An optimal regulatory regime is explored in which regulating a non-deg
radable pollution stock, e.g., the accumulation of greenhouse gases (G
HGs) in the atmosphere, would be based on a model of optimal statistic
al decisions where it is shown when it pays to 'act and learn' and whe
n to 'learn and act'. The value of information in reducing uncertainty
can be shown to be sensitive to accuracy and likelihood of scientific
research results. Some interesting policy results are obtained for th
e dynamic intertemporal decision situation when the value of new infor
mation is an outcome of stochastic optimization with learning.