High impact, low probability

1
High impact, low probability?

• A brief empirical analysis of risk in the
  economics of climate change
• Dr Simon Dietz
• Grantham Research Institute on Climate Change and
  the Environment
• London School of Economics

2
This paper

• The question
• To what extent do the results of CBA of climate
  change depend on low-probability, high-impact
  events?
• The approach
• Intuitive empirical analysis with an integrated
  assessment model (PAGE)
• The answer
• Results strongly depend on tail risks, but for a
  set of plausible assumptions discounting still
  matters
• The intuition
• Tail risk of (arbitrarily) huge losses in utility
• But (by definition) not very likely and very far
  off (t2100)

3
Weitzmans four exhibits

• Exhibit A geologically instantaneous pulse of
  greenhouse gases
• OK SRES and post-SRES BAU scenarios fit the bill
• Exhibit B long-tailed climate sensitivity
• Problem usually deterministic, or sampled from
  distribution with relatively low (and finite)
  upper support
• Exhibit C significant positive feedback from
  carbon cycle
• Problem often ignored in modelling, sometimes
  included but perhaps not on an adequate scale
• Exhibit D welfare loss from extreme warming
• Problem? Depends on assumption (can assume
  anything), but most models have quadratic loss
  function or less steep

4
Modelling strategy

• Emissions
• Non-probabilistic in PAGE (2002 version). Use A2
  scenario (high emissions) and compare BAU with
  550 stabilisation
• Climate sensitivity
• Compare truncated Stern Review pdf with
  Stainforth et al. pdf from climateprediction.net
• Carbon cycle feedbacks
• Probabilistic natural stimulation of CO2 in PAGE
• Loss function
• Compare truncated Stern Review pdf with lognormal
  distribution fitted over minimal a priori
  assumptions

5
Climate sensitivities compared
6
Loss functions compared
7
Probability of catastrophe
8
Probability of catastrophe BAU
9
Probability of catastrophe 550
10
Dependence on arbitrary upper bound
11
The intuition

• For a CRRA utility function with strictly
  positive relative risk
• aversion, marginal utility tends to infinity as
  consumption tends to
• zero

12
Social cost of carbon as a function of the
arbitrary upper bound
13
Sensitivity to risk and discounting
14
Sensitivity to risk and discounting similar
magnitudes
15
High impact, low probability?

• A brief empirical analysis of risk in the
  economics of climate change
• Dr Simon Dietz
• Grantham Research Institute on Climate Change and
  the Environment
• London School of Economics