introduction to system dynamics - cepal
TRANSCRIPT
Introduction to System Dynamics
Model-based policy formulationModels for national development planning (T21)
Main Objectives
Analytical knowledge and skills:• SD method: Basic knowledge of the System
Dynamics method;• Behavioral analysis: Ability to relate a
system’s behavior to the underlying structure;
• Understanding complexity: basic elements of complexity in common social, economic and environmental issues.
Main Objectives (2)
Technical knowledge and skills:• Software: Knowledge of basic modeling
techniques with Vensim (www.vensim.com)• Modeling: Ability of representing economic,
social and environmental issues through simple simulation models;
• Simulation techniques: Ability to run and compare alternative simulation scenarios.
Model-Based Policy Formulation
Challenges and Guidelines
Outline
1. Setting the context– Public policy and formal models
2. The Process of Policy Formulation– Steps, content and context
3. Model-based policy formulation– Methodologies
– Challenges
– Guidelines
3. Model-based policy formulation
3.1 Why are models useful3.2 Methodologies
• Scenarios• Mental models• Formal models
– Optimization– Econometrics– Simulation
3.3 Challenges3.4 Guidelines
3.1. Why are models useful
• Accurate predictions about the future would be nice to have, but can we really get them?
“Essentially, all models are wrong, but some are useful.”
“Remember that all models are wrong; the practical question is how wrong do they have to be
to not be useful.”
Box, George E. P.; Norman R. Draper (1987). Empirical Model-Building and Response Surfaces. Wiley.
3.1. Why are models useful (2)
• Models can help policymaking in various ways: – Improving understanding of the possible consequences of
policy choices, – Deepening policymakers’ comprehension of the underlying
problems and issues, – Clarifying decision-makers’ assumptions and values helping to
build understandable narratives (“stories”) in support of policy proposals,
– Informing dialogue among stakeholders and policymakers, – Providing a framework for negotiation and consensus building.
• Policymaking is about trying to affect the future: to maintain or improve on the status quo of public wellbeing.
3.2. Methodologies: scenarios and mental models
• Scenarios: exploration of a wide range of possible futures. – No attempt to identify the most or least probable
among them, aimed at finding resiliency.
• Mental models: someone's explanation of how something works in the real world.– Psychological biases and cognitive limitations may
undermine the logical application of the model.
3.2. Methodologies: Formal models
• Optimization models, which generate “astatement of the best way to accomplish some goal” (Sterman, 1998), are normative, or prescriptive, models.
• Econometrics measures economic relations, running statistical analysis of economic data and finding correlation between specific selected variables.
• Simulation models aim at representing what the main drivers for the behavior of the system are.
3.2. Why Use a Formal Model?
Model
Perfect Information
Represented in a model
Simulation
Alternative scenarios
Real World(reality)
Interpretation of information
Mentalmodels
Strategy,structure, decision
rules
Decision
• Complexity of dynamic systems (descriptive model);• Bounded rationality and misperceptions of
feedbacks and delays (descriptive model);• Limited information (simulation);• Wrong deductions re. the dynamic behavior of
systems (model validation and analysis);• Defensive routines and personal emotional
involvement (alternative scenarios).
3.3. Challenges: barriers to learning
Reference: Sterman, 2000
3.3. Challenges: Methodologies
• Optimization: correct definition of an objective function, the extensive use of linearity, the lack of feedback and lack of dynamics.
• Econometrics: full rationality of human behavior, availability of perfect information and market equilibrium.
• Simulation: correct definition of boundaries and a realistic identification of the causal relations.
3.3. Challenges: Model-based policy formulation
• There is a need for integrated tools that could serve as a mean to close the gap between dynamic and all embracing thinking and conventional methodologies and models.
• Methodologies should be combined to:– Set targets (optimization)– Define a proposal (econometrics)– Refine a bill (system dynamics)
Models for NationalDevelopment Planning
Overview
1. What is National Development Planning (NDP)?
2. From strategy to implementation
3. How can models help?
4. Why System Dynamics?
5. Example: The Threshold21 (T21) model
6. Summary
National Vision
National Development Plan
Yearly Budgets
Mid Term Strategic Plans
1. What is NDP?1. What is NDP?
Cascade Planning System
National Development Planning is a:1. Planning process at the central government
level (e.g. Min. Finance)2. Defines the strategic axes for the country’s
medium/long-term development3. Based on the long-term objectives and forms
the basis for short-term strategic plans.
A definition1. What is NDP?1. What is NDP?
Some examples of mid-long term issues:• Poverty• Economic growth• Access to social services
– Education– Health
• Environmental sustainability• Quality of institutions• Urban planning, land use planning• Disaster risk management
Type of issues at stake1. What is NDP?1. What is NDP?
Strategic planning and Policy Development
Budget Preparation
Budget Execution
Accounting, Monitoring and Internal Audit
Reporting and External Audit
Policy Review and Revision
Implementation - Generic2. From Strategy to Implementation2. From Strategy to Implementation
strategy
objectives
currentsituation
decisionsinformation
feedback
3. How Can Models Help?3. How Can Models Help?
A Learning Process
strategy
objectives
currentsituation
decisions
planningmodels
informationfeedback
simulatedresults
3. How Can Models Help?3. How Can Models Help?
Role of Planning Models
Formal models provide the possibility to test policies beforehand and accelerate learning
Necessary characteristics formedium - long term planning models:
1. Endogenously represent key variables (E)2. Comprehensive (C)3. Properly represent dynamic complexity (D)4. Transparent (T)
Necessary characteristics
Endogenous Key Variables (E)real gdp at factor cost
4e+012
3e+012
2e+012
1e+012
01990 1995 2000 2005 2010 2015 2020 2025
Time (Year)
real gdp at factor cost : MODEL cfa87/Yearreal gdp at factor cost : DATA cfa87/Year
Mid-Long TermShort Term
INERTIA
FUNDAMENTAL CHANGES
GDPGov. Revenue Gov. Expenditure
Economy
Society
Environment
Comprehensive (C)
GDPGov. Revenue Gov. Expenditure
Economy
Society
Environment
Non-Linearity
00.10.20.30.40.50.60.70.80.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
MAJOR DELAYSNon-Linearity
00.10.20.30.40.50.60.70.80.9
1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
MAJOR DELAYS
Dynamic Complexity (D)
???
Output
Input
Transparency (T)
Approach Software E C D T
Disaggregated consistency MS-EXCEL NO NO NO YES
Macro-econometrics EVIEWS YES NO YES NO
Computable general eq. GAMS YES NO YES NO
Integrated simulation VENSIM YES YES YES YES
Existing approaches to NDP
Why SD models?
1. Focus on endogenous explanation2. Support multidisciplinary approach3. Proper representation of complexity4. Transparent – User friendly
Strengths of SD approach
Brief History of Economic Models
• Political Economy and conceptual models• Introduction of quantification to Economics• Linear models and Econometrics• Linear accounting frameworks: RMSM• Matrix models: I-O, SAM, and CGE• Broader Systemic models: T21
Linear Models and Econometrics
• Look at limited set of relations and variablesGDP = a*INV + cINV = 1/a * GDP - c
• Help understand importance of relationships in sub-sector mode
• Limited range of variables considered• No relation to other economic variables
Linear Accounting Frameworks
• Address “whole” economy in a single frameworkGDP = CONS + INV + XP - IMP = CONS + SAVINV - SAV = XP - IMPGOV BAL = REV + AID + BOR - EXP - DSERBoP = XP - IMP + TRANS + NetCap
• Help see how “whole” economy balances• Dominated by exogenous assumptions and
accounting balances• Few internal links and fewer links to other
important factors in development
Matrix Models: Input-Output
• Incorporate links among production activities
• Help understand how production sectors interact• Linear structure with many exogenous factors• Lack links with other economic factors or the rest
of society
Input-Output TableAgriculture Industry Services Total
Agriculture 25 53 18 96Industry 45 87 31 163Services 26 23 12 61Total 96 163 61
Matrix Models: SAMs
• Incorporate the rest of the economy with SAM
• Better view of “whole” economy with interaction among agents and equilibrium
• Heavy data requirements• Limited relations beyond SAM
Agriculture Industry Services Households Government RoW TotalAgriculture 25 53 18 43 12 19 170Industry 45 87 31 65 28 40 296Services 26 23 12 24 27 13 125Households 48 55 40 13 12 168Government 15 34 12 15 24 100RoW 11 44 12 21 20 108Total 170 296 125 168 100 108
Matrix Models: CGEs
• Convert SAM entries into equations
• Non-linear relations, but ‘static’ solutions from ‘black box’• Require all markets to clear in equilibrium• Lack links with social and environmental factors which
affect economy, e.g. MDGs
Agriculture Industry Services Households Government RoW TotalAgriculture A1=f(K,L,ais) I1=f(K,L,ais) S1=f(K,L,ais) Dag=f(Inc,Pref) Dgag=F(bud) AgX=f(ForD) Sum AgIndustry A2=f(K,L,ais) I2=f(K,L,ais) S2=f(K,L,ais) Din=f(Inc,Pref) Dgin=F(bud) InX=f(ForD) Sum InServices A3=f(K,L,ais) I3=f(K,L,ais) S3=f(K,L,ais) Dse=f(Inc,Pref) Dgse=F(bud) SeX=f(ForD) Sum SeHouseholds AW=f(K,L,ais) IW=f(K,L,ais) SW=f(K,L,ais) Trans=f(bud) Rem=f(Emig) Sum HHGovernment Atax=f(Ag) Itax=f(In) Stax=f(Se) Hhtax=f(Inc) Aid=f(ForAid) Sum GovRoW AgM=f(AG) InM=f(In) SeM=f(Se) DM=f(Inc,Exr) Gpay=f(debt) Sum RoWTotal Sum Ag Sum In Sum Se Sum HH Sum Gov Sum RoW
Broader Systemic Models: T21
• Addresses the WHOLE system, including, economic, social and environmental factors
• Takes account of interactions across the WHOLE system• Generates long-term scenarios to show effects over time• Helps users analyze and understand how national systems function
Production
Investment
Capital
Income
Consumption Loans/debt
Health, Edu., Fam.Planning
Educationlevel
Population
Labor force
Labor productivity
Liveexpenctancy
Pollution controlResourceconservation
Architecture
37
Step 1: Refinement of focus issuesStep 2: Discussion on key elements to be considered,
via a series of open sessionsStep 3: Elaboration of results from open sessions into a
simulation modelStep 4: Testing and validating the modelStep 5: Analysis and discussion of results
Implementation process outlineImplementation process outline
38
KeyKey SuccessSuccess FactorsFactors
1. Solid Modela) Datab) Participation
2. Local Modeling Capacitya) Trainingb) Practical use
3. Local Ownershipa) Commitmentb) Ongoing Development and Use
39
Looking More Closely at T21
• Original systemic model applied to sustainable development
• More applications and experience adapting transparently to countries
• Includes deeper coverage of important non-economic social factors, environment, MDGs, poverty accounting
• Easier to use and less expensive
T21 Fits into Planning Toolkits
• Macro modelsProvide Macro Balances, MTEF, IFI discussionsShort term -- need longer-term, x-sector validation
• CGE ModelsSAM, Detailed relations, Optimum effectsComparative static -- need more transparent paths
• Threshold 21Long term, Cross sector links, Transparent resultsNot as detailed, builds on local data and input from other tools
The model was originally built for serving three purposes:
(1) Studying mid-long term development issues
(2) Testing alternative policies
(3) Enhancing learning about system
=> Support mid-long term planning through understanding of the system
Focus of the Threshold 215. Example: T215. Example: T21
1. Consistency check of data and assumptions
2. Identification of future potential issues
3. Identification of alternative strategies4. Basis for monitoring and evaluation
Benefits from using T215. Example: T215. Example: T21
• Mid-long term approach: does not focus on short-term dynamics
• National perspective: does not consider diversity among different regions
• Medium-high level of aggregation: parameters are averaged by sector
• Requires active involvement of client in definition of model’s structure
Limitations of T21 approach5. Example: T215. Example: T21
Key Messages (1)• NDP is a medium to long term planning activity• NDP needs formal models to speed-up
learning process• NDP models should:
– endogenously represent key variables;– be comprehensive;– properly represent dynamic complexity;– be transparent.
• SD is well suited to develop models in accordance to the above criteria.
• T21 is built using the SD approach, and it is rapidly diffusing worldwide.
6. Summary6. Summary
Key Messages (2)• T21 is the results synthesis of best models and
internal• T21 is innovative in the way sectors are linked
together• T21 is useful at four levels in the planning
process:– Check of data and assumptions– Identification of future potential issues– Identification of alternative strategies– Basis for monitoring and evaluation
6. Summary6. Summary
Introduction to System Dynamics
Introduction to System Dynamics
The objective of System Dynamics is:• To improve our understanding of the
interdependencies existing between the structure of a system and its behavior and the extent to which various policies influence its functioning mechanisms. Such policies can then eventually be used as levers for future development.
Models and Methodologies
Type of models:• Econometric; • Geographical maps;• Behavioral;• Language;• Mental;• …
Mental Models
Within System Dynamics, a “mental model” is defined as:
• Our beliefs and theories on causes and effects that define and underlie the structure and behavior of a system,with the limitations/boundaries of the model.
Foundations of System Dynamics and T21: Stocks and FlowsStock and Flows• Stock: accumulations ruled by a flow;• Flow: the rate of change of a stock.
ExamplesPopulation (stock);Fertility rate (positive flow);Mortality rate (negative flow).Money in a bank account (stock);Interest rate on the same account (flow).
Stocks and flows
The stock describes the actual situation
The flow changes the stock and the actual state of the system
Applications Worldwide
T 21 Countries
MI Partner
MI Head Office
MEG Countries
Why Take a Systemic View?
To Avoid Unexpected Results!
Delays
Velocity
Strength
Vensim interface
Vensim interface
Vensim interface
Motivation for this study
• There is a need for integrated tools that could serve as a mean to close the gap between dynamic and all embracing thinking and static available models;
• These tools are required when facing critical issues such as the upcoming energy transition and climate change, because conventional modeling tools do not examine their broader causes and impacts.
Contextualizing issues
The approach proposed includes: • (1) the analysis of the context in which energy
issues arise, whether they are global, regional and national, and
• (2) the study of various policy options that are being considered for solving energy, environmental and national security issues (which are normally implemented at the national level and have narrow boundaries).
Core Capabilities
• Illustration of the synergies and implications of different options across a broad framework
• Provision of a basis for productive long-term planning and unite various parties around consistent policies
• Deeper understanding of the interrelations existing among critical issues
• Support for the creation of cooperation among stakeholders at the planning and technical level
Some results
• Emergence of various unexpected side effects is likely;• Elements of policy resistance arise over the medium
and longer term due to the interrelations existing between energy and society, economy and environment;
• Side effects or unintended consequences may arise both within the energy sector and in the other spheres of the model; nevertheless, these behavioral changes influence all society, economy and environment spheres.
T21-Ecuador - Sectors
Concept – Energy Sector
Macro Feedbacks – Energy Sector
Policies Analyzed
• Subsidizing electricity prices– To reduce households’ costs
• Investing 1% of GDP in energy efficiency– To reduce electricity demand and energy costs
• Investing in Renewable Energy– To reduce thermal electricity generation and GHG
emissions from the power sector and export more oil• Increasing Electricity Imports
– To further reduce domestic thermal generation and export oil, also to reduce natural gas trafficking
Some Results
• A: Investing 1% of GDP in energy efficiency– Higher income– Higher gov. revenues (from oil) and GDP– Lower energy demand, but increasing– Lower emissions
• B: A + Investing in renewable energy– Same income and GDP– Higher employment and lower emissions
• Avoided costs and added gov. revenues are reinvested in social services (edu and infrastructure)
GHG Emissions
Why new insights?
• Side effects or unintended consequences arise from within the energy sector and influencing both the same sector as well as society, economy and environment.
• Results emerge from a combination of:– Four integrated “spheres”; – The representation of feedback, nonlinearity and delays; – A participatory and transparent approach.
The approach used contributes to the representation and understanding of the context (social, economic, environmental and political) in which issues arise and within policies are formulated and implemented.