Ciclo di seminari dipartimentali a.a. 2018/2019

Model ambiguity in decision making under uncertainty

Interviene: Georg Pflug (University of Vienna)

ABSTRACT

Abstract Many decision models in finance, energy, insurance, business 
etc. contain models of uncertainty. In some cases, these models are 
found ad hoc ("assume that this quantitity has a normal distibution with 
mean m and standard deviation s"). In other cases, they are based on 
same statistical analysis of available data. Hovever it is often ignored 
that the estmates come with some estimation (model) error. The 
incomplete information about the correct model is called "ambiguity".
We review some recent results about how the (nonparametric) model error 
can be incorporated into the decision process. To this end, we introduce 
the ambiguity set as the set of models compatible with our observations 
and extend the basic minimzation problem to a minimax (saddle point) 
problem.
We show examples from portfolio optimization (ambiguity of the whole 
model or only of the compula), multistage hydro reservoir management, 
management of a thermal power plant and last but not least from pricing 
of insurance contracts and electricity futures.
In all these cases, the price contains an additional component, which is 
based on the degree of model uncertainty (the ambiguity premium), which 
comes on top of the usual risk premium.