151
2
Integrated Risk and
Uncertainty Assessment
of Climate Change
Response Policies
Coordinating Lead Authors:
Howard Kunreuther (USA), Shreekant Gupta (India)
Lead Authors:
Valentina Bosetti (Italy), Roger Cooke (USA), Varun Dutt (India), Minh Ha-Duong (France),
Hermann Held (Germany), Juan Llanes-Regueiro (Cuba), Anthony Patt (Austria / Switzerland),
Ekundayo Shittu (Nigeria / USA), Elke Weber (USA)
Contributing Authors:
Hannes Böttcher (Austria / Germany), Heidi Cullen (USA), Sheila Jasanoff (USA)
Review Editors:
Ismail Elgizouli (Sudan), Joanne Linnerooth-Bayer (Austria / USA)
Chapter Science Assistants:
Siri-Lena Chrobog (Germany), Carol Heller (USA)
This chapter should be cited as:
Kunreuther H., S. Gupta, V. Bosetti, R. Cooke, V. Dutt, M. Ha-Duong, H. Held, J. Llanes-Regueiro, A. Patt, E. Shittu, and E.
Weber, 2014: Integrated Risk and Uncertainty Assessment of Climate Change Response Policies. In: Climate Change 2014:
Mitigation of Climate Change. Contribution of Working Group III to the Fifth Assessment Report of the Intergovernmental
Panel on Climate Change [Edenhofer, O., R. Pichs-Madruga, Y. Sokona, E. Farahani, S. Kadner, K. Seyboth, A. Adler, I. Baum,
S. Brunner, P. Eickemeier, B. Kriemann, J. Savolainen, S. Schlömer, C. von Stechow, T. Zwickel and J.C. Minx (eds.)]. Cam-
bridge University Press, Cambridge, United Kingdom and New York, NY, USA.
152152
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
Contents
Executive Summary � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 154
2�1 Introduction � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 155
2�2 Metrics of uncertainty and risk � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 157
2�3 Risk and uncertainty in climate change � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 157
2�3�1 Uncertainties that matter for climate policy choices
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 157
2�3�2 What is new on risk and uncertainty in AR5
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 159
2�4 Risk perception and responses to risk and uncertainty � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 160
2�4�1 Considerations for design of climate change risk reduction policies
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 160
2�4�2 Intuitive and deliberative judgment and choice
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 160
2�4�3 Consequences of intuitive decision making
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 161
2.4.3.1 Importance of the status quo
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 161
2.4.3.2 Focus on the short term and the here-and-now
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 162
2.4.3.3 Aversion to risk, uncertainty, and ambiguity
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 163
2�4�4 Learning
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 164
2�4�5 Linkages between different levels of decision making
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 165
2�4�6 Perceptions of climate change risk and uncertainty
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 166
2�5 Tools and decision aids for analysing uncertainty and risk � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 168
2�5�1 Expected utility theory
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 168
2.5.1.1 Elements of the theory
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 168
2.5.1.2 How can expected utility improve decision making?
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
2�5�2 Decision analysis
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 169
2.5.2.1 Elements of the theory
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 169
2.5.2.2 How can decision analysis improve
decision making? . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
2�5�3 Cost-benefit analysis
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 170
2.5.3.1 Elements of the theory
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
2.5.3.2 How can CBA improve decision making?
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
2.5.3.3 Advantages and limitations of CBA
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 170
153153
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
2�5�4 Cost-effectiveness analysis � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 171
2.5.4.1 Elements of the theory
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
2.5.4.2 How can CEA improve decision making?
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
2.5.4.3 Advantages and limitations of CEA over CBA
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
2�5�5 The precautionary principle and robust decision making
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 172
2.5.5.1 Elements of the theory
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 172
2�5�6 Adaptive management
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 173
2�5�7 Uncertainty analysis techniques
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 173
2.5.7.1 Structured expert judgment
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 173
2.5.7.2 Scenario analysis and ensembles
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 175
2�6 Managing uncertainty, risk and learning � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 177
2�6�1 Guidelines for developing policies
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 177
2�6�2 Uncertainty and the science/policy interface
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 178
2�6�3 Optimal or efficient stabilization pathways (social planner perspective)
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 178
2.6.3.1 Analyses predominantly addressing climate or damage response uncertainty
. . . . . . . . . . . . . . . . . . . . . . . 178
2.6.3.2 Analyses predominantly addressing policy response uncertainty
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
2�6�4 International negotiations and agreements
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 181
2.6.4.1 Treaty formation
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 181
2.6.4.2 Strength and form of national commitments
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
2.6.4.3 Design of measurement, verification regimes, and treaty compliance
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 182
2�6�5 Choice and design of policy instruments
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 183
2.6.5.1 Instruments creating market penalties for GHG emissions
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 183
2.6.5.2 Instruments promoting technological RDD&D
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 184
2.6.5.3 Energy efficiency and behavioural change
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
2.6.5.4 Adaptation and vulnerability reduction
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 186
2�6�6 Public support and opposition to climate policy
� � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 187
2.6.6.1 Popular support for climate policy
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 187
2.6.6.2 Local support and opposition to infrastructure projects
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 188
2�7 Gaps in knowledge and data � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 189
2�8 Frequently Asked Questions � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 189
References � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � � 192
154154
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
Executive Summary
The scientific understanding of climate change and the impact it
has on different levels of decision-making and policy options has
increased since the publication of the Intergovernmental Panel on
Climate Change (IPCC) Fourth Assessment Report (AR4). In addi-
tion, there is a growing recognition that decision makers often rely
on intuitive thinking processes rather than undertaking a systematic
analysis of options in a deliberative fashion. It is appropriate that
climate change risk management strategies take into account both
forms of thinking when considering policy choices where there is risk
and uncertainty.
Consideration of risk perception and decision processes can
improve risk communication, leading to more effective poli-
cies for dealing with climate change By understanding the sys-
tematic biases that individuals utilize in dealing with climate change
problems, one can more effectively communicate the nature of the
climate change risk. An understanding of the simplified decision
rules employed by decision makers in making choices may be helpful
in designing policies that encourage the adoption of mitigation and
adaptation measures. [Section 2.4]
Decision processes often include both deliberative and intuitive
thinkingWhen making mitigation and adaptation choices, decision
makers sometimes calculate the costs and benefits of their alterna-
tives (deliberative thinking). They are also likely to utilize emotion- and
rule-based responses that are conditioned by personal past experience,
social context, and cultural factors (intuitive thinking). [2.4.2]
Laypersons tend to judge risks differently than experts Layper-
sons’ perceptions of climate change risks and uncertainties are often
influenced by past experience, as well as by emotional processes that
characterize intuitive thinking. This may lead them to overestimate or
underestimate the risk. Experts engage in more deliberative thinking
than laypersons by utilizing scientific data to estimate the likelihood
and consequences of climate change. [2.4.6]
Cost-benefit analysis (CBA) and cost-effectiveness analysis
(CEA) can enable decision makers to examine costs and ben-
efits, but these methodologies also have their limitations Both
approaches highlight the importance of considering the likelihood of
events over time and the importance of focusing on long-term hori-
zons when evaluating climate change mitigation and adaptation poli-
cies. CBA enables governments and other collective decision-making
units to compare the social costs and benefits of different alternatives.
However, CBA cannot deal well with infinite (negative) expected utili-
ties arising from low probability catastrophic events often referred to
as ‘fat tails’. CEA can generate cost estimates for stabilizing green-
house gas (GHG) concentrations without having to take into account
the uncertainties associated with cost estimates for climate change
impacts. A limitation of CEA is that it takes the long-term stabilization
as a given without considering the economic efficiency of the target
level. [2.5.3, 2.5.4]
Formalized expert judgment and elicitation processes improve
the characterization of uncertainty for designing climate
change strategies (high confidence). Experts can quantify uncer-
tainty through formal elicitation processes. Their judgments can char-
acterize the uncertainties associated with a risk but not reduce them.
The expert judgment process highlights the importance of undertaking
more detailed analyses to design prudent climate policies. [2.5.6]
Individuals and organizations that link science with policy grap-
ple with several different forms of uncertainty� These uncertain-
ties include absence of prior agreement on framing of problems and
ways to scientifically investigate them (paradigmatic uncertainty), lack
of information or knowledge for characterizing phenomena (epistemic
uncertainty), and incomplete or conflicting scientific findings (transla-
tional uncertainty). [2.6.2]
The social benefit from investments in mitigation tends to
increase when uncertainty in the factors relating GHG emissions
to climate change impacts are considered (medium confidence).
If one sets a global mean temperature (GMT) target, then normative
analyses that include uncertainty on the climate response to elevated
GHG concentration, suggest that investments in mitigation measures
should be accelerated. Under the assumption of nonlinear impacts of
a GMT rise, inclusion of uncertainty along the causal chain from emis-
sions to impacts suggests enhancing mitigation. [2.6.3]
The desirability of climate policies and instruments are affected
by decision makers’ responses to key uncertaintiesAt the
national level, uncertainties in market behaviour and future regulatory
actions have been shown to impact the performance of policy instru-
ments designed to influence investment patterns. Both modelling and
empirical studies have shown that uncertainty as to future regulatory
and market conditions adversely affects the performance of emission
allowance trading markets [2.6.5.1]. Other studies have shown that
subsidy programmes (e. g., feed-in tariffs, tax credits) are relatively
immune to market uncertainties, but that uncertainties with respect to
the duration and level of the subsidy program can have adverse effects
[2.6.5.2]. In both cases, the adverse effects of uncertainty include less
investment in low-carbon infrastructure, increasing consumer prices,
and reducing the pressure for technological development.
Decision makers in developing countries often face a particu-
lar set of challenges associated with implementing mitigation
policies under risk and uncertainty (medium confidence). Manag-
ing risk and uncertainty in the context of climate policy is of particular
importance to developing countries that are resource constrained and
face other pressing development goals. In addition, institutional capac-
ity in these countries may be less developed compared to advanced
economies. Therefore, decision makers in these countries (governments
and economic agents such as firms, farmers, households, to name a
155155
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
few) have less room for ‘error’ (uncertain outcomes and / or wrong or
poorly implemented policies). The same applies to national, regional
and local governments in developed countries who can ill afford to
waste scarce resources through policy errors. [Box 2.1]
2.1 Introduction
This framing chapter considers ways in which risk and uncertainty can
affect the process and outcome of strategic choices in responding to
the threat of climate change.
‘Uncertainty’ denotes a cognitive state of incomplete knowledge that
results from a lack of information and / or from disagreement about
what is known or even knowable. It has many sources ranging from
quantifiable errors in the data to ambiguously defined concepts or ter-
minology to uncertain projections of human behaviour. The Guidance
Note for Lead Authors of the IPCC Fifth Assessment Report on Consis-
tent Treatment of Uncertainties (Mastrandrea etal., 2010) summarizes
alternative ways of representing uncertainty. Probability density func-
tions and parameter intervals are among the most common tools for
characterizing uncertainty.
‘Risk’ refers to the potential for adverse effects on lives, livelihoods,
health status, economic, social and cultural assets, services (includ-
ing environmental), and infrastructure due to uncertain states of the
world. To the extent that there is a detailed understanding of the char-
acteristics of a specific event, experts will normally be in agreement
regarding estimates of the likelihood of its occurrence and its resulting
consequences. Risk can also be subjective in the sense that the likeli-
hood and outcomes are based on the knowledge or perception that a
person has about a given situation. There may also be risks associated
with the outcomes of different climate policies, such as the harm aris-
ing from a change in regulations.
There is a growing recognition that today’s policy choices are highly
sensitive to uncertainties and risk associated with the climate system
and the actions of other decision makers. The choice of climate policies
can thus be viewed as an exercise in risk management (Kunreuther
etal., 2013a). Figure 2.1 suggests a risk management framework that
serves as the structure of the chapter.
After defining risk and uncertainty and their relevant metrics (Section
2.2), we consider how choices with respect to climate change policy
options are sensitive to risk and uncertainty (Section 2.3). A taxon-
omy depicts the levels of decision making ranging from international
agreements to actions undertaken by individuals in relation to climate
change policy options under conditions of risk and uncertainty that
range from long-term global temperature targets to lifestyle choices.
The goals and values of the different stakeholders given their immedi-
ate and long-term agendas will also influence the relative attractive-
ness of different climate change policies in the face of risk and uncer-
tainty.
Sections 2.4, 2.5 and 2.6 characterize descriptive and normative
theories of decision-making and models of choice for dealing with
risk and uncertainty and their implications for prescriptive analysis.
Descriptive refers to theories of actual behaviour, based on experi-
mental evidence and field studies that characterize the perception
of risk and decision processes. Normative in the context of this chap-
ter refers to theories of choice under risk and uncertainty based on
abstract models and axioms that serve as benchmarks as to how
decision makers should ideally make their choices. Prescriptive refers
to ways of improving the decision process and making final choices
(Kleindorfer etal., 1993).
A large empirical literature has revealed that individuals, small groups
and organizations often do not make decisions in the analytic or ratio-
nal way envisioned by normative models of choice in the economics
and management science literature. People frequently perceive risk
in ways that differ from expert judgments, posing challenges for risk
communication and response. There is a tendency to focus on short
time horizons, utilize simple heuristics in choosing between alterna-
tives, and selectively attend to subsets of goals and objectives.
To illustrate, the voting public in some countries may have a wait-
and-see attitude toward climate change, leading their governments to
postpone mitigation measures designed to meet specified climate tar-
gets (Sterman, 2008; Dutt and Gonzalez, 2011). A coastal village may
decide not to undertake measures for reducing future flood risks due
to sea level rise (SLR), because their perceived likelihood that SLR will
cause problems to their village is below the community council’s level
of concern.
Section 2.4 provides empirical evidence on behavioural responses to
risk and uncertainty by examining the types of biases that influence
individuals’ perception of the likelihood of an event (e. g., availability,
learning from personal experience), the role that emotional, social, and
cultural factors play in influencing the perception of climate change
risks and strategies for encouraging decision makers to undertake
cost-effective measures to mitigate and adapt to the impacts of cli-
mate change.
A wide range of decision tools have been developed for evaluating
alternative options and making choices in a systematic manner even
when probabilities are difficult to characterize and / or outcomes are
uncertain. The relevance of these tools for making more informed
decisions depends on how the problem is formulated and framed, the
nature of the institutional arrangements, and the interactions between
stakeholders (Hammond etal., 1999; Schoemaker and Russo, 2001).
Governments debating the merits of a carbon tax may turn to cost-
benefit analysis or cost-effectiveness analysis to justify their positions.
They may need to take into account that firms who utilize formal
156156
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
approaches, such as decision analysis, may not reduce their emissions
if they feel that they are unlikely to be penalized because the carbon
tax will not be well enforced. Households and individuals may find the
expected utility model or decision analysis to be useful tools for evalu-
ating the costs and benefits of adopting energy efficient measures
given the trajectory of future energy prices.
Section 2.5 delineates formal methodologies and decision aids for ana-
lysing risk and uncertainty when individuals, households, firms, com-
munities and nations are making choices that impact their own well-
being and those of others. These tools encompass variants of expected
utility theory, decision analysis, cost-benefit analyses or cost-effective-
ness analyses that are implemented in integrated assessment models
(IAMs). Decision aids include adaptive management, robust decision
making and uncertainty analysis techniques such as structured expert
judgment and scenario analysis. The chapter highlights the importance
of selecting different methodologies for addressing different problems.
Developing robust policy response strategies and instruments should
take into account how the relevant stakeholders perceive risk and their
behavioural responses to uncertain information and data (descriptive
analysis). The policy design process also needs to consider the meth-
odologies and decision aids for systematically addressing issues of
risk and uncertainty (normative analysis) that suggest strategies for
improving outcomes at the individual and societal level (prescriptive
analysis).
Section 2.6 examines how the outcomes of particular options, in terms
of their efficiency or equity, are sensitive to risks and uncertainties and
affect policy choices. After examining the role of uncertainty in the sci-
ence / policy interface, it examines the role of integrated assessment
models (IAMs) from the perspective of the social planner operating
at a global level and the structuring of international negotiations and
paths to reach agreement. Integrated assessment models combined
with an understanding of the negotiation process for reaching inter-
national agreements may prove useful to delegates for justifying the
positions of their country at a global climate conference. The section
also examines the role that uncertainty plays in the performance of dif-
ferent technologies now and in the future as well as how lifestyle deci-
sions such as investing in energy efficient measures can be improved.
Figure 2�1 | A risk management framework. Numbers in brackets refer to sections where more information on these topics can be found.
Managing Uncertainty, Risk and Learning
(Prescriptive Analysis)
[Section 2.6]
Risk Perception and Responses
to Risk and Uncertainty
(Descriptive Analysis)
[Section 2.4]
Tools and Decisions Aids for
Analysing Uncertainty and Risk
(Normative Analysis)
[Section 2.5]
Impact of Risk and Uncertainty on
Climate Change Policy Choices
[Sections 2.2 and 2.3]
157157
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
The section concludes by examining the roles that risk and uncertainty
play in support of or opposition to climate policies.
The way climate change is managed will have an impact on policy
choices as shown by the feedback loop in Figure 2.1, suggesting that
the risk management process for addressing climate change is itera-
tive. The nature of this feedback can be illustrated by the following
examples. Individuals may be willing to invest in solar panels if they
are able to spread the upfront cost over time through a long-term
loan. Firms may be willing to promote new energy technologies that
provide social benefits with respect to climate change if they are
given a grant to assist them in their efforts. National governments
are more likely to implement carbon markets or international trea-
ties if they perceive the short-term benefits of these measures to be
greater than the perceived costs. Education and learning can play key
roles in how climate change is managed through a reconsideration
of policies for managing the risks and uncertainties associated with
climate change.
2.2 Metrics of uncertainty
and risk
The IPCC strives for a treatment of risk and uncertainty that is consis-
tent across all three Working Groups based the Guidance Note (GN)
for Lead Authors of the IPCC Fifth Assessment Report on Consistent
Treatment of Uncertainties (Mastrandrea et al., 2010). This section
summarizes key aspects of the GN that frames the discussion in this
chapter.
The GN indicates that author teams should evaluate the associated
evidence and agreement with respect to specific findings that involve
risk and uncertainty. The amount of evidence available can range from
small to large, and can vary in quality and consistency. The GN recom-
mends reporting the degree of certainty and / or uncertainty of a given
topic as a measure of the consensus or agreement across the scien-
tific community. Confidence expresses the extent to which the IPCC
authors do in fact support a key finding. If confidence is sufficiently
high, the GN suggests specifying the key finding in terms of probabil-
ity. The evaluation of evidence and degree of agreement of any key
finding is labelled a traceable account in the GN.
The GN also recommends taking a risk-management perspective by
stating that “sound decision making that anticipates, prepares for,
and responds to climate change depends on information about the
full range of possible consequences and associated probabilities.”
The GN also notes that, “low-probability outcomes can have signifi-
cant impacts, particularly when characterized by large magnitude, long
persistence, broad prevalence, and / or irreversibility.” For this reason,
the GN encourages the presentation of information on the extremes
of the probability distributions of key variables, reporting quantitative
estimates when possible and supplying qualitative assessments and
evaluations when appropriate.
2.3 Risk and uncertainty
in climate change
Since the publication of AR4, political scientists have documented the
many choices of climate policy and the range of interested parties con-
cerned with them (Moser, 2007; Andonova etal., 2009; Bulkeley, 2010;
Betsill and Hoffmann, 2011; Cabré, 2011; Hoffmann, 2011; Meckling,
2011; Victor, 2011).
There continues to be a concern about global targets for mean surface
temperature and GHG concentrations that are discussed in Chapter 6
of this report. This choice is normally made at the global level with
some regions, countries, and sub-national political regions setting their
own targets consistent with what they believe the global ones should
be. Policymakers at all levels of decision making face a second-order
set of choices as to how to achieve the desired targets. Choices in this
vein that are assessed in Chapters 7 12 of this report, include tran-
sition pathways for various drivers of emissions, such as fossil fuels
within the energy system, energy efficiency and energy-intensive
behavioural patterns, issues associated with land-use and spatial plan-
ning, and / or the emissions of non- CO
2
greenhouse gases.
The drivers influencing climate change policy options are discussed in
more detail in Chapters 13 16 of this report. These options include
information provision, economic instruments (taxes, subsidies, fines),
direct regulations and standards, and public investments. At the same
time, individuals, groups and firms decide what actions to take on their
own. These choices, some of which may be in response to governmen-
tal policy, include investments, lifestyle and behaviour.
Decisions for mitigating climate change are complemented by climate
adaptation options and reflect existing environmental trends and driv-
ers. The policy options are likely to be evaluated with a set of crite-
ria that include economic impacts and costs, equity and distributional
considerations, sustainable development, risks to individuals and soci-
ety and co-benefits. Many of these issues are discussed in Chapters 3
and 4.
2�3�1 Uncertainties that matter for climate
policy choices
The range and number of interested parties who are involved in cli-
mate policy choices have increased significantly in recent years. There
has been a widening of the governance forums within which climate
158158
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
policies and international agreements are negotiated at the global
level (Victor, 2011), across multiple networks within national gov-
ernments (Andonova etal., 2009; Hoffmann, 2011), and at the local,
regional and / or interest group level (Moser, 2007; Bulkeley, 2010). At
the same time, the number of different policy instruments under active
discussion has increased, from an initial focus on cap-and-trade and
carbon tax instruments (Betsill and Hoffmann, 2011; Hoffmann, 2011),
to feed-in tariffs or quotas for renewable energy (Wiser etal., 2005;
Mendonça, 2007), investments in research and development (Sagar
and van der Zwaan, 2006; De Coninck etal., 2008; Grubler and Riahi,
2010), and reform of intellectual property laws (Dechezleprêtre etal.,
2011; Percival and Miller, 2011).
Choices are sensitive to the degree of uncertainty with respect to a
set of parameters that are often of specific importance to particular
climate policy decisions. Here, and as shown in Figure 2.2, we group
these uncertainties into five broad classes, consistent with the
approach taken in Patt and Weber (2014):
• Climate responses to greenhouse gas (GHG) emissions, and their
associated impacts. The large number of key uncertainties with
respect to the climate system are discussed in Working GroupI
(WGI). There are even greater uncertainties with respect to the
impacts of changes in the climate system on humans and the eco-
logical system as well as their costs to society. These impacts are
assessed in WGII.
• Stocks and flows of carbon and other GHGs. The large uncertain-
ties with respect to both historical and current GHG sources and
sinks from energy use, industry, and land-use changes are assessed
in Chapter 5. Knowledge gaps make it especially difficult to esti-
mate how the flows of greenhouse gases will evolve in the future
under conditions of elevated atmospheric CO
2
concentrations and
their impact on climatic and ecological processes.
• Technological systems. The deployment of technologies is likely to
be the main driver of GHG emissions and a major driver of climate
vulnerability. Future deployment of new technologies will depend
on how their price, availability, and reliability evolve over time as a
result of technological learning. There are uncertainties as to how
fast the learning will take place, what policies can accelerate learn-
ing and the effects of accelerated learning on deployment rates of
new technologies. Technological deployment also depends on the
degree of public acceptance, which in turn is typically sensitive to
perceptions of health and safety risks.
• Market behaviour and regulatory actions. Public policies can create
incentives for private sector actors to alter their investment behav-
iour, often in the presence of other overlapping regulations. The
extent to which firms change their behaviour in response to the
policy, however, often depends on their expectations about other
highly uncertain market factors, such as fossil fuel prices. There are
also uncertainties concerning the macro-economic effects of the
aggregated behavioural changes. An additional factor influencing
the importance of any proposed or existing policy-driven incen-
tive is the likelihood with which regulations will be enacted and
enforced over the lifetime of firms’ investment cycles.
• Individual and firm perceptions. The choices undertaken by key
decision makers with respect to mitigation and adaptation mea-
sures are impacted by their perceptions of risk and uncertainties,
as well as their perceptions of the relevant costs and expected
benefits over time. Their decisions may also be influenced by the
actions undertaken by others.
Section 2.6 assesses the effects of uncertainties of these different
parameters on a wide range of policy choices, drawing from both
empirical studies and the modelling literature. The following three
examples illustrate how uncertainties in one or more of the above fac-
tors can influence choices between alternative options.
Example 1: Designing a regional emissions trading system (ETS). Over
the past decade, a number of political jurisdictions have designed and
implemented ETSs, with the European ETS being the one most stud-
ied. In designing the European system, policymakers took as their
starting point pre-defined emissions reduction targets. It was unclear
whether these targets would be met, due to uncertainties with respect
to national baseline emissions. The stocks and flows of greenhouse
gas emissions were partly determined by the uncertainty of the perfor-
mance of the technological systems that were deployed. Uncertainties
in market behaviour could also influence target prices and the number
of emissions permits allocated to different countries (Betsill and Hoff-
mann, 2011).
Example 2: Supporting scientific research into solar radiation manage-
ment (SRM). SRM may help avert potentially catastrophic temperature
increases, but may have other negative impacts with respect to global
and regional climatic conditions (Rasch etal., 2008). Research could
reduce the uncertainties as to these other consequences (Robock etal.,
2010). The decision to invest in specific research activities requires an
assessment as to what impact SRM will have on avoiding catastrophic
temperature increases. Temperature change will be sensitive to the
stocks and flows of greenhouse gases (GHG) and therefore to the
responses by key decision makers to the impacts of GHG emissions. The
decision to invest in specific research activities is likely to be influenced
by the perceived uncertainty in the actions undertaken by individuals
and firms (Blackstock and Long, 2010).
Example 3: Renting an apartment in the city versus buying a house
in the suburbs. When families and households face this choice, it is
likely to be driven by factors other than climate change concerns. The
decision, however, can have major consequences on CO
2
emissions as
well as on the impacts of climate change on future disasters such as
damage from flooding due to sea level rise. Hence, governments may
seek to influence these decisions as part of their portfolio of climate
change policies through measures such as land-use regulations or the
159159
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
pricing of local transportation options. The final choice is thus likely to
be sensitive to uncertainties in market behaviour as well as actions
undertaken by individuals and firms.
To add structure and clarity to the many uncertainties that different
actors face for different types of problems, we introduce a taxonomy
shown in Figure 2.2 that focuses on levels of decision making (the
rows) that range from international organizations to individuals and
households, and climate policy options (the columns) that include
long-term targets, transition pathways, policy instruments, resource
allocation and lifestyle options. The circles that overlay the cells in Fig-
ure 2.2 highlight the principal uncertainties relevant to decision-mak-
ing levels and climate policy choices that appear prominently in the
literature associated with particular policies. These are reviewed in
Section 2.6 of this chapter and in many of the following chapters of
WGIII. The literature appraises the effects of a wide range of uncertain-
ties, which we group according to the five types described above.
2�3�2 What is new on risk and uncertainty in
AR5
Chapter 2 in WGIII AR4 on risk and uncertainty, which also served as a
framing chapter, illuminated the relationship of risk and uncertainty to
decision making and reviewed the literature on catastrophic or abrupt
climate change and its irreversible nature. It examined three pillars for
dealing with uncertainties: precaution, risk hedging, and crisis preven-
tion and management. The report also summarized the debate in the
economic literature about the limits of cost-benefit analysis in situa-
tions of uncertainty.
Since the publication of AR4, a growing number of studies have con-
sidered additional sources of risk and uncertainties, such as regulatory
and technological risks, and examined the role they play in influenc-
ing climate policy. There is also growing awareness that risks in the
extremes or tail of the distribution make it problematic to rely on his-
torical averages. As the number of political jurisdictions implement-
ing climate policies has increased, there are now empirical findings to
supplement earlier model-based studies on the effects of such risks. At
the local level, adaptation studies using scenario-based methods have
been developed (ECLACS, 2011).
This chapter extends previous reports in four ways. First, rather than
focusing solely at the global level, this chapter expands climate-related
decisions to other levels of decision making as shown in Figure 2.2.
Second, compared to AR4, where judgment and choice were primar-
ily framed in rational-economic terms, this chapter reviews the psy-
chological and behavioural literature on perceptions and responses to
risk and uncertainty. Third, the chapter considers the pros and cons of
alternative methodologies and decision aids from the point of view of
practitioners. Finally, the chapter expands the scope of the challenges
associated with developing risk management strategies in relation to
Figure 2�2 | Taxonomy of levels of decision making and climate policy choices. Circles show type and extent of uncertainty sources as they are covered by the literature. Numbers in
brackets refer to sections where more information on these uncertainty sources can be found.
International
Agreement
National
Government
Local or Regional
Government or
Interest Group
Industry
or Firm
Household or
Individual
Long-Term
Targets
Transition
Pathway
Policy
Instrument
Resource
Allocation
Lifestyle and
Behavior
Climate Responses
and Associated Impacts
[Section 2.6.3.1]
Technological Systems
[2.6.3.2]
[2.6.4]
[2.6.5]
Stocks and Flows
of Carbon and GHGs
[2.6.4.3]
Market Behavior &
Regulatory Actions
[2.6.5]
Individual and
Firm Perceptions
[2.6.5.3]
[2.6.6]
Climate Policy Choices
Scale of
Action
160160
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
AR4 that requires reviewing a much larger body of published research.
To illustrate this point, the chapter references more than 50 publica-
tions on decision making under uncertainty with respect to integrated
assessment models (IAMs), the first time such a detailed examination
of this literature has been undertaken.
2.4 Risk perception and
responses to risk
and uncertainty
2�4�1 Considerations for design of climate
change risk reduction policies
When stakeholders are given information about mitigation and adap-
tation measures to reduce climate change risks, they make the fol-
lowing judgments and choices: How serious is the risk? Is any action
required? Which options are ruled out because the costs seem prohibi-
tive? Which option offers the greatest net expected benefits?
In designing such measures and in deciding how to present them to
stakeholders, one needs to recognize both the strengths and limita-
tions of decision makers at the different levels delineated in Figure 2.2.
Decision makers often have insufficient or imperfect knowledge about
climate risks, a deficit that can and needs to be addressed by better
data and public education. However, cognitive and motivational bar-
riers are equally or more important in this regard (Weber and Stern,
2011).
Normative models of choice described in Section 2.5 indicate how
decisions under risk and uncertainty should be made to achieve effi-
ciency and consistency, but these approaches do not characterize how
choices are actually made. Since decision makers have limitations in
their ability to process information and are boundedly rational (Simon,
1955), they often use simple heuristics and rules of thumb (Payne etal.,
1988). Their choices are guided not only by external reality (objective
outcomes and their likelihood) but also by the decision makers’ inter-
nal states (e. g., needs and goals) and their mental representation of
outcomes and likelihood, often shaped by previous experience. In other
words, a descriptive model of choice needs to consider cognitive and
motivational biases and decision rules as well as factors that are con-
sidered when engaging in deliberative thinking. Another complicating
factor is that when groups or organizations make decisions, there is the
potential for disagreement and conflict among individuals that may
require interpersonal and organizational facilitation by a third party.
Mitigation and adaptation decisions are shaped also by existing eco-
nomic and political institutional arrangements. Policy and market tools
for addressing climate change, such as insurance, may not be feasible
in developing countries that have no history of this type of protection;
however, this option may be viewed as desirable in a country with an
active insurance sector (see Box 2.1). Another important determinant
of decisions is the status quo, because there is a tendency to give more
weight to the negative impacts of undertaking change than the equiv-
alent positive impacts (Johnson etal., 2007). For example, proposing
a carbon tax to reduce GHG emissions may elicit much more concern
from affected stakeholders as to how this measure will impact on
their current activities than the expected climate change benefits from
reducing carbon emissions. Choices are also affected by cultural differ-
ences in values and needs (Maslow, 1954), in beliefs about the exis-
tence and causes of climate change (Leiserowitz etal., 2008), and in
the role of informal social networks for cushioning catastrophic losses
(Weber and Hsee, 1998). By considering actual judgment and choice
processes, policymakers can more accurately characterize the effective-
ness and acceptability of alternative mitigation policies and new tech-
nologies. Descriptive models also provide insights into ways of framing
mitigation or adaptation options so as to increase the likelihood that
desirable climate policy choices are adopted. Descriptive models, with
their broader assumptions about goals and processes, also allow for
the design of behavioural interventions that capitalize on motivations
such as equity and fairness.
2�4�2 Intuitive and deliberative judgment and
choice
The characterization of judgment and choice that distinguishes intui-
tive processes from deliberative processes builds on a large body of
cognitive psychology and behavioural decision research that can
be traced to William James (1878) in psychology and to Friedrich
Nietzsche (2008) and Martin Heidegger (1962) in philosophy. A recent
summary has been provided by Kahneman (2003; 2011) as detailed in
Table 2.1:
Table 2�1 | Intuitive and deliberative process characteristics.
Intuitive Thinking (System 1)
Operates automatically and quickly, with little or no effort and no voluntary control.
Uses simple and concrete associations, including emotional reactions or simple rules of
conduct that have been acquired by personal experience with events and their consequences.
Deliberative Thinking (System 2)
Initiates and executes effortful and intentional abstract cognitive
operations when these are seen as needed.
These cognitive operations include simple or complex computations or formal logic.
Even though the operations of these two types of processes do not
map cleanly onto distinct brain regions, and the two systems often
operate cooperatively and in parallel (Weber and Johnson, 2009), the
distinction between Systems 1 and 2 helps to clarify the tension in the
human mind between the automatic and largely involuntary processes
of intuitive decisions, versus the effortful and more deliberate pro-
cesses of analytic decisions (Kahneman, 2011).
161161
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
Many of the simplified decision rules that characterize human judg-
ment and choice under uncertainty utilize intuitive (System 1) pro-
cesses. Simplification is achieved by utilizing the experiences, expec-
tations, beliefs, and goals of the interested parties involved in the
decision. Such shortcuts require much less time and effort than a
more detailed analysis of the tradeoffs between options and often
leads to reasonable outcomes. If one takes into account the con-
straints on time and attention and processing capacity of decision
makers, these decisions may be the best we can do for many choices
under uncertainty (Simon, 1955). Intuitive processes are utilized not
only by the general public, but also by technical experts such as insur-
ers and regulators (Kunreuther etal., 2013c) and by groups and orga-
nizations (Cyert and March, 1963; Cohen etal., 1972; Barreto and
Patient, 2013).
Intuitive processes work well when decision makers have copious
data on the outcomes of different decisions and recent experience is
a meaningful guide for the future, as would be the case in station-
ary environments (Feltovich etal., 2006). These processes do not work
well, however, for low-probability high-consequence events for which
the decision maker has limited or no past experience (Weber, 2011).
In such situations, reliance on intuitive processes for making decisions
will most likely lead to maintaining the status quo and focusing on the
recent past. This suggests that intuitive decisions may be problematic
in dealing with climate change risks such as increased flooding and
storm surge due to sea level rise, or a surge in fossil fuel prices as
a result of an unexpected political conflict. These are risks for which
there is limited or no personal experience or historical data and con-
siderable disagreement and uncertainty among experts with respect to
their risk assessments (Taleb, 2007).
The formal models and tools that characterize deliberative (System 2)
thinking require stakeholders to make choices in a more abstract and
systematic manner. A deliberative process focuses on potential short-
and long-term consequences and their likelihoods, and evenly evalu-
ates the options under consideration, not favouring the status quo. For
the low-probability high-consequence situations for which decision
makers have limited experience with outcomes, alternative decision
frameworks that do not depend on precise specification of probabili-
ties should be considered in designing risk management strategies for
climate change (Charlesworth and Okereke, 2010; Kunreuther etal.,
2013a).
The remainder of this section is organized as follows. Section 2.4.3
describes some important consequences of the intuitive processes uti-
lized by individuals, groups, and organizations in making decisions.
The predicted effectiveness of economic or technological climate
change mitigation solutions typically presuppose rational delibera-
tive thinking and evaluation without considering how perceptions
and reactions to climate risks impose on these policy options. Sec-
tion 2.4.4 discusses biases and heuristics that suggest that individu-
als learn in ways that differ significantly from deliberative Bayesian
updating. Section 2.4.5 addresses how behaviour is affected by social
amplification of risk and considers the different levels of decision
making in Figure 2.2 by discussing the role of social norms, social
comparisons, and social networks in the choice process. Section 2.4.6
characterizes the general public’s perceptions of climate change risks
and uncertainty and their implications for communicating relevant
information.
Empirical evidence for the biases associated with climate change
response decisions triggered by intuitive processes exists mostly at
the level of the individual. As discussed in Sections 2.5 and 2.6, intui-
tive judgment and choice processes at other levels of decision making,
such as those specified in Figure 2.2, need to be acknowledged and
understood.
2�4�3 Consequences of intuitive decision
making
The behaviour of individuals are captured by descriptive models of
choice such as prospect theory (Kahneman and Tversky, 1979) for
decisions under risk and uncertainty and the beta-delta model (Laib-
son, 1997) for characterizing how future costs and benefits are evalu-
ated. While individual variation exists, the patterns of responding to
potential outcomes over time and the probabilities of their occur-
rence have an empirical foundation based on controlled experiments
and well-designed field studies examining the behaviour of technical
experts and the general public (Loewenstein and Elster, 1992; Cam-
erer, 2000).
2�4�3�1 Importance of the status quo
The tendency to maintain the current situation is a broadly observed
phenomenon in climate change response contexts (e. g., inertia in
switching to a non-carbon economy or in switching to cost-effective
energy efficient products) (Swim etal., 2011). Sticking with the current
state of affairs is the easy option, favoured by emotional responses in
situations of uncertainty (“better the devil you know than the devil
you don’t”), by many proverbs or rules (“when in doubt, do nothing”),
and observed biases in the accumulation of arguments for different
choice options (Weber etal., 2007). Overriding the status quo requires
commitment to change and effort (Fleming etal., 2010).
Loss aversion and reference points
Loss aversion is an important property that distinguishes prospect the-
ory (Tversky and Kahneman, 1992) from expected utility theory (von
Neumann and Morgenstern, 1944) by introducing a reference-depen-
dent valuation of outcomes, with a steeper slope for perceived losses
than for perceived gains. In other words, people experience more pain
from a loss than they get pleasure from an equivalent gain. The status
quo is often the relevant reference point that distinguishes outcomes
perceived as losses from those perceived as gains. Given loss aversion,
the potential negative consequences of moving away from the current
162162
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
state of affairs are weighted much more heavily than the potential
gains, often leading the decision maker not to take action. This behav-
iour is referred to as the status quo bias (Samuelson and Zeckhauser,
1988).
Loss aversion explains a broad range of decisions in controlled labora-
tory experiments and real world choices that deviate from the predic-
tions of rational models like expected utility theory (Camerer, 2000).
Letson etal. (2009) show that adapting to seasonal and inter-annual
climate variability in the Argentine Pampas by allocating land to dif-
ferent crops depends not only on existing institutional arrangements
(e. g., whether the farmer is renting the land or owns it), but also on
individual differences in farmers’ degree of loss aversion and risk
aversion. Greene etal. (2009) show that loss aversion combined with
uncertainty about future cost savings can explain why consumers fre-
quently appear to be unwilling to invest in energy-efficient technology
such as a more expensive but more fuel-efficient car that has posi-
tive expected utility. Weber and Johnson (2009) distinguish between
perceptions of risk, attitudes towards risk, and loss aversion that have
different determinants, but are characterized by a single ‘risk attitude’
parameter in expected utility models. Distinguishing and measuring
these psychologically distinct components of individual differences in
risk taking (e. g., by using prospect theory and adaptive ways of elicit-
ing its model parameters; Toubia etal., 2013) provides better targeted
entry points for policy interventions.
Loss aversion influences the choices of experienced decision makers
in high-stakes risky choice contexts, including professional financial
markets traders (Haigh and List, 2005) and professional golfers (Pope
and Schweitzer, 2011). Yet, other contexts fail to elicit loss aversion,
as evidenced by the failure of much of the global general public to be
alarmed by the prospect of climate change (Weber, 2006). In this and
other contexts, loss aversion does not arise because decision makers
are not emotionally involved (Loewenstein etal., 2001).
Use of framing and default options for the design of decision
aids and interventions
Descriptive models not only help explain behaviours that deviate from
the predictions of normative models of choice but also provide entry
points for the design of decision aids and interventions collectively
referred to as choice architecture, indicating that people’s choices
depend in part on the ways that possible outcomes of different
options are framed and presented (Thaler and Sunstein, 2008). Pros-
pect theory suggests that changing decision makers’ reference points
can impact on how they evaluate outcomes of different options and
hence their final choice. Patt and Zeckhauser (2000) show, for exam-
ple, how information about the status quo and other choice options
can be presented differently to create an action bias with respect to
addressing the climate change problem. More generally, choice archi-
tecture often involves changing the description of choice options and
the context of a decision to overcome the pitfalls of intuitive (System
1) processes without requiring decision makers to switch to effortful
(System 2) thinking (Thaler and Sunstein, 2008).
One important choice architecture tool comes in the form of behav-
ioural defaults, that is, recommended options that will be implemented
if no active decision is made (Johnson and Goldstein, 2013). Default
options serve as a reference point so that decision makers normally
stick with this option due to loss aversion (Johnson etal., 2007; Weber
etal., 2007). ‘Green’ energy defaults have been found to be very effec-
tive in lab studies involving choices between different lighting tech-
nologies (Dinner etal., 2011), suggesting that environmentally friendly
and cost-effective energy efficient technology will find greater deploy-
ment if it were to show up as the default option in building codes and
other regulatory contexts. Green defaults are desirable policy options
because they guide decision makers towards individual and social
welfare maximizing options without reducing choice autonomy. In a
field study, German utility customers adopted green energy defaults, a
passive choice that persisted over time and was not changed by price
feedback (Pichert and Katsikopoulos, 2008). Moser (2010) provides
other ways to frame climate change information and response options
in ways consistent with the communication goal and characteristics of
the audience.
2�4�3�2 Focus on the short term and the here-and-now
Finite attention and processing capacity imply that unaided intuitive
choices are restricted in their scope. This makes individuals susceptible
to different types of myopia or short-sightedness with respect to their
decisions on whether to invest in measures they would consider cost-
effective if they engaged in deliberative thinking (Weber and Johnson,
2009; Kunreuther etal., 2013b).
Present bias and quasi-hyperbolic time discounting
Normative models suggest that future costs and benefits should be
evaluated using an exponential discount function, that is, a constant
discount rate per time period (i. e., exponentially), where the discount
rate should reflect the decision maker’s opportunity cost of money (for
more details see Section 3.6.2). In reality, people discount future costs
or benefits much more sharply and at a non-constant rate (i. e., hyper-
bolically), so that delaying an immediate receipt of a benefit is viewed
much more negatively than if a similar delay occurs at a future point in
time (Loewenstein and Elster, 1992). Laibson (1997) characterized this
pattern by a quasi-hyperbolic discount function, with two parameters:
(1) present bias, i. e., a discount applied to all non-immediate outcomes
regardless how far into the future they occur, and (2) a rational dis-
counting parameter. The model retains much of the analytical tracta-
bility of exponential discounting, while capturing the key qualitative
feature of hyperbolic discounting.
Failure to invest in protective measures
In the management of climate-related natural hazards such as flood-
ing, an extensive empirical literature reveals that adoption rates of
protective measures by the general public are much lower than if indi-
viduals had engaged in deliberative thinking by making relevant trad-
eoffs between expected costs and benefits. Thus, few people living in
163163
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
flood prone areas in the United States voluntarily purchase flood insur-
ance, even when it is offered at highly subsidized premiums under the
National Flood Insurance Program (NFIP) (Kunreuther etal., 1978). In
the context of climate change mitigation, many efficient responses like
investments in household energy efficiency are not adopted because
decision makers focus unduly on the upfront costs of these measures
(due to hyperbolic discounting amplified by loss aversion) and weight
the future benefits of these investments less than predicted by norma-
tive models (see Sections 2.6.4.3 and 3.10). The failure of consumers
to buy fuel-efficient cars because of their higher upfront costs (Section
8.3.5) is another example of this behaviour.
At a country or community level, the upfront costs of mitigating CO
2
emissions or of building seawalls to reduce the effects of sea level rise
loom large due to loss aversion, while the uncertain and future ben-
efits of such actions are more heavily discounted than predicted by
normative models. Such accounting of present and future costs and
benefits on the part of consumers and policymakers might make it dif-
ficult for them to justify these investments today and arrive at long-
term sustainable decisions (Weber, 2013).
Focus on short-term goals
Krantz and Kunreuther (2007) emphasize the importance of goals
and plans as a basis for making decisions. In the context of climate
change, protective or mitigating actions often require sacrificing
short-term goals that are highly weighted in people’s choices in order
to meet more abstract, distant goals that are typically given very low
weight. A strong focus on short-term goals (e. g., immediate survival)
may have been helpful as humans evolved, but may have negative
consequences in the current environment where risks and challenges
are more complex and solutions to problems such as climate change
require a focus on long time horizons. Weber etal. (2007) succeeded
in drastically reducing people’s discounting of future rewards by
prompting them to first generate arguments for deferring consump-
tion, contrary to their natural inclination to focus initially on rationales
for immediate consumption. To deal with uncertainty about future
objective circumstances as well as subjective evaluations, one can
adopt multiple points of view (Jones and Preston, 2011) or multiple
frames of reference (De Boer et al., 2010); a generalization of the
IPCC’s scenario approach to an uncertain climate future is discussed
in Chapter 6.
Mental accounting as a protection against short-term focus
People often mentally set up separate ‘accounts’ for different classes
of expenditures and do not treat money as fungible between these
accounts (Thaler, 1999). Mental accounts for different expenditures
serve as effective budgeting and self-control devices for decision mak-
ers with limited processing capacity and self-control. A focus on short-
term needs and goals can easily deplete financial resources, leaving not
enough for long(er)-term goals. Placing a limit on short-term spending
prevents this from happening. But such a heuristic also has a down-
side by unduly limiting people’s willingness to invest in climate change
mitigation or adaptation measures (e. g., flood proofing or solar pan-
els) that exceed their allocated budget for this account, regardless of
future benefits. Such constraints (real or mental) often lead to the use
of lexicographic (rather than compensatory) choice processes, where
option sets are created or eliminated sequentially, based on a series of
criteria of decreasing importance (Payne etal., 1988).
Mental accounting at a nonfinancial level may also be responsible for
rebound effects of a more psychological nature, in addition to the eco-
nomically based rebound effects discussed in Section 8.3.5. Rebound
effects describe the increase in energy usage that sometimes fol-
lows improvements in household, vehicle, or appliance efficiency. For
example, households who weatherize their homes tend to increase
their thermostat settings during the winter afterwards, resulting in a
decrease in energy savings relative to what is technologically achiev-
able (Hirst etal., 1985). While rebound effects on average equal only
10 30 % of the achievable savings, and therefore do not cancel out
the benefits of efficiency upgrades (Ehrhardt-Martinez and Laitner,
2010), they are significant and may result from fixed mental accounts
that people have for environmentally responsible behaviour. Having
fulfilled their self-imposed quota by a particular action allows decision
makers to move on to other goals, a behaviour also sometimes referred
to as the single-action bias (Weber, 2006).
2�4�3�3 Aversion to risk, uncertainty, and ambiguity
Most people are averse to risk and to uncertainty and ambiguity when
making choices. More familiar options tend to be seen as less risky, all
other things being equal, and thus more likely to be selected (Figner
and Weber, 2011).
Certainty effect or uncertainty aversion
Prospect theory formalizes a regularity related to people’s perceptions
of certain versus probabilistic prospects. People overweight outcomes
they consider certain, relative to outcomes that are merely proba-
ble a phenomenon labelled the certainty effect (Kahneman and Tver-
sky, 1979). This frequently observed behaviour can explain why the
certain upfront costs of adaptation or mitigation actions are viewed as
unattractive when compared to the uncertain future benefits of under-
taking such actions (Kunreuther etal., 2013b).
Ambiguity aversion
Given the high degree of uncertainty or ambiguity in most forecasts
of future climate change impacts and the effects of different mitiga-
tion or adaptation strategies, it is important to consider not only deci-
sion makers’ risk attitudes, but also attitudes towards ambiguous out-
comes. The Ellsberg paradox (Ellsberg, 1961) revealed that, in addition
to being risk averse, most decision makers are also ambiguity averse,
that is, they prefer choice options with well-specified probabilities
over options where the probabilities are uncertain. Heath and Tversky
(1991) demonstrated, however, that ambiguity aversion is not present
when decision makers believe they have expertise in the domain of
choice. For example, in contrast to the many members of the general
164164
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
public who consider themselves to be experts in sports or the stock
market, relatively few people believe themselves to be highly compe-
tent in environmentally relevant technical domains such as the trad-
eoffs between hybrid electric versus conventional gasoline engines in
cars, so they are likely to be ambiguity averse. Farmers who feel less
competent with respect to their understanding of new technology are
more ambiguity averse and less likely to adopt farming innovations (in
Peru; Engle-Warnick and Laszlo, 2006; and in the USA; Barham et al.,
2014). With respect to the likelihood of extreme events, such as natural
disasters, insurers feel they do not have special expertise in estimating
the likelihood of these events so they also tend to be ambiguity averse
and set premiums that are considerably higher than if they had more
certainty with respect to the likelihood of their occurrence (Kunreuther
etal., 1993; Cabantous etal., 2011).
2�4�4 Learning
The ability to change expectations and behaviour in response to new
information is an important survival skill, especially in uncertain and
non-stationary environments. Bayesian updating characterizes learning
when one engages in deliberative thinking. Individuals who engage
in intuitive thinking are also highly responsive to new and especially
recent feedback and information, but treat the data differently than
that implied by Bayesian updating (Weber etal., 2004).
Availability bias and the role of salience
People’s intuitive assessment of the likelihood of an uncertain event
is often based on the ease with which instances of its occurrence can
be brought to mind, a mechanism called availability by Tversky and
Kahneman (1973). Sunstein (2006) discusses the use of the availabil-
ity heuristics in response to climate change risks and how it differs
among groups, cultures, and nations. Availability is strongly influenced
by recent personal experience and can lead to an underestimation of
low-probability events (e. g., typhoons, floods, or droughts) before they
occur, and their overestimation after an extreme event has occurred.
The resulting availability bias can explain why individuals first pur-
chase insurance after a disaster has occurred and cancel their policies
several years later, as observed for earthquake (Kunreuther et al., 1978)
and flood insurance (Michel-Kerjan et al., 2012). It is likely that most
of these individuals had not suffered any losses during this period
and considered the insurance to be a poor investment. It is difficult
to convince insured individuals that the best return on their policy is
no return at all. They should celebrate not having suffered a loss (Kun-
reuther etal., 2013c).
Linear thinking
A majority of people perceive climate in a linear fashion that reflects
two common biases (Sterman and Sweeney, 2007; Cronin etal., 2009;
Dutt and Gonzalez, 2011). First, people often rely on the correlation
heuristic, which means that people wrongly infer that an accumulation
(CO
2
concentration) follows the same path as the inflow (CO
2
emis-
sions). This implies that cutting emissions will quickly reduce the con-
centration and damages from climate change (Sterman and Sweeney,
2007). According to Dutt (2011) people who rely on this heuristic likely
demonstrate wait-and-see behaviour on policies that mitigate climate
change because they significantly underestimate the delay between
reductions in CO
2
emissions and in the CO
2
concentration. Sterman and
Sweeny (2007) show that people‘s wait-and-see behaviour on mitiga-
tion policies is also related to a second bias whereby people incorrectly
infer that atmospheric CO
2
concentration can be stabilized even when
emissions exceeds absorption.
Linear thinking also leads people to draw incorrect conclusions from
nonlinear metrics, like the miles-per-gallon (mpg) ratings of vehicles’
gasoline consumption in North America (Larrick and Soll, 2008). When
given a choice between upgrading to a 15-mpg car from a 12-mpg car,
or to a 50-mpg car from a 29-mpg car, most people choose the latter
option. However, for 100 miles driven under both options, it is easily
shown that the first upgrade option saves more fuel (1.6 gallons for
every 100 miles driven) than the second upgrade option (1.4 gallons
for every 100 miles driven).
Effects of personal experience
Learning from personal experience is well predicted by reinforcement
learning models (Weber etal., 2004). Such models describe and predict
why the general public is less concerned about low-probability high-
impact climate risks than climate scientists would suggest is warranted
by the evidence (Gonzalez and Dutt, 2011). These learning models also
capture the volatility of the public’s concern about climate change
over time, for example in reaction to the personal experience of local
weather abnormalities (an abnormal cold spell or heat wave) that have
been shown to influence belief in climate change (Li etal., 2011).
Most people do not differentiate very carefully between weather, cli-
mate (average weather over time), and climate variability (variations
in weather over time). People confound climate and weather in part
because they have personal experience with weather and weather
abnormalities but little experience with climate change, an abstract
statistical concept. They thus utilize weather events in making judg-
ments about climate change (Whitmarsh, 2008). This confusion has
been observed in countries as diverse as the United States (Bostrom
et al., 1994; Cullen, 2010) and Ethiopia (BBC World Service Trust,
2009).
Personal experience can differ between individuals as a function of
their location, history, and / or socio-economic circumstances (Figner
and Weber, 2011). Greater familiarity with climate risks, unless accom-
panied by alarming negative consequences, could actually lead to a
reduction rather than an increase in the perceptions of its riskiness
(Kloeckner, 2011). On the other hand, people’s experience can make
climate a more salient issue. For example, changes in the timing and
extent of freezing and melting (and associated effects on sea ice, flora,
and fauna) have been experienced since the 1990s in the American
and Canadian Arctic and especially indigenous communities (Laidler,
2006), leading to increased concern with climate change because tra-
165165
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
ditional prediction mechanisms no longer can explain these phenom-
ena (Turner and Clifton, 2009).
People’s expectations of change (or stability) in climate variables also
affect their ability to detect trends in probabilistic environments. For
instance, farmers in Illinois were asked to recall growing season tem-
perature or precipitation statistics for seven preceding years. Farmers
who believed that their region was affected by climate change recalled
precipitation and temperature trends consistent with this expectation,
whereas farmers who believed in a constant climate, recalled precipita-
tions and temperatures consistent with that belief (Weber, 1997). Rec-
ognizing that beliefs shape perception and memory provides insight
into why climate change expectations and concerns vary between seg-
ments of the US population with different political ideologies (Leise-
rowitz etal., 2008).
The evidence is mixed when we examine whether individuals learn
from past experience with respect to investing in adaptation or miti-
gation measures that are likely to be cost-effective. Even after the
devastating 2004 and 2005 hurricane seasons in the United States, a
large number of residents in high-risk areas had still not invested in
relatively inexpensive loss-reduction measures, nor had they under-
taken emergency preparedness measures (Goodnough, 2006). Surveys
conducted in Alaska and Florida, regions where residents have been
exposed more regularly to physical evidence of climate change, show
greater concern and willingness to take action (ACI, 2004; Leiserowitz
and Broad, 2008; Mozumder etal., 2011).
A recent study assessed perceptions and beliefs about climate change of
a representative sample of the Britain public (some of whom had expe-
rienced recent flooding in their local area). It also asked whether they
would reduce personal energy use to reduce greenhouse gas emission
(Spence etal., 2011). Concern about climate change and willingness to
take action was greater in the group of residents who had experienced
recent flooding. Even though the flooding was only a single and local
data point, this group also reported less uncertainty about whether cli-
mate change was really happening than those who did not experience
flooding recently, illustrating the strong influence of personal experi-
ence. Other studies fail to find a direct effect of personal experience with
flooding generating concern about climate risks (Whitmarsh, 2008).
Some researchers find that personal experience with ill health from air
pollution affects perceptions of and behavioural responses to climate
risks (Bord etal., 2000; Whitmarsh, 2008), with the negative effects
from air pollution creating stronger pro-environmental values. Myers
etal. (2012) looked at the role of experiential learning versus moti-
vated reasoning among highly engaged individuals and those less
engaged in the issue of climate change. Low-engaged individuals
were more likely to be influenced by their perceived personal experi-
ence of climate change than by their prior beliefs, while those highly
engaged in the issue (on both sides of the climate issue) were more
likely to interpret their perceived personal experience in a manner that
strengthens their pre-existing beliefs.
Indigenous climate change knowledge contributions from Africa
(Orlove et al., 2010), the Arctic (Gearheard et al., 2009), Australia
(Green et al., 2010), or the Pacific Islands (Lefale, 2010), derive from
accumulated and transmitted experience and focus mostly on pre-
dicting seasonal or interannual climate variability. Indigenous knowl-
edge can supplement scientific knowledge in geographic areas with
a paucity of data (Green and Raygorodetsky, 2010) and can guide
knowledge generation that reduces uncertainty in areas that matter
for human responses (ACI, 2004). Traditional ecological knowledge is
embedded in value-institutions and belief systems related to historical
modes of experimentation and is transferred from generation to gen-
eration (Pierotti, 2011).
Underweighting of probabilities and threshold models of
choice
The probability weighting function of prospect theory indicates that
low probabilities tend to be overweighted relative to their objective
probability unless they are perceived as being so low that they are
ignored because they are below the decision maker’s threshold level
of concern. Prior to a disaster, people often perceive the likelihood of
catastrophic events occurring as below their threshold level of con-
cern, a form of intuitive thinking in the sense that one doesn’t have
to reflect on the consequences of a catastrophic event (Camerer and
Kunreuther, 1989). The need to take steps today to deal with future cli-
mate change presents a challenge to individuals who are myopic. They
are likely to deal with this challenge by using a threshold model that
does not require any action for risks below this level. The problem is
compounded by the inability of individuals to distinguish between low
likelihoods that differ by one or even two orders of magnitude (e. g.,
between 1 in 100 and 1 in 10,000) (Kunreuther etal., 2001).
2�4�5 Linkages between different levels of
decision making
Social amplification of risk
Hazards interact with psychological, social, institutional, and cultural
processes in ways that may amplify or attenuate public responses to
the risk or risk event by generating emotional responses and other
biases associated with intuitive thinking. Amplification may occur
when scientists, news media, cultural groups, interpersonal networks,
and other forms of communication provide risk information. The ampli-
fied risk leads to behavioural responses, which, in turn, may result in
secondary impacts such as the stigmatization of a place that has expe-
rienced an adverse event (Kasperson etal., 1988; Flynn etal., 2001).
The general public’s overall concern about climate change is influ-
enced, in part, by the amount of media coverage the issue receives as
well as the personal and collective experience of extreme weather in a
given place (Leiserowitz etal., 2012; Brulle etal., 2012).
Social norms and social comparisons
Individuals’ choices are often influenced by other people’s behaviour,
especially under conditions of uncertainty. Adherence to formal rules
166166
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
(e. g., standard operating procedures or best practices in organizations)
or informal rules of conduct is an important way in which we intui-
tively decide between different courses of action (Weber and Linde-
mann, 2007). “When in doubt, copy what the majority is doing” is not
a bad rule to follow in many situations, as choices adopted by oth-
ers are assumed to be beneficial and safe (Weber, 2013). In fact, such
social imitation can lead to social norms. Section 3.10.2 describes the
effects of social norms in greater detail. Goldstein etal. (2008) demon-
strate the effectiveness of providing descriptive norms (“this is what
most people do”) versus injunctive norms (“this is what you should
be doing”) to reduce energy use in US hotels. The application of social
norms to encourage investment in energy efficient products and tech-
nology is discussed in Section 2.6.5.3.
Social comparisons are another effective way to evaluate and learn
about the quality of obtained outcomes (Weber, 2004). It helps, for
example, to compare one’s own energy consumption to that of neigh-
bours in similar-sized apartments or houses to see how effective
efforts at energy conservation have been. Such non-price interventions
can substantially change consumer behaviour, with effects equivalent
to that of a short-run electricity price increase of 11 % to 20 % (Alcott,
2011). Social comparisons, imitation, and norms may be necessary to
bring about lifestyle changes that are identified in Chapter 9 as reduc-
ing GHG emissions from the current levels (Sanquist etal., 2012).
Social learning and cultural transmission
Section 9.3.10 suggests that indigenous building practices in many
parts of the world provide important lessons for affordable low-
energy housing design and that developed countries can learn from
traditional building practices, transmitted over generations, the social-
scale equivalent of ‘intuitive’ processing and learning at the individual
level.
Risk protection by formal (eg�, insurance) and informal
institutions (eg�, social networks)
Depending on their cultural and institutional context, people can pro-
tect themselves against worst-case and / or potentially catastrophic
economic outcomes either by purchasing insurance (Kunreuther etal.,
2013c) or by developing social networks that will help bail them out or
assist them in the recovery process (Weber and Hsee, 1998). Individual-
ist cultures favour formal insurance contracts, whereas collectivist soci-
eties make more use of informal mutual insurance via social networks.
This distinction between risk protection by either formal or informal
means exists at the individual level and also at the firm level, e. g., the
chaebols in Korea or the keiretsus in Japan (Gilson and Roe, 1993).
Impact of uncertainty on coordination and competition
Adaptation and especially mitigation responses require coordination
and cooperation between individuals, groups, or countries for many
of the choices associated with climate change. The possible outcomes
often can be viewed as a game between players who are concerned
with their own payoffs but who may still be mindful of social goals and
objectives. In this sense they can be viewed in the context of a pris-
oners’ dilemma (PD) or social dilemma. Recent experimental research
on two-person PD games reveals that individuals are more likely to
be cooperative when payoffs are deterministic than when the out-
comes are probabilistic. A key factor explaining this difference is that
in a deterministic PD game, the losses of both persons will always be
greater when they both do not cooperate than when they do. When
outcomes are probabilistic there is some chance that the losses will be
smaller when both parties do not cooperate than when they do, even
though the expected losses to both players will be greater if they both
decide not to cooperate than if they both cooperate (Kunreuther etal.,
2009).
In a related set of experiments, Gong etal. (2009) found that groups
are less cooperative than individuals in a two-person deterministic PD
game; however, in a stochastic PD game, where defection increased
uncertainty for both players, groups became more cooperative than
they were in a deterministic PD game and more cooperative than indi-
viduals in the stochastic PD game. These findings have relevance to
behaviour with respect to climate change where future outcomes of
specific policies are uncertain. Consider decisions made by groups of
individuals, such as when delegations from countries are negotiating
at the Conference of Parties (COP) to make commitments for reduc-
ing GHG emissions where the impacts on climate change are uncer-
tain. These findings suggest that there is likely to be more cooperation
between governmental delegations than if each country was repre-
sented by a single decision maker.
Cooperation also plays a crucial role in international climate agree-
ments. There is a growing body of experimental literature that looks at
individuals’ cooperation when there is uncertainty associated with oth-
ers adopting climate change mitigation measures. Tavoni etal. (2011)
found that communication across individuals improves the likelihood
of cooperation. Milinski etal. (2008) observed that the higher the risky
losses associated with the failure to cooperate in the provision of a
public good, the higher the likelihood of cooperation. If the target for
reducing CO
2
is uncertain, Barrett and Dannenberg (2012) show in an
experimental setting that cooperation is less likely than if the target is
well specified.
2�4�6 Perceptions of climate change risk and
uncertainty
Empirical social science research shows that the perceptions of climate
change risks and uncertainties depend not only on external reality but
also on the observers’ internal states, needs, and the cognitive and
emotional processes that characterize intuitive thinking. Psychological
research has documented the prevalence of affective processes in the
intuitive assessment of risk, depicting them as essentially effort-free
inputs that orient and motivate adaptive behaviour, especially under
conditions of uncertainty that are informed and shaped by personal
experience over time (Finucane etal., 2000; Loewenstein etal., 2001;
Peters etal., 2006).
167167
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
Two important psychological risk dimensions have been shown to
influence people’s intuitive perceptions of health and safety risks
across numerous studies in multiple countries (Slovic, 1987). The first
factor, ‘dread risk’, captures emotional reactions to hazards like nuclear
reactor accidents, or nerve gas accidents, that is, things that make peo-
ple anxious because of a perceived lack of control over exposure to
the risks and because consequences may be catastrophic. The second
factor, ‘unknown risk‘, refers to the degree to which a risk (e. g., DNA
technology) is perceived as new, with unforeseeable consequences
and with exposures not easily detectable.
Perceptions of the risks associated with a given event or hazard are
also strongly influenced by personal experience and can therefore dif-
fer between individuals as a function of their location, history, and / or
socio-economic circumstances (see Box 2.1) (Figner and Weber, 2011).
Whereas personal exposure to adverse consequences increases fear
and perceptions of risk, familiarity with a risk can lower perceptions
of its riskiness unless it is accompanied by alarming negative conse-
quences (Kloeckner, 2011). Seeing climate change only as a simple
and gradual change from current to future average temperatures and
precipitation may make it seem controllable the non-immediacy
of the danger seems to provide time to plan and execute protective
responses (Weber, 2006). These factors suggest that laypersons differ
in their perception of climate risks more than experts who engage in
deliberative thinking and estimate the likelihood and consequences of
climate change utilizing scientific data.
Impact of uncertainties in communicating risk
If the uncertainties associated with climate change and its future
impact on the physical and social system are not communicated accu-
rately, the general public may misperceive them (Corner and Hahn,
2009). Krosnick etal. (2006) found that perceptions of the seriousness
of global warming as a national issue in the United States depended on
the degree of certainty of respondents as to whether global warming is
Box 2�1 | Challenges facing developing countries
One of the key findings on developing countries is that non-state
actors such as tribes, clans, castes, or guilds may be of substantial
influence on how climate policy choices are made and diffused
rather than having the locus of decision making at the level of the
individual or governmental unit. For instance, a farming tribe / caste
may address the climate risks and uncertainties faced by their com-
munity and opt for a system of crop rotation to retain soil fertil-
ity or shift cultivation to preserve the nutritious state of farmlands.
Research in developing countries in Africa has shown that people
may understand probabilistic information better when it is pre-
sented in a group where members have a chance to discuss it (Patt
et al., 2005; Roncoli, 2006). This underscores why the risks and
uncertainty associated with climate change has shifted governmen-
tal responsibility to non-state actors (Rayner, 2007).
In this context, methodologies and decision aids used in individual-
centred western societies for making choices that rely on uncertain
probabilities and uncertain outcomes may not apply to develop-
ing countries. Furthermore, methodologies, such as expected utility
theory, assume an individual decision maker whereas in develop-
ing countries, decisions are often made by clans or tribes. In addi-
tion, tools such as cost-benefit analysis, cost-effectiveness analysis
and robust decision making may not always be relevant for devel-
oping countries since decisions are often based on social norms,
traditions, and customs
The adverse effects of climate change on food, water, security,
and incidences of temperature-influenced diseases (Shah and Lele,
2011), are further fuelled by a general lack of awareness about
climate change in developing countries (UNDP, 2007); conse-
quently, policymakers in these countries support a wait-and-see
attitude toward climate change (Dutt, 2011). Resource allocation
and investment constraints may also lead policy-makers to post-
pone policy decisions to deal with climate change, as is the case
with respect to integration of future energy systems in small island
states (UNFCCC, 2007). The delay may prevent opportunities for
learning and increase future vulnerabilities. It may also lock in
countries into infrastructure and technologies that may be difficult
to alter.
The tension between short- and long-term priorities in low income
countries is often accentuated by uncertainties in political culture
and regulatory policies (Rayner, 1993). This may lead to policies
that are flawed in design and / or implementation or those that
have unintended negative consequences. For example, subsidies
for clean fuels such as liquefied petroleum gas (LPG) in a country
like India often do not reach their intended beneficiaries (the poor),
and at the same time add a large burden to the exchequer (Gov-
ernment of India, Ministry of Finance, 2012; IISD, 2012).
Other institutional and governance factors impede effective cli-
mate change risk management in developing countries. These
include lack of experience with insurance (Patt etal., 2010), dearth
of data, and analytical capacity. A more transparent and effec-
tive civil service would also be helpful, for instance in stimulating
investments in renewable energy generation capacities (Komen-
dantova et al., 2012). Financial constraints suggest the impor-
tance of international assistance and private sector contribution to
implement adaptation and mitigation strategies for dealing with
climate change in developing countries.
168168
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
occurring and will have negative consequences coupled with their belief
that humans are causing the problem and have the ability to solve it.
Accurately communicating the degree of uncertainty in both climate
risks and policy responses is therefore a critically important challenge
for climate scientists and policymakers (Pidgeon and Fischhoff, 2011).
Roser-Renouf etal. (2011), building upon the work of Krosnick etal.
(2006), apply social cognitive theory to develop a model of climate
advocacy to increase the attention given to climate change in the spirit
of social amplification of risk. They found that campaigns looking to
increase the number of citizens contacting elected officials to advocate
climate policy action should focus on increasing the belief that global
warming is real, human-caused, a serious risk, and solvable. These four
key elements, coupled with the understanding that there is strong sci-
entific agreement on global warming (Ding etal., 2011), are likely to
build issue involvement and support for action to reduce the impacts
of climate change.
The significant time lags within the climate system and a focus on
short-term outcomes lead many people to believe global warming
will have only moderately negative impacts. This view is reinforced
because adverse consequences are currently experienced only in some
regions of the world or are not easily attributed to climate change. For
example, despite the fact that “climate change currently contributes to
the global burden of disease and premature deaths” (IPCC, 2007) rela-
tively few people make the connection between climate change and
human health risks.
One challenge is how to facilitate correct inferences about the role of
climate change as a function of extreme event frequency and sever-
ity. Many parts of the world have seen increases in the frequency and
magnitude of heat waves and heavy precipitation events (IPCC, 2012).
In the United States, a large majority of Americans believe that climate
change exacerbated extreme weather events (Leiserowitz etal., 2012).
That said, the perception that the impact of climate change is neither
immediate nor local persists (Leiserowitz etal., 2008), leading many
to think it rational to advocate a wait-and-see approach to emissions
reductions (Sterman, 2008; Dutt and Gonzalez, 2013).
Differences in education and numeracy
Individual and group differences in education and training and the
resulting different cognitive and affective processes have additional
implications for risk communication. It may help to supplement the
use of words to characterize the likelihood of an outcome recom-
mended by the current IPCC Guidance Note (GN) with numeric prob-
ability ranges (Budescu etal., 2009). Patt and Dessai (2005) show that
in the IPCC Third Assessment Report (TAR), words that characterized
numerical probabilities were interpreted by decision makers in incon-
sistent and often context-specific ways, a phenomenon with a long his-
tory in cognitive psychology (Wallsten etal., 1986; Weber and Hilton,
1990). These context-specific interpretations of probability words are
deeply rooted, as evidenced by the fact that the likelihood of using
the intended interpretation of TAR probability words did not differ with
level of expertise (attendees of a UN COP conference versus students)
or as a function of whether respondents had read the TAR instructions
that specify how the probability words characterized numerical prob-
abilities (Patt and Dessai, 2005).
Numeracy, the ability to reason with numbers and other mathemati-
cal concepts, is a particularly important individual and group differ-
ence in this context as it has implications for the presentation of likeli-
hood information using either numbers (for example, 90 %) or words
(for example, “very likely” or “likely”) or different graphs or diagrams
(Peters etal., 2006; Mastrandrea etal., 2011). Using personal experi-
ence with climate variables has been shown to be effective in com-
municating the impact of probabilities (e. g., of below-, about-, and
above-normal rainfall in an El Ni
~
no year) to decision makers with low
levels of numeracy, for example subsistence farmers in Zimbabwe (Patt
etal., 2005).
2.5 Tools and decision
aids for analysing
uncertainty and risk
This section examines how more formal approaches can assist deci-
sion makers in engaging in more deliberative thinking with respect to
climate change policies when faced with the risks and uncertainties
characterized in Section 2.3.
2�5�1 Expected utility theory
Expected utility [E(U)] theory (Ramsey, 1926; von Neumann and Mor-
genstern, 1944; Savage, 1954); remains the standard approach for pro-
viding normative guidelines against which other theories of individual
decision making under risk and uncertainty are benchmarked. Accord-
ing to the E(U) model, the solution to a decision problem under uncer-
tainty is reached by the following four steps:
1. Define a set of possible decision alternatives.
2. Quantify uncertainties on possible states of the world.
3. Value possible outcomes of the decision alternatives as utilities.
4. Choose the alternative with the highest expected utility.
This section clarifies the applicability of expected utility theory to the
climate change problem, highlighting its potentials and limitations.
2�5�1�1 Elements of the theory
E(U) theory is based on a set of axioms that are claimed to have nor-
mative rather than descriptive validity. Based on these axioms, a per-
169169
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
son’s subjective probability and utility function can be determined by
observing preferences in structured choice situations. These axioms
have been debated, strengthened, and relaxed by economists, psy-
chologists, and other social scientists over the years. The axioms have
been challenged by controlled laboratory experiments and field stud-
ies discussed in Section 2.4 but they remain the basis for parsing deci-
sion problems and recommending options that maximize expected
utility.
2�5�1�2 How can expected utility improve decision
making?
E(U) theory provides guidelines for individual choice, such as a farmer
deciding what crops to plant or an entrepreneur deciding whether to
invest in wind technology. These decision makers would apply E(U) the-
ory by following the four steps above. The perceptions and responses
to risk and uncertainty discussed in Section 2.5 provide a rationale
for undertaking deliberative thinking before making final choices.
More specifically, a structured approach, such as the E(U) model, can
reduce the impact of probabilistic biases and simplified decision rules
that characterize intuitive thinking. At the same time, the limitations
of E(U) must be clearly understood, as the procedures for determining
an optimal choice do not capture the full range of information about
outcomes and their risks and uncertainties.
Subjective versus objective probability
In the standard E(U) model, each individual has his / her own subjec-
tive probability estimates. When there is uncertainty on the scientific
evidence, experts’ probability estimates may diverge from each other,
sometimes significantly. With respect to climate change, observed rela-
tive frequencies are always preferred when suitable sets of observa-
tions are accessible. When these data are not available, one may want
to utilize structured expert judgment for quantifying uncertainty (see
Section 2.5.7).
Individual versus social choice
In applying E(U) theory to problems of social choice, a number of
issues arise. Condorcet’s voting paradox shows that groups of ratio-
nal individuals deciding by majority rule do not exhibit rational prefer-
ences. Using a social utility or social welfare function to determine an
optimal course of action for society requires some method of mea-
suring society’s preferences. In the absence of these data the social
choice problem is not a simple exercise of maximizing expected utility.
In this case, a plurality of approaches involving different aggregations
of individual utilities and probabilities may best aid decision makers.
The basis and use of the social welfare function are discussed in Sec-
tion 3.4.6.
Normative versus descriptive
As noted above, the rationality axioms of E(U) are claimed to have
normative as opposed to descriptive validity. The paradoxes of Allais
(1953) and Ellsberg (1961) reveal choice behaviour incompatible
with E(U); whether this requires modifications of the normative the-
ory is a subject of debate. McCrimmon (1968) found that business
executives willingly corrected violations of the axioms when they
were made aware of them. Other authors (Kahneman and Tversky,
1979; Schmeidler, 1989; Quiggin, 1993; Wakker, 2010) account for
such paradoxical choice behaviour by transforming the probabilities
of outcomes into decision weight probabilities that play the role of
likelihood in computing optimal choices but do not obey the laws
of probability. However, Wakker (2010, p.350) notes that decision
weighting fails to describe some empirically observed behavioural
patterns.
2�5�2 Decision analysis
2�5�2�1 Elements of the theory
Decision analysis is a formal approach for choosing between alterna-
tives under conditions of risk and uncertainty. The foundations of deci-
sion analysis are provided by the axioms of expected utility theory. The
methodology for choosing between alternatives consists of the follow-
ing elements that are described in more detail in Keeney (1993):
1. Structure the decision problem by generating alternatives and
specifying values and objectives or criteria that are important to
the decision maker.
2. Assess the possible impacts of different alternatives by determin-
ing the set of possible consequences and the probability of each
occurring.
3. Determine preferences of the relevant decision maker by develop-
ing an objective function that considers attitudes toward risk and
aggregates the weighted objectives.
4. Evaluate and compare alternatives by computing the expected util-
ity associated with each alternative. The alternative with the high-
est expected utility is the most preferred one.
To illustrate the application of decision analysis, consider a homeowner
that is considering whether to invest in energy efficient technology as
part of their lifestyle options as depicted in Figure 2.2:
1. The person focuses on two alternatives: (A1) Maintain the status
quo, and (A2) Invest in solar panels, and has two objectives: (O1)
Minimize cost, and (O2) Assist in reducing global warming.
2. The homeowner would then determine the impacts of A1 and A2
on the objectives O1 and O2 given the risks and uncertainties
associated with the impact of climate change on energy usage as
well as the price of energy.
3. The homeowner would then consider his or her attitude toward
risks and then combine O1 and O2 into a multiattribute utility
function.
4. The homeowner would then compare the expected utility of A1
and A2, choosing the one that had the highest expected utility.
170170
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
2�5�2�2 How can decision analysis improve
decision
making?
Decision analysis enables one to undertake sensitivity analyses with
respect to the uncertainties associated with the various consequences
and to different value structures. Suppose alternative A1 had the high-
est expected utility. The homeowner could determine when the deci-
sion to invest in solar panels would be preferred to maintaining the
status quo by asking questions such as:
• What would the minimum annual savings in energy expenses have
to be over the next 10 years to justify investing in solar panels?
• What is the fewest number of years one would have to reside in
the house to justify investing in solar panels?
• What impact will different levels of global warming have on the
expected costs of energy over the next 10 years for the home-
owner to want to invest in solar panels?
• How will changing the relative weights placed on minimizing cost
(O1) and assisting in reducing global warming (O2) affect the
expected utility of A1 and A2?
2�5�3 Cost-benefit analysis
2�5�3�1 Elements of the theory
Cost-benefit analysis (CBA) compares the costs and benefits of differ-
ent alternatives with the broad purpose of facilitating more efficient
allocation of society’s resources. When applied to government deci-
sions, CBA can indicate the alternative that has the highest social net
present value based on a discount rate, normally constant over time,
that converts future benefits and costs to their present values (Board-
man et al., 2005; see also the extensive discussion in Section 3.6).
Social, rather than private, costs and benefits are compared, including
those affecting future generations (Brent, 2006). In this regard, bene-
fits across individuals are assumed to be additive. Distributional issues
may be addressed by putting different weights on specific groups to
reflect their relative importance. Under conditions of risk and uncer-
tainty, one determines expected costs and benefits by weighting out-
comes by their likelihoods of occurrence. In this sense, the analysis is
similar to expected utility theory and decision analysis discussed in
Sections 2.5.1 and 2.5.2.
CBA can be extremely useful when dealing with well-defined problems
that involve a limited number of actors who make choices among dif-
ferent mitigation or adaptation options. For example, a region could
examine the benefits and costs over the next fifty years of building
levees to reduce the likelihood and consequences of flooding given
projected sea level rise due to climate change.
CBA can also provide a framework for defining a range of global
long-term targets on which to base negotiations across countries
(see for example Stern, 2007). However, CBA faces major challenges
when defining the optimal level of global mitigation actions for the
following three reasons: (1) the need to determine and aggregate
individual welfare, (2) the presence of distributional and intertempo-
ral issues, and (3) the difficulty in assigning probabilities to uncertain
climate change impacts. The limits of CBA in the context of climate
change are discussed at length in Sections 3.6 and 3.9. The discus-
sion that follows focuses on challenges posed by risk and uncer-
tainty.
2�5�3�2 How can CBA improve decision making?
Cost-benefit analysis assumes that the decision maker(s) will even-
tually choose between well-specified alternatives. To illustrate this
point, consider a region that is considering measures that coastal vil-
lages in hazard-prone areas can undertake to reduce future flood risks
that are expected to increase in part due to sea level rise. The different
options range from building a levee (at the community level) to pro-
viding low interest loans to encourage residents and businesses in the
community to invest in adaptation measures to reduce future damage
to their property (at the level of an individual or household).
Some heuristics and resulting biases discussed in the context of
expected utility theory also apply to cost-benefit analysis under uncer-
tainty. For example, the key decision maker, the mayor, may utilize a
threshold model of choice by assuming that the region will not be
subject to flooding because there have been no floods or hurricanes
during the past 25 years. By relying solely on intuitive processes there
would be no way to correct this behaviour until the next disaster
occurred, at which time the mayor would belatedly want to protect the
community. The mayor and his advisors may also focus on short-time
horizons, and hence do not wish to incur the high upfront costs associ-
ated with building flood protection measures such as dams or levees.
They are unconvinced that that such an investment will bring signifi-
cant enough benefits over the first few years when these city officials
are likely to be held accountable for the expenditures associated with
a decision to go forward on the project.
Cost-benefit analysis can highlight the importance of considering
the likelihood of events over time and the need to discount impacts
exponentially rather than hyperbolically, so that future time periods
are given more weight in the decision process. In addition, CBA can
highlight the tradeoffs between efficient resource allocation and distri-
butional issues as a function of the relative weights assigned to differ-
ent stakeholders (e. g., low income and well-to-do households in flood
prone areas).
2�5�3�3 Advantages and limitations of CBA
The main advantage of CBA in the context of climate change is that it is
internally coherent and based on the axioms of expected utility theory.
171171
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
As the prices used to aggregate costs and benefits are the outcomes
of market activity, CBA is, at least in principle, a tool reflecting people’s
preferences. Although this is one of the main arguments in favour of
CBA (Tol, 2003), this line of reasoning can also be the basis for rec-
ommending that this approach not be employed for making choices if
market prices are unavailable. Indeed, many impacts associated with
climate change are not valued in any market and are therefore hard to
measure in monetary terms. Omitting these impacts distorts the cost-
benefit relationship.
Several ethical and methodological critiques have been put forward
with respect to the application of CBA to climate policy (Charlesworth
and Okereke, 2010; Caney, 2011). For example, the uncertainty sur-
rounding the potential impacts of climate change, including possible
irreversible and catastrophic effects on ecosystems, and their asym-
metric distribution around the planet, suggests CBA may be inappro-
priate for assessing optimal responses to climate change in these cir-
cumstances.
A strong and recurrent argument against CBA (Azar and Lindgren,
2003; Tol, 2003; Weitzman, 2009, 2011) relates to its failure in dealing
with infinite (negative) expected utilities arising from low-probability
catastrophic events often referred to as ‘fat tails’. In these situations,
CBA is unable to produce meaningful results, and thus more robust
techniques are required. The debate concerning whether fat tails are
indeed relevant to the problem at hand is still unsettled (see for exam-
ple Pindyck, 2011). Box 3.9 in Chapter 3 addresses the fat tail problem
and suggests the
importance of understanding the impacts associated
with low probability, high impact climate change scenarios in evaluat-
ing alternative mitigation strategies.
One way to address the fat tail problem would be to focus on the
potential catastrophic consequences of low-probability, high-impact
events in developing GHG emissions targets and to specify a thresh-
old probability and a threshold loss. One can then remove events from
consideration that are below these critical values in determining what
mitigation and / or adaptation to adopt as part of a risk management
strategy for dealing with climate change (Kunreuther et al., 2013c).
Insurers and reinsurers specify these thresholds and use them to deter-
mine the amount of coverage that they are willing to offer against
a particular risk. They then diversify their portfolio of policies so the
annual probability of a major loss is below a pre-specified thresh-
old level of concern (e. g., 1 in 1000) (Kunreuther etal., 2013c). This
approach is in the spirit of a classic paper by Roy (1952) on safety-
first behaviour and can be interpreted as an application of probabilis-
tic cost-effectiveness analysis (i. e., chance constrained programming)
discussed in the next section. It was applied in a somewhat different
manner to environmental policy by Ciriacy-Wantrup (1971) who con-
tended that “a safe minimum standard is frequently a valid and rel-
evant criterion for conservation policy.
One could also view uncertainty or risk associated with different
options as one of the many criteria on which alternatives should be
evaluated. Multi-criteria analysis (MCA) is sometimes proposed to
overcome some of the limitations of CBA (see more on its basic fea-
tures in Chapter 3 and for applications in Chapter 6). MCA implies that
the different criteria or attributes should not be aggregated by convert-
ing all of them into monetary units. MCA techniques commonly apply
numerical analysis in two stages:
• Scoring: for each option and criterion, the expected consequences
of each option are assigned a numerical score on a strength of
preference scale. More (less) preferred options score higher (lower)
on the scale. In practice, scales often extend from 0 to 100, where
0 is assigned to a real or hypothetical least preferred option, and
100 is assigned to a real or hypothetical most preferred option. All
options considered in the MCA would then fall between 0 and 100.
• Weighting: numerical weights are assigned to define their relative
performance on a chosen scale that will often range from 0 (no
importance) to 1 (highest importance) (Dodgson etal., 2009).
2�5�4 Cost-effectiveness analysis
2�5�4�1 Elements of the theory
Cost-effectiveness analysis (CEA) is a tool based on constrained optimi-
zation for comparing policies designed to meet a pre-specified target.
The target can be defined through CBA, by applying a specific guide-
line such as the precautionary principle (see Section 2.5.5), or by speci-
fying a threshold level of concern or environmental standard in the
spirit of the safety-first models discussed above. The target could be
chosen without the need to formally specify impacts and their respec-
tive probabilities. It could also be based on an ethical principle such as
minimizing the worst outcome, in the spirit of a Rawlsian fair agree-
ment, or as a result of political and societal negotiation processes.
Cost-effectiveness analysis does not evaluate benefits in monetary
terms. Rather, it attempts to find the least-cost option that achieves a
desired quantifiable outcome. In one sense CEA can be seen as a spe-
cial case of CBA in that the technique replaces the criterion of choos-
ing a climate policy based on expected costs and benefits with the
objective of selecting the option that minimizes the cost of meeting
an exogenous target (e. g., equilibrium temperature, concentration, or
emission trajectory).
Like CBA, CEA can be generalized to include uncertainty. One solution
concept requires the externally set target to be specified with certainty.
The option chosen is the one that minimizes expected costs. Since
temperature targets cannot be met with certainty (den Elzen and van
Vuuren, 2007; Held etal., 2009), a variation of this solution concept
requires that the likelihood that an exogenous target (e. g., equilib-
rium temperature) will be exceeded is below a pre-defined threshold
probability. This solution procedure, equivalent to chance constrained
172172
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
programming (CCP) (Charnes and Cooper, 1959), enables one to use
stochastic programming to examine the impacts of uncertainty with
respect to the cost of meeting a pre-specified target. Chance con-
strained programming is a conceptually valid decision-analytic frame-
work for examining the likelihood of attaining climate targets when
the probability distributions characterizing the decision maker’s state
of knowledge is held constant over time (Held etal., 2009).
2�5�4�2 How can CEA improve decision making?
To illustrate how CEA can be useful, consider a national government
that wants to set a target for reducing greenhouse gas (GHG) emis-
sions in preparation for a meeting of delegates from different countries
at the Conference of Parties (COP). It knows there is uncertainty as to
whether specific policy measures will achieve the desired objectives.
The uncertainties may be related to the outcomes of the forthcoming
negotiation process at the COP and / or to the uncertain impacts of
proposed technological innovations in reducing GHG emissions. Cost-
effectiveness analysis could enable the government to assess alterna-
tive mitigation strategies (or energy investment policies) for reduc-
ing GHG emissions in the face of these uncertainties by specifying a
threshold probability that aggregate GHG emissions will not be greater
than a pre-specified target level.
2�5�4�3 Advantages and limitations of CEA over CBA
Cost-effectiveness analysis has an advantage over CBA in tackling
the climate problem in that it does not require formalized knowledge
about global warming impact functions (Pindyck, 2013). The focus of
CEA is on more tangible elements, such as energy alternatives, where
scientific understanding is more established (Stern, 2007). Still, CEA
does require scientific input on potential risks associated with climate
change. National and international political processes specify tempera-
ture targets and threshold probabilities that incorporate the prefer-
ences of different actors guided by data from the scientific community.
The corresponding drawback of CEA is that the choice of the target is
specified without considering its impact on economic efficiency. Once
costs to society are assessed and a range of temperature targets is
considered, one can assess people’s preferences by considering the
potential benefits and costs associated with different targets. However,
if costs of a desirable action turn out to be regarded as too high, then
CEA may not provide sufficient information to support taking action
now. In this case additional knowledge on the mitigation benefit side
would be required.
An important application of CEA in the context of climate change is
evaluating alternative transition pathways that do not violate a pre-
defined temperature target. Since a specific temperature target can-
not be attained with certainty, formulating probabilistic targets as a
CCP problem is an appropriate solution technique to use. However,
introducing anticipated future learning so that probability distribu-
tions change over time can lead to infeasible solutions (Eisner etal.,
1971). Since this is a problem with respect to specifying temperature
targets, Schmidt etal. (2011) proposed an approach that that com-
bines CEA and CBA. The properties of this hybrid model (labelled ‘cost
risk analysis’) require further investigation. At this time, CEA through
the use of CCP represents an informative concept for deriving miti-
gation costs for the case where there is no learning over time. With
learning, society would be no worse off than the proposed CEA solu-
tion.
2�5�5 The precautionary principle and robust
decision making
2�5�5�1 Elements of the theory
In the 1970s and 1980s, the precautionary principle was proposed for
dealing with serious uncertain risks to the natural environment and
to public health (Vlek, 2010). In its strongest form the precautionary
principle implies that if an action or policy is suspected of having a risk
that causes harm to the public or to the environment, precautionary
measures should be taken even if some cause and effect relationships
are not established. The burden of proof that the activity is not harmful
falls on the proponent of the activity rather than on the public. A con-
sensus statement to this effect was issued at the Wingspread Confer-
ence on the Precautionary Principle on 26 January 1998.
The precautionary principle allows policymakers to ban products or
substances in situations where there is the possibility of their caus-
ing harm and / or where extensive scientific knowledge on their risks is
lacking. These actions can be relaxed only if further scientific findings
emerge that provide sound evidence that no harm will result. An influ-
ential statement of the precautionary principle with respect to climate
change is principle 15 of the 1992 Rio Declaration on Environment and
Development: “where there are threats of serious or irreversible dam-
age, lack of full scientific certainty shall not be used as a reason for
postponing cost-effective measures to prevent environmental degra-
dation.
Robust decision making (RDM) is a particular set of methods devel-
oped over the last decade to address the precautionary principle in a
systematic manner. RDM uses ranges or, more formally, sets of plau-
sible probability distributions to describe uncertainty and to evaluate
how well different policies perform with respect to different outcomes
arising from these probability distributions. RDM provides decision
makers with tradeoff curves that allow them to debate how much
expected performance they are willing to sacrifice in order to improve
outcomes in worst case scenarios. RDM thus captures the spirit of the
precautionary principle in a way that illuminates the risks and benefits
of different policies. Lempert etal. (2006) and Hall etal. (2012) review
the application of robust approaches to decision making with respect
to mitigating or adapting to climate change.
173173
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
The tolerable windows approach can also be regarded as a ‘robust
method’. Temperature targets are specified and the bundle of decision
paths compatible with the targets is characterized. Mathe matically, the
tolerable windows approach incorporates the features of CEA or CCP
without optimization. The selection of the relevant targets and the paths
to achieving it are left to those making the decision. (See Bruckner and
Zickfeld (2008) for an introduction and an overview to peer-reviewed
literature on the tolerable windows approach.)
2�5�6 Adaptive management
Adaptive management is an approach to governance that that grew
out of the field of conservation ecology in the 1970s and incorporates
mechanisms for reducing uncertainty over time (Holling, 1978; Walters
and Hilborn, 1978). Paraphrasing the IPCC Special Report on Extreme
Events (SREX) (IPCC, 2012), adaptive management represents struc-
tured processes for improving decision making and policy over time,
by incorporating lessons learned. From the theoretical literature, two
strands of adaptive management have been developed for improving
decision making under uncertainty: passive and active.
Passive adaptive management (PAM) involves carefully designing
monitoring systems, at the relevant spatial scales, so as to be able to
track the performance of policy interventions and improve them over
time in response to what has been learned. Active adaptive manage-
ment (AAM) extends PAM by designing the interventions themselves
as controlled experiments, so as to generate new knowledge. For
example, if a number of political jurisdictions were seeking to imple-
ment support mechanisms for technology deployment, in an AAM
approach they would deliberately design separate mechanisms that
are likely to differ across jurisdictions. By introducing such variance
into the management regime, however, one would collectively learn
more about how industry and investors respond to a range of interven-
tions. All jurisdictions could then use this knowledge in a later round
of policymaking, reflecting the public goods character of institutional
knowledge.
With respect to the application of PAM, Nilsson (2005) reports on a
case study of Sweden, in which policymakers engaged in repetitive ex
post analyses of national climate policy, and then responded to the les-
sons learned by modifying their goals and strategies. There are many
documented cases of PAM applications in the area of climate change
adaptation (Lawler etal., 2008; Berkes etal., 2000; Berkes and Jolly,
2001; Joyce etal., 2009; Armitage, 2011). The information gathering
and reporting requirements of the UNFCCC are also in the spirit of
PAM with respect to policy design, as are the diversity of approaches
implemented for renewable energy support across the states and prov-
inces of North America and the countries in Europe. The combination of
the variance in action with data gathered about the consequences of
these actions by government agencies has allowed for robust analysis
on the relative effectiveness of different instruments (Blok, 2006; Men-
donça, 2007; Butler and Neuhoff, 2008).
Individuals relying on intuitive thinking are unlikely to undertake
experimentation that leads to new knowledge, as discussed in Section
2.4.3.1. In theory, adaptive management ought to correct this problem
by making the goal of learning through experimentation an explicit
policy goal. Lee (1993) illustrates this point by presenting a paradig-
matic case of AAM designed to increase salmon stocks in the Columbia
River watershed in the western United States and Canada. In this case,
there was the opportunity to introduce a number of different manage-
ment regimes on the individual river tributaries, and to reduce uncer-
tainty about salmon population dynamics. As Lee (1993) documented,
policymakers on the Columbia River were ultimately not able to carry
through with AAM: local constituencies, valuing their own immediate
interests over long-term learning in the entire region, played a crucial
role in blocking it. One could imagine such political and institutional
issues hindering the application of AAM at a global scale with respect
to climate change policies.
To date, there are no cases in the literature specifically documenting
climate change policies explicitly incorporating AAM. However, there
are a number of examples where policy interventions implicitly fol-
low AAM principles. One of these is promotion of energy research and
development (R&D). In this case the government invests in a large
number of potential new technologies, with the expectation that some
technologies will not prove practical, while others will be successful
and be supported by funding in the form of incentives such as subsi-
dies (Fischer and Newell, 2008).
2�5�7 Uncertainty analysis techniques
Uncertainty analysis consists of both qualitative and quantitative
methodologies (see Box 2.2 for more details). A Qualitative Uncer-
tainty Analysis (QLUA) helps improve the choice process of decision
makers by providing data in a form that individuals can easily under-
stand. QLUA normally does not require complex calculations so that it
can be useful in helping to overcome judgmental biases that character-
ize intuitive thinking. QLUA assembles arguments and evidence and
provides a verbal assessment of plausibility, frequently incorporated in
a Weight of Evidence (WoE) narrative.
A Quantitative Uncertainty Analysis (QNUA) assigns a joint distribu-
tion to uncertain parameters of a specific model used to characterize
different phenomena. Quantitative Uncertainty Analysis was pioneered
in the nuclear sector in 1975 to determine the risks associated with
nuclear power plants (Rasmussen, 1975). The development of QNUA
and its prospects for applications to climate change are reviewed by
Cooke (2012).
2�5�7�1 Structured expert judgment
Structured expert judgment designates methods in which experts
quantify their uncertainties to build probabilistic input for complex
174174
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
decision problems (Morgan and Henrion, 1990; Cooke, 1991; O’Hagan
et al., 2006). A wide variety of activities fall under the heading of
expert judgment that includes blue ribbon panels, Delphi surveys, and
decision conferencing.
Elements
Structured expert judgment such as science-based uncertainty quan-
tification was pioneered in the Rasmussen Report on risks of nuclear
power plants (Rasmussen, 1975). The methodology was further elabo-
rated in successive studies and involves protocols for expert selection
and training, elicitation procedures and performance-based combi-
nations that are described in more detail in Goossens etal. (2000).
In large studies, multiple expert panels provide inputs to computer
models with no practical alternative for combining expert judg-
ments except to use equal weighting. Hora (2004) has shown that
equal weight combinations of statistically accurate (‘well calibrated’)
experts loses statistical accuracy. Combinations based on experts’
statistical accuracy have consistently given more accurate and infor-
mative results (see for example Cooke and Goossens, 2008; Aspinall,
2010).
How can this tool improve decision making under uncertainty?
Structured expert judgment can provide insights into the nature of
the uncertainties associated with a specific risk and the importance of
undertaking more detailed analyses to design meaningful strategies
and policies for dealing with climate change in the spirit of deliberative
thinking. In addition to climate change (Morgan and Keith, 1995; Zick-
feld etal., 2010), structured expert judgment has migrated into many
fields such as volcanology (Aspinall, 1996, 2010), dam/dyke safety
(Aspinall, 2010), seismicity (Klügel, 2008), civil aviation (Ale et al.,
2009), ecology (Martin et al., 2012; Rothlisberger etal., 2012), toxi-
cology (Tyshenko etal., 2011), security (Ryan etal., 2012), and epidemi-
ology (Tuomisto etal., 2008).
The general conclusions emerging from experience with structured
expert judgments to date are: (1) formalizing the expert judgment pro-
cess and adhering to a strict protocol adds substantial value to under-
standing the importance of characterizing uncertainty; (2) experts
differ greatly in their ability to provide statistically accurate and infor-
mative quantifications of uncertainty; and (3) if expert judgments must
be combined to support complex decision problems, the combination
Box 2�2 | Quantifying uncertainty
Natural language is not adequate for propagating and com-
municating uncertainty. To illustrate, consider the U. S. National
Research Council 2010 report Advancing the Science of Climate
Change (America’s Climate Choices: Panel on Advancing the Sci-
ence of Climate Change; National Research Council, 2010). Using
the AR4 calibrated uncertainty language, the NRC is highly confi-
dent that (1) the Earth is warming and that (2) most of the recent
warming is due to human activities.
What does the second statement mean? Does it mean the NRC is
highly confident that the Earth is warming and the recent warm-
ing is anthropogenic or that, given the Earth is warming, are they
highly confident humans cause this warming? The latter seems
most natural, as the warming is asserted in the first statement. In
that case the ‘high confidence’ applies to a conditional statement.
The probability of both statements being true is the probability
of the condition (Earth is warming) multiplied by the probability
of this warming being caused by humans, given that warming is
taking place. If both statements enjoy high confidence, then in
the calibrated language of AR4 where high confidence implies a
probability of 0.8, the statement that both are true would only be
“more likely than not” (0.8 x 0.8 = 0.64).
Qualitative uncertainty analysis easily leads the unwary to errone-
ous conclusions. Interval analysis is a semi-qualitative method in
which ranges are assigned to uncertain variables without distribu-
tions and can mask the complexities of propagation, as attested
by the following statement in an early handbook on risk analysis:
“The simplest quantitative measure of variability in a parameter or
a measurable quantity is given by an assessed range of the values
the parameter or quantity can take. This measure may be adequate
for certain purposes (e. g., as input to a sensitivity analysis), but in
general it is not a complete representation of the analyst’s knowl-
edge or state of confidence and generally will lead to an unreal-
istic range of results if such measures are propagated through an
analysis”, (U. S. NRC, 1983, Chapter12, p.12).
The sum of 10 independent variables each ranging between
zero and ten, can assume any value between zero and 100. The
upper (lower) bound can be attained only if ALL variables take
their maximal (minimal) values, whereas values near 50 can
arise through many combinations. Simply stating the interval
[0, 100] conceals the fact that very high (low) values are much
more exceptional than central values. These same concepts are
widely represented throughout the uncertainty analysis literature.
According to Morgan and Henrion (1990): “Uncertainty analysis
is the computation of the total uncertainty induced in the out-
put by quantified uncertainty in the inputs and models […] Fail-
ure to engage in systematic sensitivity and uncertainty analysis
leaves both analysts and users unable to judge the adequacy of
the analysis and the conclusions reached”, (Morgan and Henrion,
1990, p.39).
175175
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
method should be subjected to the following quality controls: statisti-
cal accuracy and informativeness (Aspinall, 2010).
As attested by a number of governmental guidelines, structured expert
judgment is increasingly accepted as quality science that is applicable
when other methods are unavailable (U. S. Environmental Protection
Agency, 2005). Some expert surveys of economists concerned with
climate change examine damages (Nordhaus, 1994) and appropri-
ate discount rates (Weitzman, 2001). Structured expert judgments of
climate scientists were recently used to quantify uncertainty in the
ice sheet contribution to sea level rise, revealing that experts’ uncer-
tainty regarding the 2100 contribution to sea level rise from ice sheets
increased between 2010 and 2012 (Bamber and Aspinall, 2013).
Damages or benefits to ecosystems from invasions of non-indigenous
species are difficult to quantify and monetize on the basis of histori-
cal data. However ecologists, biologists and conservation economists
have substantial knowledge regarding the possible impacts of inva-
sive species. Recent studies applied structured expert judgment with
a performance-based combination and validation to quantify the costs
and benefits of the invasive species introduced since 1959 into the U. S.
Great Lakes by opening the St. Lawrence Seaway (Rothlisberger etal.,
2009, 2012). Lessons from studies such as these reveal that experts
may have applicable knowledge that can be captured in a structured
elicitation when historical data have large uncertainties associated
with them.
Advantages and limitations of structured expert judgment
Expert judgment studies do not reduce uncertainty; they merely quan-
tify it. If the uncertainties are large, as indeed they often are, then deci-
sion makers cannot expect science to relieve them of the burden of
deciding under conditions of ambiguity. Since its inception, structured
expert judgment has been met with scepticism in some quarters; it is,
after all, just opinions and not hard facts. Its steady growth and widen-
ing acceptance over 35 years correlates with the growth of complex
decision support models. The use of structured expert judgment must
never justify a diminution of effort in collecting hard data.
2�5�7�2 Scenario analysis and ensembles
Scenario analysis develops a set of possible futures based on extrapo-
lating current trends and varying key parameters, without sampling in
a systematic manner from an uncertainty distribution. Utilizing suffi-
ciently long time horizons ensures that structural changes in the sys-
tem are considered. The futurist Herman Kahn and colleagues at the
RAND Corporation are usually credited with inventing scenario analy-
sis (Kahn and Wiener, 1967). In the climate change arena, scenarios are
currently presented as different emission pathways or Representative
Concentration Pathways (RCPs). Predicting the effects of such path-
ways involves modelling the Earth’s response to changes in GHG con-
centrations from natural and anthropogenic sources. Different climate
models will yield different projections for the same emissions scenario.
Model Intercomparison studies generate sets of projections termed
‘ensembles’ (van Vuuren etal., 2011).
Elements of the theory
Currently, RCPs are carefully constructed on the bases of plausible
storylines while insuring (1) they are based on a representative set of
peer-reviewed scientific publications by independent groups, (2) they
provide climate and atmospheric models as inputs, (3) they are harmo-
nized to agree on a common base year, and (4) they extend to the year
2100. The four RCP scenarios, shown in Figure 2.3 relative to the range
of baseline scenarios in the literature, roughly span the entire scenario
literature, which includes control scenarios reaching 430 ppm CO
2
eq
or lower by 2100. The scenarios underlying the RCPs were originally
developed by four independent integrated assessment models, each
with their own carbon cycle. To provide the climate community with
four harmonized scenarios, they were run through the same carbon
cycle / climate model (Meinshausen etal., 2011). Note that a represen-
tative set is not a random sample from the scenarios as they do not
represent independent samples from some underlying uncertainty dis-
tribution over unknown parameters.
Ensembles of model runs generated by different models, called multi-
model ensembles or super-ensembles, convey the scatter of the climate
response and natural internal climate variability around reference sce-
narios as sampled by a set of models, but cannot be interpreted proba-
bilistically without an assessment of model biases, model interdepen-
dence, and how the ensemble was constructed (see WGI AR5 Section
12.2; Knutti et al., 2010). In many cases the assessed uncertainty is
larger than the raw model spread, as illustrated in Figure 2.4. The
shaded areas (+ / - 1 standard deviation) around the time series do not
imply that 68 % are certain to fall in the shaded areas, but the model-
ers’ assessed uncertainty (likely ranges, vertical bars on the right) are
larger. These larger ranges reflect uncertainty in the carbon cycle and
the full range of climate sensitivity (WGI AR4 Section 10.5.4.6 and Box
10.3; Knutti etal., 2008) but do not reflect other possible sources of
uncertainty (e. g., ice sheet dynamics, permafrost, or changes in future
solar and volcanic forcings). Moreover, many of these models have
common ancestors and share parameterizations or code (Knutti etal.,
2013) creating dependences between different model runs. Probability
statements on global surface warming require estimating the models’
bias and interdependence (see WGI AR5 Sections 12.2 and 12.4.1.2).
WGI AR5 assigns likelihood statements (calibrated language) to global
temperature ranges for the RCP scenarios (WGI AR5 Table SPM.2) but
does not provide probability density functions (PDFs), as there is no
established formal method to generate PDFs based on results from dif-
ferent published studies.
Advantages and limitation of scenario and ensemble analyses
Scenario and ensemble analyses are an essential step in scoping the
range of effects of human actions and climate change. If the scenarios
span the range of possible outcomes, they may be seen as providing
support for uncertainty distributions in a formal uncertainty analysis. If
specific assumptions are imposed when generating the scenarios, then
176176
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
Figure 2�3 | Total radiative forcing (left panel) and cumulative carbon emissions since 1751 (right panel) in baseline scenario literature compared to RCP scenarios. Forcing was estimated
ex-post from models with full coverage using the median output from the MAGICC results. Secondary axis in the left panel expresses forcing in CO
2
eq concentrations. Scenarios are depicted
as ranges with median emboldened; shading reflects interquartile range (darkest), 5th 95th percentile range (lighter), and full extremes (lightest). Source: Figure 6.6 from WGIII AR5.
0
1
2
3
4
5
6
7
8
9
10
Total Radiative Forcing [W/m
2
]
1600
1200
900
700
550
450
0
1
2
3
2010 2030 2050 2070 2090 2010 2030 2050 2070 2090
Cumulative Carbon Emissions [TtC]
0.55 TtC (1751-2010)
CO
2
-Equivalent Concentration
[ppm CO
2
eq]
Percentile0-100
th
5-95
th
25-75
th
RCP 8.5
RCP 6.0
RCP 4.5
RCP 2.6
Figure 2�4 | Solid lines are multi-model global averages of surface warming (relative to 1980 1999) for the scenarios A2, A1B and B1, shown as continuations of the 20th century
simulations. Shading denotes the ± 1 standard deviation range of individual model annual averages. The orange line is for the experiment where concentrations were held constant
at year 2000 values. The grey bars at right indicate the best estimate (solid line within each bar) and the likely range assessed for the six families of emissions scenarios discussed
in the IPCC’s Fourth Assessment Report (AR4). The assessment of the best estimate and likely ranges in the grey bars includes the Atmosphere-Ocean General Circulation Models
(AOGCMs) in the left part of the figure, as well as results from a hierarchy of independent models and observational constraints. Based on: Figure SPM.5 from WGI AR5.
177177
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
the support is conditional on these assumptions (see Section 6.2.3).
The advantage of scenario / ensemble analyses is that they can be per-
formed without quantifying the uncertainty of the underlying unknown
parameters. On the downside, it is easy to read more into these analy-
ses than is justified. Analysts often forget that scenarios are illustra-
tive possible futures along a continuum. They tend to use one of those
scenarios in a deterministic fashion without recognizing that they have
a low probability of occurrence and are only one of many possible out-
comes. The use of probabilistic language in describing the swaths of
scenarios (such as standard deviations in Figure 2.4) may also encour-
age the misunderstandings that these represent science-based ranges
of confidence.
The study of representative scenarios based on probabilistic fore-
casts have been shown to facilitate strategic planning by professional
groups such as military commanders, oil company managers, and poli-
cymakers (Schoemaker, 1995; Bradfield etal., 2005). Recent work on
ice sheet modelling (Little etal., 2013) points in this direction. Using
modelling assumptions and prior distributions on model coefficients,
Monte Carlo simulations are used to produce probabilistic predictions.
Expert informed modelling is methodologically intermediate between
structured expert judgment (Bamber and Aspinall, 2013) and non-
probabilistic scenario sweeps. Structured expert judgment leaves the
modelling assumptions to the experts who quantify their uncertainty
on future observables.
2.6 Managing uncertainty,
risk and learning
2�6�1 Guidelines for developing policies
This section assesses how the risks and uncertainties associated with
climate change can affect choices with respect to policy responses,
strategies, and instruments. At the time of the AR4, there was some
modelling-based literature on how uncertainties affected policy design,
but very few empirical studies. In the intervening years, international
negotiations failed to establish clear national emissions reductions
targets, but established a set of normative principles, such as limit-
ing global warming to 2 °C. These are now reflected in international,
national, and subnational planning processes and have affected the
risks and uncertainties that matter for new climate policy develop-
ment. Greater attention and effort has been given to finding syner-
gies between climate policy and other policy objectives, so that it is
now important to consider multiple benefits of a single policy instru-
ment. For example, efforts to protect tropical rainforests (McDermott
etal., 2011), rural livelihoods (Lawlor etal., 2010), biodiversity (Jin-
nah, 2011), public health (Stevenson, 2010), fisheries (Axelrod, 2011),
arable land (Conliffe, 2011), energy security (Battaglini etal., 2009),
and job creation (Barry etal., 2008) have been framed as issues that
should be considered when evaluating climate policies.
The treatment here complements the examination of policies and
instruments in later chapters of this report, such as Chapter 6 (which
assesses the results of IAMs) and Chapters 13 15 (which assess policy
instruments at a range of scales). Those later chapters provide greater
details on the overall tradeoffs to be made in designing policies. The
focus here is on the special effects of various uncertainties and risks on
those tradeoffs.
• Section 2.6.2 discusses how institutions that link science with pol-
icy grapple with several different forms of uncertainty so that they
meet both scientific and political standards of accountability.
• Section 2.6.3 presents the results of integrated assessment models
(IAMs) that address the choice of a climate change temperature
target or the optimal transition pathway to achieve a particular
target. IAMs normally focus on a social planner operating at the
global level.
• Section 2.6.4 summarizes the findings from modelling and empiri-
cal studies that examine the processes and architecture of interna-
tional treaties.
• Section 2.6.5 presents the results of modelling studies and the
few empirical analyses that examine the choice of particular policy
instruments at the sovereign state level for reducing GHG emis-
sions. It also examines how the adoption of energy efficiency prod-
ucts and technologies can be promoted at the firm and household
levels. Special attention is given to how uncertainties affect the
performance and effectiveness of these policy instruments.
• Section 2.6.6 discusses empirical studies of people’s support or
opposition with respect to changes in investment patterns and
livelihood or lifestyles that climate policies will bring about. These
studies show people’s sensitivity to the impact that climate change
will have on their personal health or safety and their perceptions
of the health and safety risks associated with the new technolo-
gies addressing the climate change problem.
Linking intuitive thinking and deliberative thinking processes for deal-
ing with uncertainties associated with climate change and climate
policy should increase the likelihood that instruments and robust poli-
cies will be implemented. In this sense, the concepts presented in this
section should be viewed as a starting point for integrating descriptive
models with normative models of choice for developing risk manage-
ment strategies.
178178
Integrated Risk and Uncertainty Assessment of Climate Change Response Policies
2
Chapter 2
2�6�2 Uncertainty and the science/policy