management policies. In our model of an ecosystem with an unknown, reversible threshold, com-
monly stated goals for managing multistate ecosystems - maintaining resilience and applying a
precautionary principle in decisionmaking - are completely consistent with, and can be justified
by, economic theory. Thus, we suggest that economic optimization approaches and empirical
scientific approaches for ecosystem management are quite complementary. Thus, the kinds of
policies suggested by economic analysis, including incentive-based management schemes such as
taxes, subsidies, and tradable permit markets, and some command-and-control approaches, should
be considered as potential instruments for scientifically-based ecosystem management in the pres-
ence of thresholds. In particular, it may be effective to take the current regulatory framework and
adjust the level of existing damaging activities based on both expected ecosystem resilience and
uncertainty. This proposal may be viewed as an economically-derived equivalent to the concept of
“bet-hedging” against uncertainty [8, 10]. Such adjustments may be considerably easier to imple-
ment than large-scale stakeholder involvement schemes and would be both flexible and adaptable
to future advances in scientific knowledge about ecosystem dynamics.
5 Conclusions
Many natural systems have the potential to switch between alternative system dynamics. We ana-
lyze a multistate system with two distinct domains, each with its own equation of motion. While
earlier studies of multistate systems rely on numerical simulations, by considering both uncertainty
of threshold location and a random component to the underlying dynamic natural process we are
able to formulate the manager’s decision as a stochastic dynamic programming problem and show
that the value function is differentiable, even at the threshold. We show that utility maximization
leads to a decision rule with precautionary behavior that increases system resilience, if the system
is thought to be close to the threshold. We find that increasing uncertainty (both uncertainty associ-
ated with natural processes and uncertainty of the decisionmaker about threshold location) can lead
to nonmonotonic changes in precautionary actvity. In particular, as the variance in the stochastic
component of the natural system increases, the level of precautionary activity may first increase,
but for large enough variance, precaution will eventually always decrease. Similarly, there is also
a nonmonotonic relationship between the uncertainty of the utility maximizer about the unknown
threshold and precautionary behavior. If the decisionmaker is certain that he/she is right below
the threshold, there is no expected benefit from engaging in precautionary activities. If uncertainty
about threshold location increases, so does the probability that the threshold will be crossed and
19