Stochastic optimization using the stochastic Bellman equation, and probabilities approximated by way of Monte-Carlo, is a fairly unexplored field -- especially when the probabilities are affected by the controls. In this thesis, we establish a framework for such optimization problems, and solve problems of optimal reinsurance. The results are then investigated in order to reveal potential weaknesses and/or strengths of the methodology.
We found that the methodology produces unreliable results. We conclude that further research and improvement of the methodology is essential if it is to be used on real problems.