Page 28 - 36-1
P. 28

Ambiguity Increases and Insurance Deductibles




                            increase if the ambiguity increase is consistent with Definition 2 and
                            preserves the cumulative loss probability at the initial optimal deductible
                              *
                            D .
                              F

               Under this assumption, at D , the expected utility in the loss state covered by the insurer
                                        *
                                         F
               is unaffected by the ambiguity increase. In other words, the net marginal utility benefit of
               lowering the deductible is unchanged. This leads to the same result as the assumption of

                                                          *
               an ambiguity increase affecting the loss below D ; that is, G(x;π ) = G(x;π ) and G(x;¯π ) =
                                                                       _ F
                                                                                            F
                                                          F
                                                                                _ T
                                       *
               G(x;¯π ) for losses above D , which implies a specific ambiguity increase. Similarly, Gollier
                    T
                                       F
                                                                  *
               (2014) considers the ambiguity affecting losses below D  to study the optimal insurance
               contract under ambiguity aversion formulated by the smooth ambiguity aversion model.
                                                                                   19
                    To clarify Definition 3, we present a numerical example in Figure 4.  Consider a
               discrete loss x that takes a value in the set {0,1,2,3,4} and w = 5. The full specifications
               of the loss distributions are provided in Panels A and B in Appendix C. We assume τ =
               0.2, the CARA coefficient γ = 0.5, and α = 0.56, as assumed for the numerical example
               of Proposition 3. Before an ambiguity increase, under the loss distributions G(x;π ) and
                                                                                         _ F
                                                                                    *
               G(x;¯π ) (the dark dashed and solid lines, respectively) with center G(x;π ) (the dark
                     F
               dotted line), we have D  = 2. After an ambiguity increase, the loss distributions become
                                     *
                                     F
               G(x;π ) and G(x;¯π ) (the light dashed and solid lines, respectively) with the same center. In
                   _ T
                               T
               this case, Assumptions 2 and 3 hold. The figure indicates that G(2;π ) = G(2;π ) = 0.9285
                                                                                    _ T
                                                                           _ F
               and G(2;¯π ) = G(2;¯π ) = 0.9315, which satisfies Definition 3. Thus, the ambiguity increase
                        F
                                 T
               is the specific one.
                    Suppose that a specific ambiguity increase occurs. We show how a risk- and
               ambiguity-averse individual respond to this specific ambiguity increase in the following
                                  *
               proposition, where D  denotes the optimal deductible after the ambiguity increase.
                                  T
               Proposition 4:  Suppose that ambiguity preferences can be described by an α-maxmin
                             model. Under Assumptions 1–3, a risk- and ambiguity-averse individual
                             reacts to a specific ambiguity increase as defined in Definition 3 as

                             follows:




                  19  We are grateful for an anonymous reviewer’s suggestion to illustrate this definition with a numerical
                     example.


                                                      20
   23   24   25   26   27   28   29   30   31   32   33