Interesting idea floating at Daniel Drezner. Although severance insurance may or may not be a Pareto improvement (that depends on how you spell out the payoff functions), it has the potential, by providing for the gainers from a policy change to compensate the losers, to satisfy either the Kaldor or the Hicks criteria of welfare economics. That's a non-trivial achievement. Some economists use the term "Marshallian improvement" (I stole it from one of Paul Krugman's policy books, don't remember which one, sorry) to refer to a situation in which social welfare increases in the aggregate despite losses to some agents. To visualize this, draw a production possibility curve, pick an allocation of output along it, draw a new production possibility curve to the right of the old one, and pick an allocation of output that is NOT to the northeast of the old one. You've just identified a Marshallian improvement, in the form of expanded output, but one that requires some people to give up some of their current consumption to make the improvement happen. If there are less cumbersome ways of effecting a Kaldor or a Hicks compensation, that's likely to be good.