Yes, I am totally agreed with Prof. Booth's comment. It depends on you whether you want to use Bayesian ideas or not. Bayesian ideas are very popular in many fields of science, technology and research.
the Bayesian statistics framework is the only reasonable method in the insurance area (and other areas, where decisions are being made that truly matter):
First of all, there is natural integration with the decision making under uncertainty framwork: a Bayesian regression results in a posterior distribution of the parameter of interest. The expected utility of a choice (e.g., provide an insurance for a car at a certain price) can be calculated from the posterior and a cost function. That is not possible with frequentist statistics, where only point and interval estimates are provided.
Prior knowledge can easily be accounted for (e.g., young male drivers in rural regions have a higher risk of accidents) and can continuously be updated.
Modern Bayesian regression engines provide much greater flexibility as ever before, for example, non-linear relationships and exotic random distributions.