, for a grand total cost to BP of $66 billion.  Sounds rather ominous, doesn't it?

We've tried going higher - what's a rough lower bound?  I remember reading somewhere that BP put its estimates of the total cost at $4 billion - chances are fairly good this is the lowest "reasonable" estimate we're going to find.  Meanwhile, most analysts seem to be guessing somewhere between $10-25 billion, which seems like a good middle of the road worst-case scenario cost - its satisfyingly right in the middle of our lower and upper bounds (at least logarithmically, which is how you should compare orders of magnitude).  Still, notice that I said "worst-case scenario" - no where in here have I taken the numbers into account in BP's favor, and in fact I have been quite conservative against them.  Its my sincerest hope that the reality will actually favor BP in this situation, because anything that lowers their costs here would be coming directly from a lessened environmental impact of the spill itself.

So there we have it - a worst case lower bound of about $4 billion, an upper bound of $44-66 billion, and a most likely worst case scenario cost of about $10-25 billion which we get by logarithmically scaling between the two.  Maybe I'll whip out some log normal distribution to try to justify this further, but for now my intuition satisfies me.