30 April 2010

Spill, part 2

Let's consider this oil spill situation systematically, with an eye towards information and statistics.

What did we know before the spill happened? We knew that oil rigs fail with some probability p>0.

What have we learned as a result of this spill? We may have learned that p is higher than previously thought.

What actions should we take? If p was too low we should adjust it upwards, and then re-evaluation any decisions we previously made on the basis of the value of p.

Note that depending on what prior estimates of p were it is also possible that p is still a good fit for our expected failure rates, and we shouldn't take any action in light of the spill.

(An example: You expect a lightbulb to burn out in your house every 1000 hours. A lightbulb burns out. Do you need to adjust your expectations by lowering the expected time between failures? Maybe, but not if it's been more than 1000 hours since the last burn out. And not if it's been less than 1000 hours, but not enough less to lower the mean time between failures under 1000. It's very possible that this one burn out, by itself, doesn't tell you anything new about your true lightbulb failure rate distribution.)

Let's assume that we adjust pmarginally upwards, and therefore also adjust marginally upwards the expected cost of drilling. If you already thought the benefits of drilling in American waters didn't outweigh the cost, the new information does nothing for you. If you already thought that the benefits of drilling far outweigh the costs, the new information does nothing for you. The only way the new information is relevant to you is if you thought the benefits just barely outweighed the costs, because maybe now they don't anymore.

Of course then you have to ask all sorts of other questions. Will new drilling platforms be safer than old platforms? If we don't drill here will somebody else drill somewhere else? What is the safety of platforms elsewhere? Do we care as much about spills elsewhere as we do here? What is the elasticity of supply and demand for the sort of oil produced by these wells? And on and on.

My point isn't that this is evidence for more drilling, merely that it isn't much in the way of evidence either way. One data points like this doesn't tell us much, whether we're talking about industrial disasters, terrorism, crime, medicine, whatever. There aren't any broad, new conclusions that we can reach after an incident like this unless we also answer all those other questions.

1 comment:

  1. Good information. Lucky me I found your website by chance (stumbleupon).

    I have bookmarked it for later!

    Feel free to surf to my page how to cheat in miscrits
    - ,