I see no reason that insurance history could provide insight into the cause - be it fossil fuels, freon,
History is simply a record of observations; a dataset.
By itself, it doesn't provide much of anything. It's data.
Combined with our understanding of physics and chemistry, however, and the story changes significantly. We are able to chart what we know about the materials against what we see in the data, and tease out very relevant data.
Measured data over a period of time + physical and chemical knowledge = simulation. Whether it's a flight simulator, racing game, crash simulation, or fluid dynamics - the principles are the same. Over time, the simulations become more and more complex, and more and more accurate.
The process is along the lines of:
- Take what we know, and create a computer model that accurately models previously obtained data (ie. matches known data as closely as possible)
- Next, get a new slice of reality - such as crash two real cars together and film and instrument it to collect the desired data
- We input the same initial conditions into the simulation, and run the simulation
- We compare the results between the two, and improve the simulation model
It's a simple feedback system that improves over time.
High quality simulations are not simple, but they are based on simple building blocks, just like all human knowledge. Over time, the models become very accurate (and peer reviewed, often by a competing company whose interest is in disproving your model to their gain). Eventually, the simulation becomes close enough to reality that we base our decisions on the simulation, and tool up for production using simulated data. Verifying the simulation's accuracy is often little more than a formality with an already expected outcome. (And if the outcome is different, then it's an opportunity to improve the simulation model - and profit from that knowledge).
Modern simulations have reached the point where nearly everything that happens on a human scale (be it vehicle design, structures, radio transmission, or even diaper packaging) not only can be simulated with nearly perfect accuracy, but is routine to the point of being almost boring.
This was not always so. Only a couple of decades ago, simulations were crude affairs with very approximate results. Yet these crude simulations were more than sufficient to get us to the moon and back, as well as build the most powerful heavy lift rockets ever made.
While the order of complexity for simulating the climate is many, many orders of magnitude higher than what is required to simulate the structural and aerodynamic performance of the Saturn V or N1 rockets, our ability to perform such simulations has also increased many, many orders of magnitude.
A great deal of the academic papers with respect to climate science are about finding problems (and solutions) in the simulations. While it may sound like that means the model isn't any good, the reality is the discussion has reached the point of minutiae that increase the overall accuracy, but don't actually change the overall result (or prediction) significantly.