Abstract:
In this talk, Dr. Pete Barbrook-Johnson will present research using methods from systems thinking and complexity economics to model the energy transition. In particular, he will show how both systems mapping and agent-based modelling can be applied to policy questions. Pete will show examples of these methods being used, comment on challenges in their use, and plot some future directions. You can read more about Pete’s work at https://www.barbrookjohnson.com/ and follow him on Twitter @bapeterj.
Bio:
Pete Barbrook-Johnson is a social scientist, economist, complexity scientist, and systems thinker. He regularly uses research methods such as agent-based modelling and systems mapping in his applied environmental, energy, and public health research and policy analysis. He teaches on a range of undergraduate and masters courses, focussing on the economics of environmental change, and the use of complexity and systems sciences in environmental issues.
Recent work highlights include:
His book on systems mapping, a practical guide written with Dr. Alexandra Penn.
A library of new economic modelling case studies. Pete led the production of this large report with colleagues from China, India, Brazil, the EU, and UK. It presents a range of new economic modelling approaches being used to understand and inform policy on the energy transition.
View a full list of Pete's academic publications on his Google Scholar page.
Summary:
Focus: modeling the global energy transition
Approach: systems thinking, complexity economics
Assertions:
We need climate economy models to fit the decisions we face now
We need a variety of models to mitigate bias in analysis
Observations: many existing models have a bias (e.g. Dynamic General Equilibrium Models, which are typically used by Integrated Assessment Models)
Needs of decision makers
Public policy: know they need to address climate change, need to design specific, sector focused policies that accomplish the transition and are economically beneficial
Business policy: need to make strategic decisions that position their business for success
Modeling approach:
Traditional economic models are equilibrium-based,
Predicts the optimal/stable state the system will settle into
Focused on marginal changes around the equilibrium
Examples (from least detailed to most): Systems Mapping, Economic Complexity, Stochastic Experience Curves, Disequilibrium Macro, Future Technology Transformation., System Dynamics, Agent-based Modeling
However, the target problem is explicitly a change between equilibria
Need range of possible outcomes
Dynamic description of how the system changes over time
New modeling approach
More granular, regions, sectors
Tech progress
Policy in details
More dynamic (explicit time), no equilibrium
Models need more granular data than prior generation of models
Example: Systems Mapping
Represent the causal structure in a system
Typically captured via directed graphs (edges are cause->effect relationships)
Useful for: structuring thinking about the problem, synthesizes/organizes information, helps communication, helps to extrapolate
Types:
Simple Causal Loop
Simple graph
Identifies primary feedbacks (dampening vs amplifying change)
Systems-based Theory of Change
Medium complexity
Causal assumptions behind our theory of change
Inputs + activities -> outputs + outcomes
Participatory System Map
High complexity
Captures knowledge from many stakeholders
Builds ownership by stakeholders, consensus
Often used as the first step to building a system dynamics model
Future directions:
Using data to build these
Quantitative: time series, causal inference
Text data via LLM/NLP
Blend available data sources
Example: Agent-based Modeling
Very detailed and quantitative
Modeling occupation-level unemployment in Brazilian labor market
How will different development paths affect employment in sectors, regions, occupations?
Built an occupational mobility network
Nodes: occupations
Edges: workers moving between occupations
Based on detailed dataset of occupation changes
Labor market shocks/scenarios for Total Factor Productivity (TFP)
Sourced from the World Bank
Baseline: not changing
TFP in Agriculture increases .5% per year
Impacts: 1.8% higher GDP, moderate deforestation reduction, more CO2 emissions
TFP in Manufacturing increases .5% per year
Impacts: 3.9% higher GDP, higher deforestation reduction, lower CO2 emissions
Note: Brazil is a major CO2 emitter, mostly from Agriculture and Land Use, little from energy
Propagated these shocks through the occupational mobility network
People chose/transition to new occupations at rates that have been observed in the dataset
Useful for both prediction and for identifying key mobility bottlenecks that may need to be addressed by policymakers
Modeling for policy
Modeling often used as policies are being designed; need to be maintained through the entire deployment cycle to help with policy evaluation
Challenges:
Conceptual
Practical: data and different types of outputs
Institutional: risk aversion, structures around existing models
How to start with detailed modeling?
Systems mapping, then iteratively increase complexity