In the algorithm, we need to know the policy functions ~qh;t+1(1;t+1; 2;t+1; st+1) and ~ Fl;t+1
(1;t+1; 2;t+1; st+1) with h = 1; 2; 3, l = 1; 2 and st+1 is the state of Nature. Since we cannot obtain
closed-form expressions for these functions, we take use of B-spline smooth functions with degree 3
to approximate policy functions. Therefore the approximated functions are twice continuously differentiable.
We approximate asset prices ~qh(1; 2; s) with h = 1; 2; 3 and exiting wealth ~ Fl(1; 2; s)
with l = 1; 2 parametrically by functions