Risk versus Reward

Fintan O'Toole, a man for whom I have a lot of time, has an interesting article in today's Irish Times. His opening statement reads "Economic and scientific elites are prey to delusions that the unlikely will not happen. The rest of us pick up the pieces when it does."Fintan then proceeds to dazzle us with some armchair mathematics about million to one chances coming true. The main thrust of his argument is that because rare events don't occur often enough for us to classify them we can't make estimations on the likelihood of them occurring in the future.

He mentions the odds of six specific horses winning at Cheltenham as being 1.5 million to one. Fair enough. But the odds of any six specific horses winning would also be quite high. Not following horse racing I can't comment on the specifics, but lets say there were 10 horses in each race and each horse is exactly identical then each set of six winners is equally likely. In my scenario there are exactly 1 million outcomes so no matter which set of six winners occurred that's a one in a million chance. But one of those outcomes has to occur so why do we draw any special inference on that particular outcome? Because we're human and like to see patterns in chaos.

The other thing to note is that Fintan falls into another common trap by selectively picking his data points. Lets say Cheltenham ran for 5 days, 8 races per day, 10 horses per race, 3 Irish horses per race, all horses equally likely to win. Then the odds of there being a day where there were six Irish winners is given by our good old friend the binomial distribution and works out around 2% per day or about 9% over the course of the festival. All of a sudden that's pretty far away from the million to one odds that were being quoted earlier on.

EDIT: Just looked at my spreadsheet again - the 2% and 9% figures are for 1/3 horses being Irish. If there are 3/10 Irish horses then the figures are 1% and 5.5% respectively but they are still orders of magnitude greater than a million to one.

Later in the article he state the design of the Fukushima reactors "did not take account of the possibility of an earthquake beyond what had ever been experienced before being combined with a tsunami." If that is so, and I have seen nothing elsewhere that backs this statement up, it is a damning indictment of Japanese engineering practices 40 years ago. But I seriously doubt this is actually true when one considers the amount of effort that Japan invests into earthquake and tsunami defense.

The biggest failing of Fintan's article is that he ends up saying we should be subservient to Sods Law and that to challenge the status quo is in some way hubris. This I emphatically object to. The path of progress has always been risky. The Wright Brothers were not the first people to try to fly and many died before them trying. Many have died since in aviation accidents but we don't think twice about jumping on a Ryanair flight to see a football match in a foreign country.

Similarly nuclear power is not without its risks. The specter of a reactor meltdown or a dirty bomb immediately sends shivers up the spine. However, when compared to other sources of energy it has a pretty good track record. Despite the good news story of the Chilean Miners last year, coal mining and pollution from coal burning power plants still kill thousands every year. The Gulf Oil spill caused massive ecological damage and lets not even touch the whole Iraq issue in terms of lives lost.

We need to have a sensible, rational discussion about the issues surrounding nuclear power but articles like this should be called out for what they are - FUD.