Optimal stopping
With backward iteration one should formulate the variables
Xt$max(Ht,E(Xt+ 1 j Ft)),
whereE(Xt+ 1 j Ft)is the conditional expectation ofXt+ 1 given
information available at timet.
It is not di¢ cult to show that
Xt=sup
τt
E(H(τ)j Ft)
That isXtis the best what one can get starting at timet.Xtis the same
as the value function in DP.
DeÖnition
Xtis called the Snell envelope.