not. If the broker determines the loss to be ex post insurable, the insurer and policyholder negotiate for
a settlement. If the insurer fails to make a reasonable offer, the broker inflicts reputation penalty on the
insurer. With such a breakdown, the broker no longer feels that its clients are getting good service from
this insurer and it diverts business to rivals or extracts concessions from the misbehaving insurer in the form
of lower future prices on the business it places. If the broker determines the loss to be ex post uninsurable,
no claim is made. The broker could also inflict a reputation penalty on the policyholder for trying to push
an ex post uninsurable loss onto the insurer. For example, if the broker feels that the loss is due to fraud,
or is unreasonably inflated (i.e., uninsurable), the broker could inflict a reputation cost on the policyholder.
The broker now is less willing to represent such policyholders because they may hurt its contingent fees from
the insurer or compromise its reputation with the insurer and therefore its ability to place higher quality
policyholders.
2 Related Literature
Our paper is related to the literature on costly state verification and falsification in the way that observability
and/or verifiability of states of the world is restricted which impacts the types of “contracts” that can be
written.8 The literature on costly state verification (Townsend, 1979, Bond and Crocker, 1997) considers
situations in which (some) states are only observable to the policyholder but that the insurer can verify
these state ex-post at a cost. After a claim is filed, the insurer now has the choice to audit the claim. A
“contract” is then represented by a premium, a verification set which specifies those claims to be monitored,
and a transfer from the insurer contingent on the policyholder’s claim and the auditing outcome. The
literature on costly state falsification (Crocker and Morgan, 1998) considers situations in which states are
also only observable to the policyholder but they cannot be verified at any cost. It is, however, costly to the
policyholder to falsify claims. Ex-ante, the policyholder and insurer sign a “contract” specifying a premium
and an ex-post transfer from the insurer contingent on the policyholder’s claim. In all these models, full
8Our paper also relates to the literature on implicit risk sharing (e.g. Kimball, 1988, and Kocherlakota, 1996). These papers
are applicable to village communities (e.g. farmers, Kimball) or to two households (Kocherlakota) where individuals are in close
contact and information about defect becomes easily known (e.g. through a blacklist). It is assumed that no future trade will
take place once a party defects and this thread is assumed to be credible. Our paper is different in two ways. First, we assume
that an institutional structure, the insurance m arket, is in place to organize risk sharing. We then address the question how risk
sharing can be extended to include non contractible events? In contrast to the papers above, we thus look at the coexistence
of explicit and implicit risk sharing which leads to interesting interactions. Second, we show how risk sharing can be improved
through the broker serving as a coordination mechanism. This allows for scenarios in which information does not flow easily,
i.e. outside village communities. Interestingly, not only can the broker improve implicit but also explicit risk sharing.