On Wednesday evening, a magnitude-5.2 earthquake shook the Spanish town of Lorca, causing widespread damage. It had not been predicted – at least, not quite.
The Italian Raffaele Bendandi (1891-1979) made his name making predictions, apparently based on little verifiable science, about when and where earthquakes would strike. He predicted that Rome would be reduced to rubble by an earthquake on Wednesday 11th May, 2011. There was no earthquake in Italy, but Bendandi was only a few hundred miles out.
Predicting rare and catastrophic events is a notoriously difficult task. Scientific seismological predictions can usually not be made with any reasonable degree of accuracy more than a few hours in advance. So what should we make of Bendandi’s prediction, and what can statistics tell us about the problem?
This is not the first earthquake to have been ‘correctly’ predicted by Bendandi, but neither is there any clear record of how many earthquakes he predicted, or how he reached his predictions. Perhaps, a little like Nostradamus, he made thousands of predictions, most of them incorrect and swiftly forgotten. In this light, an occasional correct prediction seems a lot less impressive than the same prediction made in isolation. We need to know how many incorrect predictions Bendandi made before we can cast him as soothsayer or charlatan.
As predicting certain natural phenomena is so difficult, modern predictions are often made as statements of probability – shades of grey, rather than black or white. Probabilities are sometimes given in weather forecasting, predicting the chance of rain in the next 24 hours, for instance.
Probabilistic forecasts provide an altogether more realistic summary of the chance of rare events occurring. They are both easily understood and easily tested, at least in the long run. But they don’t make such good headlines as cut-throat certainties.