I take the thrust your point in principle. However the data was collected over all seven years of post JAA operations and not just a single year, and therefore the maximum amount of data available was presumably used. One could argue that if seven years is not enough then how about ten, or fifty or whatever.
In the paper I cited, the comparison was between 2 years (1997/8) before JAR-FCL introduction and 2 years (2002/3) after.
The point is that what is possible from such low numbers is not very informative. One can only conclude from the data that the underlying frequency of "incidents" is somewhere between 20% higher and 20% lower. 20% safer is worth having, isn't it?
I accept that there are few incidents from which to enable large statistical sampling, but what has been done appears to the best possible from such low numbers, which in themselves might be argued are further evidence of unnecessary over regulation
I don't agree. Accidents, and in particular deaths, are sufficiently important and expensive that it's worth putting some effort into avoiding even a low number -- a safety measure designed to save one fatal accident per year is worth quite a lot of cost and effort, even if it takes 30 years for the effect to be significant in statistical terms.
But I've separated my critique of that paper from my thoughts about regulation in general. Whether JAR-FCL's extra measures get anywhere close to achieving that level of improvement at commensurate cost is another matter.