Efficient Portfolio Frontier in MSF

I received a few questions about how to calculate an efficient portfolio frontier in MS Solver Foundation. The simplest algorithm is to solve a series of QP’s. (A better way may be a parametric QP analysis). As we don’t have looping constructs in OML, it is not immediately obvious how to do this with the Excel plug-in. With the API’s we could do the looping in C# and the Quadratic Programming Problem itself in OML. Here is an alternative approach: solve a bigger problem that has all the individual QP’s embedded. Conceptually this would look like:

Obj:	x1’Qx1-λ1 μ'x1	+x2’Qx2-λ2 μ'x2	+x3’Qx3-λ3 μ'x3	+x4’Qx4-λ4 μ'x4
S.t.	∑ix1,i=1
		∑ix2,i=1
			∑ix3,i=1
				∑ix4,i=1

Note that x_k is a vector rather than a scalar, and that I ignored the non-negativity conditions x_k≥0. A similar approach can be used for any problem where the individual problems are relatively small. A good example is DEA where we achieve better performance by batching small problems into bigger ones.

The OML formulation of this combined portfolio model can look like:

Model[
   Parameters[Sets,I,K],
   Parameters[Integers,Selected[I]],
   Parameters[Reals,Return[I]],
   Parameters[Reals,Lambda[K]],
   Parameters[Reals,Covar[I,I]],

   Decisions[Reals[0,Infinity],Alloc[K,I]],
   Constraints[
       Foreach[{k,K},FilteredSum[{i,I},Selected[i]>0,Alloc[k,i]]==1]
   ],

   Goals[Minimize["Overall"->
         Sum[{k,K},
              FilteredSum[{i,I},Selected[i]>0,
                   FilteredSum[{j,I},Selected[j]>0,Alloc[k,i]*Covar[i,j]*Alloc[k,j]]]
              - Lambda[k]*FilteredSum[{i,I},Selected[i]>0,Return[i]*Alloc[k,i]]
           ]
   ]]
]

Here the parameter Selected indicates whether an instrument can be part of the portfolio.