Brylek, GREAT project. Thanks for taking the initiative on this and for all your data gathering efforts. blue_myriddn, fantastic visualization. Thanks to you as well.
About the "early years"...I know we are talking an order of magnitude when comparing 91 production volume vs. each of the last few years so perhaps it isn't practical for this project to look at the early years in the same way as you have done for the last four. But we could still come up with some very reliable estimates regarding number of vehicles on the road with no accidents vs. those with accidents vs. totaled, even though a state-by-state break-down might not be feasible. What about surveying a random sample of those years' VINs.
Earlier in this thread someone posted the idea of applying current ratios to those years and it was pointed out that this would be inaccurate as older cars had been exposed on the road for many more years and miles, which is a valid point. But if we sampled, say, 100 cars from '91 it might give us a very accurate estimate of the fate of the total production run. Of course, "100" is just a number I picked out of the air. Someone here who still remembers Statistics-101 from their college days can probably say how large the sample size would need to be in order to represent the total population with a confidence level of 98%, or 95%, or whatever we decide is "good enough".
PS: Please PM me if there is an opportunity to work on or help fund this project going forward, regardless of how you proceed.