[PPT] - Too Little Too Late, part II p Clean Innovation as Policy PowerPoint Presentation

SLIDE 1

Too Little Too Late, part II p

Clean Innovation as Policy Commitment Device

CREE annual workshop, Oslo 16-17 Sep 2013

Reyer Gerlagh, Sam Okullo, Mads Greaker

Tilburg University

SLIDE 2

Introduction / model / results / conclusion

WRE 96: IPCC92 wants too early reductions WRE 96: IPCC92 wants too early reductions

Wigley, Richels & Edmonds

Wigley, Richels & Edmonds (1996): IPCC92 reduces too early. Reasons for delay: Reasons for delay:

Return on capital (Hicks

compensation)

Vintages of existing dirty
Vintages of existing dirty

capital

Cheaper future clean

technologies technologies

Atmospheric depreciation

(increased carbon budget) Carbon price goes up with real return on capital + real depreciation

17 September 2013 2 Reyer Gerlagh

SLIDE 3

Introduction / model / results / conclusion

Reproducing WRE 96 Reproducing WRE 96

A simple Ramsey model

A simple Ramsey model

Cobb-Douglas production

with capital

Emissions proportional to
utput
Quadratic costs for

Quadratic costs for emissions reductions (linear marginal costs)



Decreasing over time



Decreasing over time

Emissions – multi-box

atmospheric CO2 – ceiling

17 September 2013 3 Reyer Gerlagh

SLIDE 4

Introduction / model / results / conclusion

Critique: Too little too late (ITC) Critique: Too little too late (ITC)

Claim: Cheap abatement

Claim: Cheap abatement becomes available only if used

> early abatement is needed
Ha-Duong et al 1997

Ha Duong et al 1997

Goulder and Mathai 2000
Van der Zwaan et al. 2002

2003 DEMETER 2003, … DEMETER

Manne and Barreto 2004
Popp 2006, ENTICE
Bosetti et al. 2006….
Gerlagh, Kverndokk &

Rosendahl 2009, 2014… ,

Acemoglu et al. 2012

17 September 2013 4 Reyer Gerlagh

SLIDE 5

Introduction / model / results / conclusion

Scientific uncertainty: stochastic targets Scientific uncertainty: stochastic targets

Scientific uncertainty → don’t know climate threshold

yet Scientific uncertainty → don t know climate threshold … yet

Assume climate threshold is known at some future date T →

Hedging (act-learn-act)

Theory: Ulph and Ulph 1997 and Webster 2000
IAMs: Manne and Richels 1995 Nordhaus and Popp 1997 Yohe
IAMs: Manne and Richels 1995, Nordhaus and Popp 1997, Yohe,

Andronova, Schlesinger 2004, Bosetti et al. 2009, Gerlagh and van der Zwaan 2011

What if objective targets ‘never’ arrive? If they need to be derived

endogenously, each period again? endogenously, each period again?



Assume International Cooperation is reached

17 September 2013 5 Reyer Gerlagh

SLIDE 6

Introduction / model / results / conclusion

Project Part I: Cost Effective Scenarios = TLTL Project Part I: Cost-Effective Scenarios = TLTL

Research question 1A: how will climate targets develop

Research question 1A: how will climate targets develop dynamically, if they are endogenous?

Frame: Scientific uncertainty → no clear climate threshold →

negative welfare as function of expected consequences

i.e. preferences over consumption stream + long-term climate

i.e. preferences over consumption stream long term climate

utcomes
Result: preferences are time-inconsistent
Climate change targets are not credible.



Assume in 2013: we find 450ppm too costly, so we go for 550ppm.



In 2030: Achieving 550ppm is equally costly, as was achieving 450 in 2013.

Result: Sequence of climate plans deviates from naive plans

17 September 2013 6 Reyer Gerlagh

SLIDE 7

Introduction / model / results / conclusion

Project Part I: Cost Effective Scenarios = TLTL Project Part I: Cost-Effective Scenarios = TLTL

Research question 1B: how big is the gap between committed

Research question 1B: how big is the gap between committed and naive climate target outcome?

Result: if we aim for 450ppm (2K) by 2000, we naively reach

570ppmv (>3K)



Sensitive to all details of model

17 September 2013 7 Reyer Gerlagh

SLIDE 8

Introduction / model / results / conclusion

Project Part II: Sophisticated policies Project Part II: Sophisticated policies

Research question 2A: what are characteristics of a sophisticated
Research question 2A: what are characteristics of a sophisticated

policy



Sophisticated policy = Markov equilibrium: each regulator predicts correctly p p y q g p y future response to current policies, and maximizes its own objectives (that are different from future objectives)

How to calculate a sophisticated scenario in an IAM?
How to calculate a sophisticated scenario in an IAM?
Does the sophisticated policy perform better/worse vis-a-vis the

naive policy naive policy



Irrelevance theorem (Iverson 2012): future climate policies don’t affect current optimal climate policies

17 September 2013 8 Reyer Gerlagh

SLIDE 9

Introduction / model / results / conclusion

Project Part II: Innovation as commitment device Project Part II: Innovation as commitment device

Research question 2B: Can clean innovation act as commitment
Research question 2B: Can clean innovation act as commitment

device to to overcome the time-inconsistency problem?

If so: clean energy innovation deserves support in excess of

carbon price. p

17 September 2013 9 Reyer Gerlagh

SLIDE 10

Introduction / model / results / conclusion

Cost effective with objective threshold Cost-effective with objective threshold

UNFCC: “Stabilization of greenhouse gases that would prevent
UNFCC: Stabilization of greenhouse gases that would prevent

dangerous anthropogenic interference with the climate system”

Conceptual framework:

Conceptual framework: max ( )

t τ t τ

W β U t s t

  

  . . ( ) s t Z t Z 

Where Z(t) is (set of) climate-variables (temperature, atmospheric

CO2, ocean acidification, cumulative emissions), and Z

_

is the CO2, ocean acidification, cumulative emissions), and Z is the ‘dangerous’ threshold

17 September 2013 10 Reyer Gerlagh

SLIDE 11

Introduction / model / results / conclusion

Cost effective with subjective threshold Cost-effective with subjective threshold

max ( ) ( )

t τ

W β U t D Z

 

   max ( ) ( ) ( )

t τ

W β U t D Z Z t Z



   

τ is the time of perspective (when planner decides on optimal path)
U*(t;τ) is the optimal utility at time t envisaged at τ
U (t;τ) is the optimal utility at time t envisaged at τ
Z*(τ) is the optimal target envisaged at τ

17 September 2013 11 Reyer Gerlagh

SLIDE 12

Introduction / model / results / conclusion

Numerical model

Basic Ramsey growth model

Numerical model

Basic Ramsey growth model
Capital, population growth, CD production function
Calibrated 3 box atmosphere ocean biosphere model
Calibrated 3-box atmosphere-ocean-biosphere model
Quadratic emission reduction costs (loss of output)
Social costs of atmospheric ppmv quadratic such that in 2000 a
Social costs of atmospheric ppmv quadratic, such that in 2000 a

450 ppmv target is optimal

ETC1 (current version): transition costs: reducing emissions by 1%

more per year (relative to BAU) adds 1% GDP costs more per year (relative to BAU) adds 1% GDP costs

ETC2 (in progress): endogenous growth choice between TFP &

emission intensity y

17 September 2013 12 Reyer Gerlagh

SLIDE 13

Introduction / model / results / conclusion

Model: technology Model: technology

 

2

1

t τ 

 

1

t

t t t

C Y K δ K    

 

2

1

1

max

ln /

5

2 27

t τ τ t t t t t τ

W ρ L C L Δ ppm

  

          

1

t

t t t

C Y K δ K



 

1 α α t t t t

X K A L





 

2 2 1

1 Ω 1

2

t t t t t t t

µ µ Y Temp φ µ X



     1

t

t t t

µ Z σ X   ;

;
Temp

f Z Z Z

1

;...

;
t

t t

Temp f Z Z Z





17 September 2013 13 Reyer Gerlagh

SLIDE 14

Introduction / model / results / conclusion

Committing regulator

Two decision variables: investments (K) & abatement (μ)

Committing regulator

Two decision variables: investments (K) & abatement (μ)
Assume that regulator controls all future decisions
Calculate optimal path at time 2000 (or 2020)
Calculate optimal path at time 2000 (or 2020)

 

2

1

1

l

/

5

27

t τ

W L C L Δ

  

    

 

1
max

ln /

5

2 27

t τ τ t t t t t τ

W ρ L C L Δ ppm



        

 

* * * * * *

 

* * * * * * , , ,

;max

;max
m

;

2

/ a 75 x

t t τ τ τ τ τ τ t τ τ τ

MRT C MRS C Δ ppm ppm τ ppm C   

17 September 2013 14 Reyer Gerlagh

SLIDE 15

Introduction / model / results / conclusion

Time inconsistency of regulator

Regulator at time τ looking at current consumption:

Time-inconsistency of regulator

Regulator at time τ, looking at current consumption:

 

* * * * * * , , ,

;max

;max
m

;

2

/ a 75 x

t t τ τ τ τ τ τ t τ τ τ

MRT C MRS C Δ ppm ppm τ ppm C   

Regulator at time τ, looking at future consumption

 

* * * * * *

Regulator at time τ+1 (looking at previous plan):

 

* * * * * * 1, 1 , , 1

;max

;max

;

275 /1
max

t t τ τ τ τ τ τ t τ τ τ

ppm p MRT C MRS ρ C Δ ppm τ C pm

  

   

Regulator at time τ+1 (looking at previous plan):

 

* * * * * * 1, 1, 1,

;max

;max

1

;
275

m / ax

t t τ τ τ τ τ τ t τ τ τ

MRT C MRS C Δ ppm pp ppm C m τ

  

   

17 September 2013 15 Reyer Gerlagh

SLIDE 16

Introduction / model / results / conclusion

Naive regulator

Regulator assumes that it controls all future decisions

Naive regulator

Regulator assumes that it controls all future decisions
But each subsequent regulator re-optimizes
Calculate optimal paths for all times 2000 2020 2030
Calculate optimal paths for all times 2000, 2020, 2030, …
Naive path is sequence of first periods

17 September 2013 16 Reyer Gerlagh

SLIDE 17

Introduction / model / results / conclusion

Sophisticated regulator (Markov equilibrium)

Regulator forecasts future response

Sophisticated regulator (Markov equilibrium)

Regulator forecasts future response
Calculate equilibrium path
1

Υ Θ

Backwards recursively estimate functions:

1

2 1 1 1 1 1

,

1 1 max

2

τ

τ Θ τ τ τ τ τ τ τ

Υ Θ Θ Θ ρ W U ΔΓ Θ



    

   

Backwards recursively estimate functions:

1 1

1

τ τ τ τ τ

Υ Θ Υ U Θ

 

   1

τ τ τ

ρ 

1 1

max

;

τ τ τ τ τ

ppm Γ Θ Γ Θ

 



17 September 2013 17 Reyer Gerlagh

SLIDE 18

Introduction / model / results / conclusion

Naive policies: Emissions

2000 proposal: raise carbon

Naive policies: Emissions

2000 proposal: raise carbon

prices quickly so that emissions reduce quickly q y after 2030.

20 years of BAU (inaction)
2020 proposal: delay action

by about 15 years, relative to 2000 l 2000-proposal.

2030 proposal: delay action

b another 5 ears by another 5 years

etcetera

17 September 2013 18 Reyer Gerlagh

SLIDE 19

Introduction / model / results / conclusion

Naive policies: Atmospheric CO2 concentrations

2000 proposal: stabilize

Naive policies: Atmospheric CO2 concentrations

2000 proposal: stabilize

atmospheric CO2 at 450 ppm. pp

20 years of BAU (inaction)
2020 proposal: stabilize at

p p 475 ppm

2030 proposal: stabilize at

490 ppm

Ultimately reached: 570 ppm

17 September 2013 19 Reyer Gerlagh

SLIDE 20

Introduction / model / results / conclusion

Sophisticated policies: virtually no difference

Sophisticated policy does not

Sophisticated policies: virtually no difference

Sophisticated policy does not

deviate (much) from Naive policy! p y



Irrelevance theorem (Iverson)

17 September 2013 20 Reyer Gerlagh

SLIDE 21

Introduction / model / results / conclusion

Commitment through ‘technology’

Assume Transition Costs

Commitment through technology

Assume Transition Costs

(very rudimentary ETC)

More abatement now =

More abatement now cheaper future abatement

Abatement = commitment

device

Sophisticated policy abates

more compared to Naive policy!

TU-SSB project: endogenous

growth specification

17 September 2013 21 Reyer Gerlagh

SLIDE 22

Introduction / model / results / conclusion

I Too little too late

Our paper specifies climate target as an endogenous variable
I. Too little too late
Our paper specifies climate target as an endogenous variable

dependent on preferences for climate stabilization, in addition to preferences for consumption streams p p

Time-inconsistency of preferences is fundamental property of

sustainability concerns? (Gerlagh & Liski 2012)

Climate targets tend to erode over time both in naive and

sophisticated policies

17 September 2013 22 Reyer Gerlagh

SLIDE 23

Introduction / model / results / conclusion

II Clean innovation as commitment device

Moving targets: Need to think about commitment mechanisms
II. Clean innovation as commitment device
Moving targets: Need to think about commitment mechanisms.
Pledges don’t commit.
Clean energy technologies can work as commitment device
Clean energy technologies can work as commitment device
Transition costs induce commitment device: sophisticated

abatement effort > naive abatement effort abatement effort > naive abatement effort

Is cheap clean technology a good commitment device / do we

need to support clean technology beyond carbon price ? pp gy y p

17 September 2013 23 Reyer Gerlagh