THE CLIMATE POLICY DILEMMA

download report

Transcript THE CLIMATE POLICY DILEMMA

THE CLIMATE POLICY DILEMMA
Robert S. Pindyck
M.I.T.
December 2012
INTRODUCTION
• Question: Should stringent climate policy be an
immediate priority for environmental economists?
• Waning political support for stringent GHG abatement.
• Economic argument for stringent policy far from clear.
– Disagreement over likelihood of alternative climate
outcomes and impacts, as well as nature and extent of
uncertainty over those outcomes and impacts.
– Disagreement over framework to use to evaluate benefits
from abatement, including social welfare function and
discount rate.
• Makes climate policy difficult to evaluate, and a hard
sell for the public at large.
INTRODUCTION (Con‘t)
• Why is it so difficult to apply cost-benefit analysis to
GHG abatement policy?
– Long time horizon: key parameters (discount rate, IRRA)
become crucial.
– Uncertainty over climate change, hard to even characterize
the uncertainty.
– Uncertainty over impact of climate change: we know
almost nothing.
• Last 20 years has seen many quantitative studies of
climate policy, including a variety of IAMs.
• Claim: We can’t make the case for immediate adoption
of stringent policy based on consensus distributions,
IAMs.
• Must consider catastrophic outcomes.
OVERVIEW
• Policy evaluation: uncertainties and areas of
disagreement.
– Welfare function, key parameters.
– Uncertainty over climate outcomes.
– Uncertainty over impact of climate change. May be
“unknowable.”
– Implies IAMs not useful for policy evaluation.
• Case for stringent policy must be based on
catastrophic outcomes.
– “Plausible” probabilities and impacts.
– What about non-climate catastrophes?
• Conclusion: Can a convincing case be made for
stringent GHG abatement now?
Economic Evaluation of Climate Policy
• Standard approach: Calculate NPV of current and
expected future costs and benefits.
– This is essence of IAM-based analysis.
– Considerable uncertainties, hard to characterize nature
and extent.
– No consensus on social welfare measurement.
• Assume we agree on CRRA utility: U(C) = C1-η/(1-η)
• What rate to discount future utility (rate of time
preference), δ?
• What value for index of risk aversion, η?
• Can’t simply choose δ and η to get desired answer.
DISCOUNT RATE
• Financial and macro data: δ is 2% to 5%.
– Even 2% makes PV of future welfare gains from GHG
abatement too small to justify any policy.
– Lower rate for intergenerational comparisons, i.e.,
time horizon of 50 or 100 years? Stern Review sets δ
close to zero.
– Should δ = 0 on “ethical” grounds? Economists have
little to say. John versus Jane.
– δ is a policy parameter. Reflects values of policy
makers. Might or might not reflect voters.
– δ might be positive, zero, or even negative.
• Problem: if δ is an arbitrary parameter, hard to
make a case for (or against) stringent policy.
Must also make case for value of δ.
INDEX OF RELATIVE RISK AVERSION
• The IRRA, η, also critical. Two effects:
– Large η implies U’(C) declines rapidly as C rises, reducing
benefits from preventing future loss of C.
– Large η implies high risk aversion, increasing benefits if
future C uncertain.
– Unless risk aversion or uncertainty is extreme, first effect
will dominate.
• What is “correct” value for η? Behavioral or policy
parameter, not an ethical one.
– If behavioral, range is 1 to 4. If policy, 1 to 3.
– If η ≥ 2, welfare gains from policy too small.
– Stern Review sets η = 1.
• Problem: if η is arbitrary, hard to make case for (or
against) stringent policy.
UNCERTAINTY OVER TEMPERATURE
• In JEEM (2012), I estimated WTP to ensure that T ≤ 3°C.
– Growth rate of C: g = g0 ─ γT, with γ calibrated to IAMs.
.01t
– Temperature: Tt  2T2100 [1  (1 / 2) ]
– T2100 is distributed as gamma, mean = 3°, SD = 2.1°
r
f (T ; r,  ,  ) 
(T   ) r 1 e   ( T  )
( r )
– Fitting to mean and SD implies r = 3.8, λ = 0.92, and θ = -1.13.
– For δ = 0, η ≥ 1.5, WTP < 2% of GDP.
– Can get WTP of 3% or more if δ = 0, η close to 1.
– Can get higher WTP by doubling impact parameter γ.
• Do other distributions for T imply higher WTP?
• Calculate WTP using Frechet (GEV Type II) and Roe-Baker
distributions. Both are fat-tailed.
WTP USING ALTERNATIVE DISTRIBUTIONS
• Frechet:
f (T ; k ,  ,  )  (1 /  )exp[ (1  kz ) 1/ k ](1  kz ) 11/ k
where z = (T-μ)/σ, k > 0, and T ≥ μ – σ/k.
– Calibrating to mean = 3°, SD = 2.1° gives k = 0.28, μ = 2.15, σ = 0.195.
• Roe-Baker:
g (T ; f ,  f ,  ) 
f
2

1
1 1 f 1/ z  
exp - 
 
2

f
2 z
 2
 

where z = T + θ. Calibrating to mean = 3°, SD = 2.1° gives 𝑓 = 0.797, σf = .0441,
and θ = 2.13. (The feedback parameter in the Roe-Baker model is normally
distributed with mean and SD 𝑓 and σf respectively.)
• Graphs compare gamma, Frechet, and Roe-Baker distributions
for T2100 and resulting WTP.
THREE DISTRIBUTIONS FOR T2100
THREE DISTRIBUTIONS: HIGH T2100
IMPLICATIONS FOR WTP
WTP WITH HIGH IMPACT
UNCERTAINTY OVER IMPACT
• Why so difficult to estimate economic impact of
climate change?
– Very little data on which to base empirical work.
– Little or no economic theory explaining impact of higher
temperatures.
– Climate change is slow, creating potential for adaptation.
How much adaptation will occur?
• Our understanding of economic impact unlikely to
improve in next 20 years.
• May be in the realm of the “unknowable.”
• Most IAMs posit ad hoc loss function relating T to GDP.
– E.g., DICE Model has L(T) = 1/(1 + π1T + π2T2 )
– Weak tool for policy.
CATASTROPHIC CLIMATE CHANGE
• Case for stringent policy must be based on chance of
catastrophic outcome.
• Not a climate outcome – a catastrophic impact of
whatever climate change might occur.
• Outside the realm of IAMs and WTP estimates.
• What to do? Roughly estimate probabilities of large
climate changes and distributions for impact, as in
studies of “consumption disasters.”
– Find “plausible” range of catastrophic outcomes, measured
by decline in productive capital. Find plausible probabilities.
– Calculate WTP to avert outcomes or to reduce probabilities.
– Is WTP large and robust to range of δ and η?
• This does not have perceived precision of IAM-based
analysis. But that precision is illusory.
• Given “unknowables,” can only rely on the “plausible.”
MULTIPLE CATASTROPHES
• Suppose analysis based on “plausible” outcomes and
probabilities yields high WTP, e.g., 10% of GDP. Are we
home?
• Maybe not. Must consider other potential catastrophes:
nuclear or biological terrorist attack, mega-virus, nonclimate environmental catastrophe ... (use your
imagination).
• Calculate WTPs for these catastrophes the same way.
• Problem: WTPs not additive. When taken as a group, WTP
for each potential catastrophe (including climate) will fall.
– Non-climate catastrophes reduce expected GDP growth,
increasing expected future marginal utility before climate
catastrophe occurs. This increases WTP to avoid climate change.
– With more catastrophes, large fraction of GDP needed to keep us
safe. This “income effect” reduces WTP for climate. It dominates.
CONCLUSIONS
• Should we push for early adoption of a stringent GHG
abatement policy? I have not answered this. But:
– Case cannot be made based on “likely outcomes,” i.e.,
distributions for temperature consistent with IPCC, and
economic impact functions used in most IAMs.
– Greatest uncertainty is with impact of climate change.
Economic loss functions for most IAMs are ad hoc. Not
surprising given how little we know.
– Economic impact of climate change may be in the realm of
the “unknowable.”
– Case for stringent abatement must be based on the
possibility of catastrophic outcome. This means economic
outcome, not climate outcome.
– Need “plausible” estimates of probabilities of various
climate outcomes, and impacts from those outcomes.
– Must consider other potential catastrophes as well.