Actors behaving badly: modelling non

Actors behaving badly:
modelling non-optimal behaviour in energy transitions
Dr. Francis Li | [email protected]
Research Associate in Energy Systems Modelling
UCL Energy Institute | www.ucl.ac.uk/energy
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
Overview
• Background to Model Development
• BLUE Model Overview and Structure
• Example Results
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
Overview
• Background to Model Development
• BLUE Model Overview and Structure
• Example Results
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
Introduction: Realising Transition Pathways (RTP)
• Interdisciplinary research consortium with 9 universities, 2008-2016
• Emphasis on exploring socio-technical change and understanding not only technological change and
market design but also the role of governance, behaviour and lifestyle factors in energy transitions
Image: Based on Foxon et al. 2010,
Developing transition pathways for a
low carbon electricity system in the UK
doi: 10.1016/j.techfore.2010.04.002
Introduction: Common Critiques of Energy Systems Modelling
Critique
Community Responses
Examples
(Hyperlinks to all references)
Model structure and data
are often not 100% transparent
Community efforts to publish
documentation, develop open
source tools
Howells et al., 2011
Model operation is overly
deterministic and constraints
merely reflect biases of model
operator
Techniques that consider parameter
uncertainty (e.g. Monte Carlo)
Techniques for exploring structural
uncertainty (e.g. MGA)
Pye et al., 2014
DeCarolis, 2011
Blanford et al., 2014
Multi-model comparison exercises
Models incorporate perfect
foresight so optimal cost paths
don’t necessarily form a good
basis for policy design (timing)
Cost minimisation implies
rational choice and perfect
markets, overlooking important
insights from behavioural
economic theory
Myopic model development
Keppo and Strubegger, 2010
Stochastic programming and robust
optimization
Labriet et al., 2015
Heterogeneous choice behaviour
McCollum et al., 2016
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
Introduction: Improving Representation of Actors and Behaviour in Energy Modelling
• Deviations from neoclassical paradigm of rational
economic choice include:
o
Actors lack perfect information or time to
consider all their options (Simon, 1956)
o
Actors influenced by price and non-price factors
e.g. aesthetics, branding, perception of
reliability (Neij et al. 2009)
o
Actors influenced by wider social norms and
customs, often termed “neighbour effects”
(e.g. Mau et al. 2008)
o
Purchase behaviour in consumer society is not
just satisfying end-use demand for
goods/services but also a means of social
signalling, and demonstrating lifestyle
aspirations (Axsen and Kurani, 2011)
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
Introduction: Improving Representation of Actors and Behaviour in Energy Modelling
• Improving the utility of energy models for policy has been argued to involve:
o
Better representation of behavioural factors
(Wilson and Dowlatabadi, 2007, Mundaca et al. 2010)
o
More explicit representation of actors and institutions
(Hughes and Strachan 2010, Trutnevyte et al. 2016)
• Whole system energy economy models that have been demonstrated with
some degree of heterogeneous choice behaviour:
o
GCAM (Kyle and Kim, 2012)
o
IMAGE/TIMER (Daioglou et al., 2012)
o
MESSAGE-Transport (McCollum et al., 2016)
o
TIMES variants (e.g. Daly et al. 2015)
o
CIMS (Jaccard and Rivers, 2005)
o
IMACLIM (Giraudet et al., 2012)
• Actors and institutions are generally mostly still represented by discrete
scenarios, with agent-based approaches yet to scale fully to whole system
analysis (Mercure et al. 2016)
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
Introduction: Insights into “Second-Best” Policy Worlds
•
•
“Second-best” theory envisages policy design as a kind of pareto
optimization across multiple fronts (Lipsey and Lancaster, 1956)
o
1st best implementation: correct pricing is all you need
o
2nd best implementation: additional policy instruments are
needed to address multiple market failures, and there are
additional (political) constraints (Bennear and Stavins, 2007)
Recent policy reversals in the UK underscore the importance of
understanding second-best outcomes:
o
UK has historically been a leader in setting climate targets
and developing an ambitious domestic emission control
framework
o
Implementation has proved much more challenging, with
long term uncertainty on policies enduring (Lockwood, 2013)
o
Since 2015, cancellation of multiple flagship energy
efficiency and technology demonstration schemes identified
on the critical cost path, with no replacements
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
Overview
• Background to Model Development
• BLUE Model Overview and Structure
• Example Results
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
BLUE: Concept and Design
• Behaviour, Lifestyles and Uncertainty Energy model
• An exploratory modelling tool for the UK to complement our existing pathway optimisation models
Desired Qualities
Implementation
Incorporate insights from behavioural economics
into technology choice framework
Heterogeneous choice behaviour for all actors
Improved representation of actors and
institutions in the energy system
Multiple actors with different behaviours, each
representing decisions taken in a single
economic sector
Suitable for exploring transitions under
uncertainty
Monte Carlo simulation for parameter
uncertainty
Very fast runtime to facilitate probabilistic and
stochastic operation, use in participatory
workshop settings
Computationally lightweight system dynamics
type simulation
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
BLUE: Structure
• Inspired by the Multi-Level Perspective on socio-technical transitions (Geels and Schot, 2007)
• Currently eight actors (A-H), each representing decisions taken in individual sectors
• Stylized model with a limited number of transition technologies (X) and changes to lifestyles (Y)
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
BLUE: Operation
• Model steps through time horizon, with changing landscape drivers, e.g.
o End-use demands
o Primary resource costs, technology costs
• Actors make decisions about capital stock replacement based on myopic expectations of levelised costs
• Actors react dynamically to each other through time
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
BLUE: Behavioural Inputs
• Actors are differentiated in their micro-economic behaviour by a
number of different parameters
• Each actor can be configured with different behaviours
Micro-economic Behaviour
Demand elasticities
Replacement/retrofit rates
Intangible/hidden costs
Hurdle rates
Heterogeneity of response
BLUE Parameters
Actors are sensitive to energy price changes
Actors experience different limits to deployment rates
Different actors can view identical technologies as having “hassle”
or barrier costs to them
Actors have different sensitivities to up front investments
Actors have different capacities to make “cost-optimal” decisions
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
Overview
• Background to Model Development
• BLUE Model Overview and Structure
• Example Results
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
BLUE: Example Results
• Demand drivers aligned with low/central/high trends used by UK Department of Energy and
Climate Change (DECC)
• Most inputs (fuel costs, technology costs, performance) captured as uncertain distributions
• Wide range of outcomes possible, e.g.
Emissions grow in spite of technological change
Deep decarbonisation of the energy system
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
BLUE: Example Results
• Differentiated actor behaviour is a key feature of the model (e.g. businesses
may be more cost conscious than consumers) but not yet explored in the early
work presented
• The presented paper compares technological transitions in the power,
transport, and building heating sectors under two scenarios:
o
All actors make near-cost optimal choices and take a long term social
planning perspective on valuing the future
o
All actors exhibit heterogeneous investment decisions and take a
private discounting perspective on the future
• Also explored are three different levels of ambition on emissions mitigation,
represented by 3 different price trajectories for CO2 (UK government range +
very high mitigation case)
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
Near-cost optimal behaviour
Social planning perspective
Power Sector
Heterogeneous choice behaviour
Private discounting perspective
Near-cost optimal behaviour
Social planning perspective
Residential Sector
Heterogeneous choice behaviour
Private discounting perspective
Near-cost optimal behaviour
Social planning perspective
Road Sector
Heterogeneous choice behaviour
Private discounting perspective
BLUE: Challenges for Future Development and Policy Engagement
Science-policy interface challenges:
o
In the UK, we have limited revealed/stated preference data for calibrating behavioural
parameters, necessitating an exploratory approach
o
Second-best outcomes and failure are not popular topics to explore with policymakers,
who often use models to reassure themselves of certainty
o
Framing such models as tools for exploring deep uncertainty, including behaviour, are
likely to be a more successful route to engagement
o
Possible need for such models to be used in an iterative, participatory fashion, following
a “modelling as a service” strategy rather than “modelling as a product” approach
(Lempert et al. 2002)
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
Closing Thoughts
•
National level policymakers risk focusing on a few deterministic model runs with very complex
tools, instead of broader analysis of uncertainties, including behaviour
•
Deterministic models with narrowly defined assumptions can give the appearance of certainty
where none exists, closing down debate and leading to ineffective policy design (Taylor et al.
2014, Stirling 2010)
•
The community has historically explored “second-best” insights by:
o
Introducing resource or technology deployment constraints to mimic real-world barriers
(e.g. Usher and Strachan, 2011)
o
Introducing delays to represent inertia in decision-making (e.g. Schaeffer et al., 2015)
•
More explicit modelling of actors and institutions may require new approaches
•
New approaches include:
o
Exploratory modelling with new tools that try to integrate techno-economic detail, actor
heterogeneity and transition pathway dynamics (Li et a. 2015)
o
Attempts to bridge analytical disciplines such as quantitative modelling and socio-technical
transition studies (Turnheim et al. 2015)
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
Thank you!
Realising Transition Pathways:
http://www.realisingtransitionpathways.org.uk/
UCL Energy Models:
https://www.ucl.ac.uk/energy-models/models
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland
Socio-Technical Modelling
Image: Based on Li et al. 2015, A
review of socio-technical energy
transition (STET) models
doi:10.1016/j.techfore.2015.07.017
2nd June 2016, Parallel Session 2E: Behaviour & People, International Energy Workshop 2016, University College Cork, Ireland