It's nice to have data that shows pretty significant predictability over a long period of time, but the problem is that when all the data is dumped into a bucket and the averages/medians are determined, they are just that...averages and medians. Just because a certain set of variables for one discrete group of workers matches another doesn't mean that there aren't huge anomalies. For instance, if I took the following variables (age, ethnicity, geographic location, number of dependents, academic achievement, work skill) and was able to show that the displacement characteristics were similar, I'll bet you the minute I then sorted by sex the similarity would erode.
What does all of this mean to an insurance underwriter? I really don't know, but I suspect that s/he would conclude that an underwriting would have to be done on an individual by individual basis. That's pretty cumbersome stuff and could be enough of a hurdle for the insurance companies to back away from even trying to offer severance insurance.
But where there's smoke, there's usually fire, and there is clearly buzz out there that some insurance company is writing severance insurance or at least some form of it. I still haven't found it, and I don't know whether it is for corporations or employees. I'll keep looking.