I have been thinking about policies for dealing with the biodiversity consequences of climate change when, as seems realistic, a policy-maker has poor (or non-existent) probability information.
One approach is to use some classical decision theory which eschews use of probability information.
Suppose one is thinking about a potential biodiversity conservation problem that may be associated with climate change and one seeks some anticipatory adaptation response to possibly limit the damage. There are three policy-relevant states of the world:
S1 = when a disaster occurs as a consequence of climate change which is successfully dealt with by policy,
S2 = when no disaster occurs
S3 = when a disaster occurs even though policy actions have been taken.
If the cost of the disaster is L and the cost of dealing with it is C then the minimax payoff matrix can be constructed for the two policy options (i) Take action, (ii) Don’t take action. According to the minimax criterion of trying to eliminate the worst possible eventuality it is best in this situation to take no policy action because, with S3 a possible state of the world, this avoids the chance of incurring cost C and at the same time the loss L. The payoff matrix (where taking action and dealing with the problem successfully is given base utility 0) is illustrated.
Another possibility is to use the minimax regret policy criterion – now one seeks to minimise the regrets future generations will experience through the current generation not taking action. The minimax regret payoff matrix is illustrated above. This gives the more sensible result that one seeks to minimise the regret experienced one will take action to avoid the threat if L-C > C or if L > 2C.
Thus, sensibly, one must weigh up the climate induced losses against the cost of policy. In particular, take action if – as seems sensible – C is small relative to L. This will be the case if the policy choosen is a ‘no regrets’, ‘all weather’ policy that yields some return even if the climate change disaster does not occur or if it does occur when the policy proves useless in offsetting it.
For example, biodiversity conservation policies in the face of climate change that involve tree planting or limiting clearing will reduce salinity problems. This is an ‘all-weather’ policy that might be advanced to improve the resiliance of biodiversity conservation but which also has spin-off benefits to agriculture.
Moreover, one can intuit arguments for policy promptness by thinking about ‘no regrets’ policies in a ‘real options’ framework. Here there are traditionally two ‘irreversibility’ forces that drag a policymaker in opposite directions with respect to the timing and intensity of greenhouse gas adaptation policy. Sunk-cost effects make one want to delay action (and to reduce the intensity of initial actions when taken) because the policy-maker can learn about the future. Other irreversibilities (such as species extinctions) make the policy-maker want to bring actions forward (and increase the initial intensity of actions). The net effect of these opposing forces on the timing an intensity of policy is an indeterminate mess.
But ‘no regrets’ options reduce the sunk cost effects here – they reduce sunk costs because they offer a payoff irrespective of the state of the world that eventuates. Hence they lead to more weight being placed on those irreversibilities that involve a case for prompt action. This overturns the bias others have deduced for waiting and motivates a prompt response.
One extension would be to allow different types of ‘all weather’ policies. Another would be to explicitly model agricultural sector benefits B(C) as a spillover consequence of spending C in addrtessing climate change. Then, I think, the minimax case for not taking action is weakened to B(C) less than C while the minimax regret argument for taking policy action is strengthened to L less than 2C -B.
Further extensions would allow for biodiversity and protection against climate change to be joint stochastic outputs contingent on investment in adaptation. This might suggest something about the desired overall scale of the adaptation intervention. I’ll pursue these arguments in future posts.
Copyright H. Clarke.