Those coal ash spills should have been expected.

Normal Accidents is a 25-year-old book by Charles Perrow, subtitled “Living with High-Risk Technologies.” Perrow, reflecting on the Three Mile Island nuclear incident and other accidents, argued that modern advanced technologies are so complex, and require such careful monitoring and management, that accidents, including potentially massive system failures, have to be considered “normal,” not exceptional, events.

The technologies he wrote about included many we consider commonplace today, but climate change and other global environmental impact risks were not among the "accidents" he anticipated.

Economists seem to have learned precious little from the book, highly acclaimed as it was. Economic calculations still get made on the basis of “expected values” — the statistically most likely outcomes — despite the fact that these values do not accommodate the virtual certainty of unexpected events.

Analyses like Environmental Impact Statements — required for major federal investments under the National Environmental Policy Act — are still based on what economists call “expected utility theory” (EUT). Based on past experience and recorded data, we project the probability of different events and use those odds in combination with the “utility” or value associated with each alternative event to arrive at an expected value for a course of action.

That doesn’t make sense … or does it?

Think of it this way. On a given day, it could rain or be sunny. You have to decide whether or not to carry an umbrella. The utility of having an umbrella if it rains is high, but if it’s sunny, the umbrella’s nothing but a pain — a negative utility. Whether you carry the umbrella is a function of what you think the odds are of rain.

Great … except that you don’t know if you can trust the weatherman’s forecasts. And you’ve failed to consider a whole mess of possible outcomes: High wind makes the umbrella useless even in rain; there’s hail or snow; if you bring your umbrella, it is sunny, and you forget it somewhere and lose it.

Not only that, but you’ve also assumed that none of the negatives are too terrible, and none of the positives too spectacularly good. If you were like the Wicked Witch of the West in The Wizard of Oz and melted if you got wet, wouldn’t you carry your umbrella no matter what the odds were?

Now let’s get back to coal ash and normal accidents.

If we buy Perrow’s line of argument, we’d have to incorporate possibilities other than rain or sun in our umbrella calculations. The logic of normal accidents is that something unexpected will happen if the technology or system is complex enough (and enough time has passed). The same way, when the coal-fired power plants were built and the ash storage designed, some allowance for a spill or other accident should have been included in the calculations of the expected value of new power plants.

Assuming that no spill will occur is not reasonable. There must have been some calculation made of the probability of accidents and the damage they might cause. But did anyone plan for or calculate the loss associated with a complete collapse of the retaining wall and a spill of the entire containment, not just a slow leak? I don’t know, but the New York Times coverage of the spill at the Kingston, Tenn., plant noted that the TVA had struggled for years to control small leaks.

They had to have known better, right?

Not quite. We also have excellent evidence on the TVA’s assumption about the likelihood of a massive spill and the costs involved. From the NYT article:

In 2003, it [the TVA] considered switching to dry disposal, but balked at the estimated cost of $25 million, according to a report in The Knoxville News Sentinel. That is less than the cost of cleaning up an ash spill in Pennsylvania in 2005 that was a 10th of the size of the one in Tennessee.

So, that tells us the TVA figured the probability of a spill multiplied by the likely cost to clean up was less than $25 million in current dollars in 2003.

They were badly wrong. Why?

The expected value calculation appears to have assigned a zero probability to a spill as massive as the one that occurred. That’s as if you carried an umbrella in the rain even though you might melt  if you got wet, and assigned a zero probability to the possibility of high winds or hail. That’s not rational.

Assigning that zero likelihood also meant that there was no serious effort to calculate the cost of such a massive failure.

What this helps to see is that the expected value calculation — and all our NEPA-mandated Environmental Impact Statements for federal projects — depend on key assumptions that are unlikely to be true even if we make conscious allowance for normal accidents we tend to ignore:

  • We know the possible outcomes or conditions (the weather states that can occur),
  • we know the probability of each those outcomes will occur within some time frame, and
  • we know the impacts on “utility” of each of those outcomes.

If we don’t know all possible outcomes, or if there is uncertainty about the likelihood of each outcome, or if we don’t fully understand or can’t measure the effects of each outcome, then the expected value calculation is a guess, not an empirically-based forecast.

The weather forecast for tomorrow, and its effects on the ability to do outdoor construction work, doesn’t meet these standards for a full empirical projection. Yet the designers of those power plants claim to be able to give us good forecasts of the stability of the ash storage facilities for the decades that a new plant will be in operation.

Good luck. Any power company planners — or state and local government regulators and oversight agency personnel — out there: Did you ever hear of the Precautionary Principle?