How Fukushima Dai-ichi core meltdown changed the probability of nuclear accidents?

How to compute the probability of a nuclear accident by using past observations? How the observation of the last catastrophe at the Fukushima Dai-ichi nuclear power plant changed the expected frequency for such events? Little has been the consensus in answering these questions. While opponents of nuclear power claim that the probability of serious accident is very high, the industry ensures that it is negligible. The aim of our paper is to shed light in this controversy by using historical data and an appropriate statistical model. In order to address these questions, we have compared four Poisson models using three different definitions of nuclear accidents from 1952 to 2012. We found that as the definition of accident takes into account other events, (i.e. not only the major nuclear disasters but also less serious accidents) the independence assumption underlying the three first Poisson models is violated. This result called for a time-varying mean model, such as the Poisson Exponentially Weighted Moving Average (PEWMA). By using this last model we find a greater increase in the risk of a core meltdown accident, owing to the accident that took place in Japan in 2011, than the estimates that we obtained using the classic Poisson models.