Uncertainty is poorly represented in existing studies of climate-chang
e impacts. Methods that have been used to characterize uncertainty in
the literature are described and the limitations of each discussed. It
is found that two broad characterizations are useful. A large number
of studies are based on several specific scenarios or attempt to rando
mize selected variables in their deterministic economic models. Other
studies describe individual or collective reaction to risk. The first
category falls short of an adequate representation of uncertainty by f
ocusing primarily on a few values of variables included to capture var
iability. The second group of studies tend to focus more on behavior t
han impacts. What is needed are Monte Carlo type simulations where ran
domness is apparent in a series of independent draws from a distributi
on suitably adjusted for climate change. Some of the benefits of impro
vements in the characterization of uncertainty are discussed.