Thoughts on Nuclear Power

An objective comparison might indeed suggest that a well-designed and vigorously regulated nuclear power plant poses less danger than, say, a coal-fired plant of comparable size. Such a comparison, however, ignores the fact that the regulation of nuclear power in the U.S. still relies on wand-waving.

Elizabeth Kolbert, The Future of Nuclear Energy Around the World (New Yorker)

As the Fukushima nuclear power plant’s woes fade from public consciousness, I’m reconsidering my stance on nuclear power. Before Fukushima I had pretty much signed on with the “it’s better than coal” point of view along with many “environmentalists”. The general equation went like this:

Coal (cons)

  • Produces greenhouse gases
  • Mining coal sucks (and kills quite a lot of people)
  • Transporting coal sucks (and kills a few people)

Nuclear Power (cons)

  • Nuclear Power plants can melt down and create gigantic clouds of radiation
  • We don’t know what to do with nuclear waste (but we’ll figure it out)
  • Nuclear power plants may lead to proliferation of nuclear weapons (but we’re just talking about countries which have them already, right?)

The first con under nuclear power plants had been largely dealt with by engineering — or so we thought. You can’t, or shouldn’t, argue against nuclear power based on the performance of past power plants since we’re not going to build past power plants. Right? So, assuming our current reactor designs are safe, then nuclear power plants are safe.

Once in a thousand years

A commonly bandied about term in electrical engineering is MTBF, or mean time between failures. Most electronics exhibit a certain pattern of failure, which is that their failure rate is highest when they’re first switched on, and then remains very small but essentially constant for the rest of their lifetime. Often, electronic devices are factory-tested for immediate failure and if they survive that they are expected to be very reliable from that point on, but have a certain (essentially fixed) chance of failing per-unit-time over their lives.

When we talk about floods, earthquakes, and so forth, it tends to be in terms such as “one in a hundred year” or “one in a thousand year” incident. The idea here is that we can somehow gauge the probability of an event exceeding some specific measure in a given year (e.g. the amount by which a river will overflow its banks or the Richter scale value of an earthquake). It’s worth bearing in mind that accurate measurement of such events has, in general, been going on for no more than 150 years, so the idea that we have any historical basis for determining what a “one in a thousand year” event might look like is pretty laughable. How many weeks do you think you’d need to measure “rainfall”, say, in order to be able to gauge what a “one in one thousand week” rainfall value would be? (Ah, but rainfall is seasonal and earthquakes aren’t… Are you sure? Are you quite sure?)

When nuclear experts talk about one in a thousand year events they’re generally talking completely outside their area of expertise. Nuclear engineering is kind of complicated; so is meteorology; so is seismology. Even more disconcertingly, the way phrases like “one in a thousand year event” are bandied about, we’re led to believe that this means that we’re safe for a thousand years, or that the odds of something bad happening this year are less than one in a thousand. Something like that.

Even if the US nuclear industry does make power plants with a one in one thousand chance of catastrophic failure within a given calendar year, there are 104 nuclear power plants in the USA. This gives us a 9.9% chance of having a catastrophic failure each year, and that’s just based on the stuff we expect.

But nuclear power plants aren’t laboratory experiments. They’re factories built by companies to make money in a regulatory environment devised by a government. Engineering estimates of quality generally assume that things are built to spec, and competently maintained and operated. For-profit companies are notorious for cutting corners, and — in the nuclear industry, as elsewhere — are frequently caught doing so.

And human beings can of course do things we can’t predict. In her 2003 piece Indian Point Blank, Kolbert mentions, among other things, that after many nuclear power plants failed to protect themselves against mock terrorist attacks (despite having advance notice) the nuclear lobby responded by demanding easier tests. The same article discusses the fictional quality of nuclear evacuation plans because people outside an evacuation zone will inevitably flee, blocking the path of the people who most need to flee. (The Japanese must have been thrilled by the US telling people in a 50 mile radius of Fukushima to evacuate when the government was saying 20 miles.)

Incidentally, the Union of Concerned Scientists quotes a study by Sandia (part of DARPA, so not tree-huggers) in 1982 that a core meltdown and radiological release from one of Indian Point’s two reactors could cause 50,000 near term deaths, and so on.

So, exactly how does one predict a one in one thousand year risk for corporate negligence, successful lobbying to reduce safeguards, or terrorist attacks?

Nuclear Waste

The most glaring problem for nuclear power is nuclear waste. We have no particularly good way to store it in the short term (and, in fact, the greatest and most probable danger at Fukushima was always posed by the spent fuel rods which aren’t in a sturdy containment chamber but simply stacked in a pool of water). And we haven’t got a good idea what to do with it in the long term. In a recent Diane Riehm show, Ken Belson (the NY Times’ nuclear correspondent) described Yucca Mountain as having been chosen by “the finest geologists in the US Senate”.

As with all of the worst problems posed by nuclear power, the issues involve people.


Perhaps the biggest obstacle to nuclear power is that it simply doesn’t make economic sense. The curious thing about proponents of nuclear power is that they’re happy to argue economics until you mention one thing: you can’t insure a nuclear power plant. If we should look to markets for guidance, the markets have spoken: we cannot find an upper bound on the risk posed by a nuclear power plant, so we can’t insure one. If you want to argue that markets are, or can be, irrational — I’ll happily accept that. But then don’t come back and argue price, because the price of nuclear power assumes a massive — apparently infinite — government subsidy.

To put it in another way — when you pay for coal power you’re paying, in part, for the life insurance or hazard pay of coal miners and the liability insurance of coal-fired power plants. Yes, coal may kill people, but you’re compensating the dead when you pay for the power it gives you. When you pay for nuclear power, you’re hoping nothing bad happens, and if it does you (or your descendants) will pay more tax to cover it. If it can be covered.

I can’t really sum up any more eloquently than Elizabeth Kolbert:

As the disaster in Japan illustrates, so starkly and so tragically, people have a hard time planning for events that they don’t want to imagine happening. But these are precisely the events that must be taken into account in a realistic assessment of risk. We’ve more or less pretended that our nuclear plants are safe, and so far we have got away with it. The Japanese have not.

Elizabeth Kolbert, The Future of Nuclear Energy Around the World (New Yorker)