R

1
RD as Greenhouse Insurance

• Erin Baker
• University of Massachusetts, Amherst
• June 2003
• Leon Clarke PNNL
• John Weyant Stanford University

2
Climate Change

• Uncertainty about how emissions today will cause
  damages tomorrow.
• But, we are learning more and more.
• Uncertainty and Learning impact current decisions
• Conclusion Uncertainty Learning less control
  of emissions.
• Kolstad
• Ulph Ulph
• Manne Richels
• Baker

3
What about RD?

• RD planning is further complicated by different
  programs
• Solar PVs versus efficiency of coal-fired
  electricity
• We consider optimal RD
• uncertainty and learning about climate damages
• choice of RD program

4
RD Policy

• Socially Optimal RD
• induced private RD
• The Gap
• The Gap is the rationale for RD policy.
• How is The Gap impacted by uncertain climate
  change damages / uncertain price of carbon?

5
Todays Talk

• Present a unifying framework for thinking about
  RD assumptions
• Discuss impact of assumptions in a sequential
  decision making problem
• Social planner point of view
• Theoretical
• Numerical
• Firm point of view
• Initial

6
The Production Function
t standard inputs e emissions
7
From production function to abatement cost curve
Production Function
Abatement Cost Curve

Cost
a
m
0
e emissions
m emission reductions
8
RD as leftward shiftsequestration?

Production Function
9
Multiplicative ShiftCost Reduction
tmax
Production Function
Cost
a
a
1-a
1-a
tmin
e
m
e
0
0
1
The abatement cost curve pivots downward
10
Emissions Reduction
Production Function
Cost
a
1- a
a
1- a
a
1- a
m
e
The abatement cost curve pivots to the right
11
RD impacts convexity of cost curve / production
function
Flatter ? RD increases in risk
More convex ? RD decreases in risk
12
Integrated Assessment Model

• William Nordhauss DICE
• Optimal Growth Climate Model
• Added uncertainty, using stochastic programming
• Added RD as a decision variable
• One time decision in 1st period before learning
• Cost reduction implemented in 50 years, after
  learning.

13
2 Types of increasing risk
Increasing Probability certain low medium high
Probability of high damage 0 .018
.05 .08333 Value of high damage -
.042 .042 .042 Value of low
damage .0035 .002794 .001473 0
Increasing Damage certain low
high Probability of high damage 0 .018
.002374 Value of high damage
- .042 .3 Value of low
damage .0035 .002794 .002794
14
Results Increasing Probability
0.4
16
14
0.3
12
10
Optimal RD
0.2
Billions of US
8
6
0.1
4
2
0
0
1
2
3
4
1
2
3
4
Risk
Risk
Emissions
Sequestration
Cost Reduction
15
Results Increasing Damages
5
0.12
4
0.08
3
Optimal RD
Billions of US dollars
2
0.04
1
0
0
1
2
3
1
2
3
Risk
Risk
Emissions
Sequestration
Cost Reduction
16
Induced RD

• Consider a firms response to an uncertain carbon
  tax.
• Firm pays carbon tax (e.g. electric generator)
• Firm produces RD

17
Induced RD initial results

• Investment in sequestration is independent of
  risk (only depends on expected value of carbon
  tax).
• The Gap Shrinks
• Emissions Reduction decreases in risk
• Cost Reduction of no-carbon alternatives
• The Gap Grows

18
Conclusions

• RD can be a hedge against uncertainty,
• But, it depends on what kind.
• Private response to uncertainty may diverge from
  social optimal even without risk aversion.
• May be a rationale for government involvement in
  RD into alternative energy sources.

19
Future Research

• Empirical
• The shape of production function, impact of RD
• Returns to RD
• Methodological
• incorporate uncertain returns to RD
• endogenous technical change
• Policy
• Private RD response to uncertainty
• Compare private response, under a variety of
  instruments to optimal RD.

20
Research Plan

• Carefully consider the impact of RD.
• Many papers make assumptions about RD
• What do these assumptions really mean?
• How do different assumptions impact results?
• Look at how optimal RD responds to uncertainty
  and learning.
• What kinds of RD are a hedge against uncertainty?

21
How can we buy insurance?

• RD, particularly into low or no-carbon
  technologies
• If we learn that climate damages will be severe,
  we can more quickly and easily reduce emissions.
• Endogenous Technical Change
• Goulder et al.
• Nordhaus
• Buonanno et al. 2003

22
Stochastic, economy-environment model with RD

• DICE

23
Uncertainty with learning

• Stochastic Programming approach
• Uncertainty for 5 periods (50 years) then perfect
  learning
• E(q2).0035

24
RD

• Alter the cost function
• Leftward shift
• Multiplicative
• Or the emissions output ratio s
• ratio shift

25
Cost of RD

• k 18 for leftward and ratio
• k 1 for multiplicative

26
Sequential Decision Making Under Uncertainty
27
Sequential Decision Making Under Uncertainty
Work Backwards
28
Sequential Decision Making Under Uncertainty
Work Backwards
29
Sequential Decision Making Under Uncertainty
Work Backwards
30
Sequential Decision Making Under Uncertainty

• Rothschild-Stiglitz show that a will increase
  with uncertainty iff Va is concave in z.

31
Sequential Decision Making Under Uncertainty

• Generally inconclusive.
• But,

RD is more likely to decrease in risk RD is
more likely to increase in risk
32
RD as leftward shiftsequestration?
t
a
a
e
e
e
33
Multiplicative Shift Cost Reduction
tmax
a
1-a
tmin
e
e
34
Emissions reduction
a
1- a
a
1- a
e