oil spills

Natural gas fracking well in Louisiana

The Deepwater Horizon oil spill clearly demonstrated that really bad accidents -- the ones that the industry (and too often the government as well) say never happen -- do actually happen sometimes.  Not merely a figment of some pointy-headed actuary, these low probability but very large damage events really do need to be built into government policy. 

I had hoped that connections would be drawn between the oil spill and nuclear accidents, as there are number of important parallels.  Like oil spills, nuclear accidents are infrequent but can cause devastating damage if they do occur.  In both markets, Congress has set caps on private sector liability by statute.  The result has been to shift liability risks away from the private operators onto surrounding residents or taxpayers, effectively subsidizing oil and nuclear power.  My hope had been that the scale of damages from the BP spill would highlight the inadequacies of nuclear liability insurance, forcing much needed strengthening of both.

Enter energy investors Howard Newman and Craig Jarchow.  In an article in Rollcall, they make exactly the opposite points:  that the nuclear liability system works so well it should be adopted almost whole cloth to deal with oil spills.  In their own words:  "For a solution, we need look no further than the nuclear-power industry's Price-Anderson Nuclear Industries Idemnity Act."

They are wrong.

While they share a similar end goal to mine ("the cost of insuring oil spill liability should be paid by the oil industry, not the taxpayer"), Newman and Jarchow overstate the coverage that Price-Anderson provides while understating what is actually needed to properly internalize operational risk in both energy markets. 

Some examples:

  • Primary insurance cover is too low.  Newman and Jarchow highlight the $300 million in insurance that the reactors must purchase to cover damages to people and property off-site.  While this may look rich compared to the $75 million limit now in place on economic damages from an oil spill, there is little cause for celebration.  There have been 50 years of actuarial data on the nuclear industry to guide underwriters, massive growth in the population and real assets at risk surrounding nuclear plants, ever higher valuations on damages to human health and loss of life, and tremendous innovation in methods to syndicate risk in the marketplace.  Yet, once you adjust for inflation the mandated coverage level for primary insurance today is actually lower than when the Price-Anderson Act first became law in the 1950s.   Arguments that insurers simply can't write more coverage seem unlikely:  plant owners routinely purchase more than ten times as much coverage to protect their own assets (onsite damage to plant and equipment and business interruption) as what they buy to protect all of us outside their gates.  Rather more likely is that they simply don't want to pay for more.
  • Aggregate coverage amounts and adequacy have been overstated.  Scale of the insurance pool is the main benefit of Price-Anderson according to Newman and Jarchow.  "[T]he entire nuclear industry is on the hook for more than $10 billion of third-party liability if an individual facility's coverage is exceeded," they note, touting this as "enormous insurance capacity without bankrupting the industry."  This is certaintly better than the situation in many other parts of the world.  But that doesn't mean it is good.  Their conclusions suffer three main problems:  
    • Value isn't really $10 billionThese "retrospective premiums" are paid in over seven years, not immediately.  This reduces the aggregate cover to roughly $7.6 billion on a present value basis.  Even that is not guaranteed.  Despite letters of credit that reactors will make good on their retrospective payment commitments, the market distress to all reactors that will follow any substantial nuclear accident will create significant financial stress on owners.  Bankruptcy, government dispensation to slow or eliminate retrospective premiums, concentrated ownership of large numbers of reactors by single firms (with each required to make separate payments each year) all suggest total take will be lower than expected.  Available insurance for non-payment of these premiums is quite limited.

      The authors note as a feature the ability for the government to fund the restropective premiums immediately, subsequently "recouping its costs from the industry over time".  This arrangement will likely provide more industry subsidies (through artificially low interest charges on delayed payments).  More importantly, there seems little novel about such an option.  Even absent any insurance program governments could front the funds and try to recover it over time (such as through litigation).  However, such an approach would not constitute an effective risk internalization strategy in my book.
    • Coverage levels are far below need.  Far from demonstrating the value of P-A, the BP oil spill actually underscores the inadequacy of the nuclear limits.   BP has already pledged to fund a segregated trust fund of $20 billion.  While this is still far short of the estimated damages of any major nuclear accident (estimates in the $100 billion range*), BP's current pledge is nearly three times the aggregate coverage available from reactor owners for all offsite damages in a nuclear power plant accident.  Shortfalls following a nuclear accident are most likely to fall on the taxpayer, or in uncompensated damages to health and property for the surrounding population. 
    • What coverage is available varies widely by type of nuclear facilityPrice-Anderson requirements differ by type of entity.  While Newman and Jarchow focus on reactors, coverage levels are lower or non-existent for other portions of the nuclear fuel chain such as enrichment facilities.

The authors underscore how the incentive alignment of Price-Anderson promotes plant safety.  The theory here is interesting (shared losses for accidents anywhere should promote more self-policing), but I think far from proven.  There have been enough near misses at US reactors since Three Mile Island to indicate that if self-policing is happening, it isn't very robust.  Their promotion of a prospective premium for oil, collected as a fee per barrel, is a good one and perhaps could be adopted in P-A reform.  The approach is not without risks, however.  One needs to be concerned that the industry characterizes the fee as covering the entire risk rather than just part of it, increasing the actuarial shortfall.

What the article doesn't mention at all is simply eliminating the liability cap.  If the upper tier of damages is as improbable as industry says it is, an actuarily-fair premium should be no big deal for either industry.  Congress wouldn't get to dump uncompensated risks on to the surrounding population or taxpayers, and Newman and Jarchow's stated goal of having the cost of liability insurance borne by the industry rather than the taxpayer would really be met.


Bob Herbert (NYT) has a more realistic view on liability in these industries here.

Thanks to Simon Carroll for links to the Rollcall and NYT articles.

*See, for example, Beyea, J., E. Lyman, and F. von Hippel. 2004. "Damages from a major release of 137Cs into the atmosphere of the United States," Science and Global Security 12:125–139.