This paper reconsiders the largely overlooked Thompson insurance mecha
nism for public choice. The mechanism's apparent defects can be remedi
ed by a multi-part mechanism that generates necessary information and
corrects potential errors. The added parts are (a) a sample of the pop
ulation of eligible voters, who prepare a program for all voters' appr
oval, and (b) the Vickrey-Clarke-Groves pivot mechanism. One can use c
lassical statistical methods to draw valid inferences from a populatio
n sample about the entire electorate, preserving incentive compatibili
ty and permitting efficient outcomes.