Insurance-sponsored studies, particularly ones presented to a public utilities rates-setting commission, should be digested with a helping of pepper to hide the foul, rotting taste of spoiled meat that goes along with the profiteering inherent in the insurance industry.
The conclusion that "sport bike rates should 89% higher" than other classes demonstrates pretty clearly what they are trying to do - utilize one correlated variable to set rates, and ignore other significant variables. This is an example of contorting statistics to meet the needs of the user, in this case a regulated entity seeking to raise rates. The insurance industry knows that by engaging in behavior like this, they can generate excess income and risk-free profit. They no doubt have, internally a more sophisticated multi-variable regression analysis that delineates exactly what variables are loosely correlated, which are strongly correlated, and which are not correlated and then have selected the ones that allow them to reasonably (to the unpracticed eye) present a case to raise rates the most with the least risk.
First of all, what is the definition of a "sport bike"? Is it defined by power-weight ratio, or average height of the unloaded Cg, or is it the presence of a certain cosmetic features etc.
Beyond that, I would guess (no data) that the single most important variable associated with insurance claims is rider experience versus type of bike (ignoring my definitional problem above). If you have 10 years riding sportbikes, your unlikely to start spontaneously crashing next year - but, if your experience is limited to sportbikes, and you begin riding a cruiser, the answer is likely different (although less different than if you have 0 years riding).
Oversimplification in specification is a classicly effective method to misuse stats.
cdg