Using Grammatical Signals Suitable to Patterns of Idea Development
IDS Impact Innovation and Learning Workshop March 2013: Day 1, Paper Session 2 Peter Loewe
1. Exploring the Potential of
Systems Dynamics Modelling
Impact Innovation and Learning: Towards a Research and
Practice Agenda for the Future
Brighton 26/27 March 2013
Peter Loewe
2. Content
• Impact paths of enterprise development projects
• Linear causal chains vs closed causal loops
• Causal loop diagrams: 3 investment strategies
• System Dynamics past and present
• The UNIDO demonstration tool
• Two simulations
• Status, lessons, way forward
3. Causal chain model of PSD poverty impact
Source: OECD Donor Committee for Enterprise Development
How Private Sector Development leads to Pro-Poor Impacts: A Framework for Evidence
4. Poverty impact paths under scrutiny
Employment
reduction
IUIntervention
Competitiveness
impact
Industrial
Development/
Economic Growth
Impact
Employment
impact
Skill content impact Additional EffectsPoverty impact
Crowding out of less
competitive SMEs
Business decline
Job-less growth
(productivity path)
High skill jobs
Low skill jobs
Non-poor growth
Increase in poverty
Reduction in local
purchasing power,
consumption,
production, SMEs,
and employment
increase in poverty
Market
access &
partnership
Cluster &
networking
Pro-poor skill
development
& training
Pro-poor
social
protection
Pro-poor
targeting
(regions, sectors,
firms, services)
Pro-poor
entrepreneur &
SME
development
Business
maintenance
High skill jobs
Competitive SMEs
(high road/costs &I U)
Business expansion
of local suppliers &
subcontractors
Business expansion
in export markets
Business expansion
in local market
Employment creation
/ job-rich growth
(expansion path) Low skill jobs
Pro-poor
growth
Increase in local
purchasing power,
consumption,
production, SMEs,
and employment
poverty reduction
Employment
maintenance
Business decline of
less competitive local
suppliers &
subcontractors
Competitive SMEs
(low road/costs)
Possible impact drivers are shown in red.
5. Causal Chains vs. Systems Dynamics
• Linear causal chain modeling (logframe)
– state-of-the-art in development cooperation
– too simplistic for complex cases
• System Dynamics Modelling (SDM)
– Causal loops instead of causal chains
– Negative feed-back loops (“goal seeking”)
– Positive feed-back loops (“exponential growth”)
– Interconnected loops
– Strong / weak coupling
– Variables can be “stocks” or “flows”
– Short term / long term behavior
– Non-linear behavior(small causes – big effects)
6. 6
Causal loop example (1)
Profits
Investment
Expansion
Competitivenes
on price
+
+
-
+
Negative feed back
loop ("goal seeking")
7. 7
Causal loop example (2)
Profits
Investment
Rationa
-lization
Expansion
Competitiveness
on price
+
+
-
+
+
+
Exponentialgrowth(buttoothebottom)
Positive feed back loop
(but jobless "low road"))
8. 8
Causal loop example (3)
Profits
Investment
Rationa
-lization
Expansion
Competitiveness
on price
+
+
-
+
+
+
Exponential growth(buttoo the bottom) Positive feed back loop
(but jobless "low road")
Competitiveness on
quality & new products
Innovation
+
-
+
Positive feed back
loop ("high road")
Negative feed back
loop (equilibrium)
9. 9
Entry points for interventions
Profits
Investment
Rationa
-lization
Expansion
Competitiveness
on price
+
+
-
+
+
+
Competitiveness on
quality & new products
Innovation
+
-
+
Business
climate
Access to
finance
Trade
liberalization
Improve national
quality and innovation
system
10. System Dynamics: Past & present
• Early applications
– Engineering: Technological feed-back systems
– Biology: Ecological systems (predator – prey systems)
• 1961: Jay Forrester (MIT) application to management :
“Industrial Dynamics”
• 1972: Forrester/Meadows: “World Model” – Limits of Growth
• 1980s: System Dynamics tools for PC (Ithink; Vensim)
• Current applications: management; traffic/city/regional
planning; energy and environment; economic development
• World System Dynamics Society: www.systemdynamics.org
• 2005: Millennium Institute “Threshold 21” simulation tool
11. 11
I believe we are proposing the “Process” of modeling rather
than particular frozen and final models. … It seems to me that
the average person will be greatly concerned if he feels that
the future and alternatives are being frozen once and for all
into a particular model, instead we are suggesting that
models will help to clarify our processes of thought; they will
help to make explicit the assumptions we are already
making; and they will show the consequences of the
assumptions. But as our understanding, our assumptions,
and our goals change, so can the models.
John Forrester (1985), “The” model versus a modeling process, in: System
Dynamics Review, 1, S. 133-134.
Importance of the modeling process
12. The UNIDO experimental simulation tool
• Part of an evaluation of “Industrial Upgrading” projects
• Concrete case: Leather project in Ethiopia
• Programmer: Sebastian Derwisch (University of Bergen)
• Consultant: Cornelia Staritz (Austrian Development
Foundation)
• Group model building workshop in December 2011
• Computer model using VENSIM software:
– About 200 internal variables / equations
– Three external factors
– Eight input variables (development interventions)
– Eleven output variables
• Presentations to project managers and management
• Recommendation to pursue
14. 3 external factors
• Import tariffs (increased imports)
• Cost of raw materials
• Increased competition on export markets
15. 8 interventions (input variables)
• Investment in equipment
• Labor intensity
• Investment in skills
• Access to credit
• Strengthening of National Quality Infrastructure
• Logistics and customs infrastructure
• “Buy local” campaign
• Promotion of labor standards
16. 11 effects (output variables)
• Equipment
• Skills
• Productivity
• Costs
• Quality
• Price
• Local demand
• International demand
• Production
• Jobs
• Wages
17. 17
"Imports increasing (Tariff reduction)"
0 1
Increase of imports
0.6
0.45
0.3
0.15
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
"Imports increasing (Tariff reduction)" : Baseline with trade liberalization
Represents an increase in competition on the national
market. The parameter can be varied between 0 (no
competition - everything produced for the domestic
market can be sold) and 1 (full competition, nothing
produced for the domestic market can be sold)
Simulation 1
External change: trade liberalization
in 2011 - 2014
Slider to change the parameter between 0 and 1
Simulation period: 2010 to 2035
18. 18
Simulation 1:
Effects of the external change
Costs
2
1.4
0.8
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".expected costs." : Baseline with trade liberalization
Production
1
0.75
0.5
0.25
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".production." : Baseline with trade liberalization
Price
1
0.8
0.6
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".expected price." : Baseline with trade liberalization
International Demand
0.2
0.1
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".international demand." : Baseline with trade liberalization
Local Demand
1
0.5
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".local demand." : Baseline with trade liberalization
Equipment
1
0.7
0.4
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".equipment." : Baseline with trade liberalization
Productivity
2
1.7
1.4
1.1
0.8
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
relative productivity : Baseline with trade liberalization
Quality
1
0.8
0.6
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
score quality : Baseline with trade liberalization
Jobs
1
0.7
0.4
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".jobs." : Baseline with trade liberalization
Wages
1
0.9
0.8
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
wages : Baseline with trade liberalization
Skill per worker
2
1.7
1.4
1.1
0.8
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
relative average skills per worker : Baseline with trade liberalization
19. Promotion of labour standards
0 5
logistics and customs upgrading program
0 5
Equipment upgrading program
0 5
Skills upgrading program
0 5
buy local campaign
0 5
access to credit
0 5
NQS upgrading program
0 5
Investment in Equipment
2
1
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
Equipment upgrading program : 8 Liberalisation response
Logistics and customs upgrading
2
1.4
0.8
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
logistics and customs upgrading program : 8 Liberalisation response
Investment in skills
4
3
2
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
Skills upgrading program : 8 Liberalisation response
NQS upgrading
2
1.4
0.8
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
NQS upgrading program : 8 Liberalisation response
Buy local campaign
2
1.7
1.4
1.1
0.8
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
buy local campaign : 8 Liberalisation response
Labour standards
2
1.5
1
0.5
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
Promotion of labour standards : 8 Liberalisation response
Access to credit
2
1.75
1.5
1.25
1
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
access to credit : 8 Liberalisation response
Interventions reduces labor requirements - Saves
on workforce and maintains production. The
parameter can be varied between -1 (full
automatization, no workers needed) and 1
(labour intensity increased by 100%)
Interventions increases labor productivity by
investments into skill building - increases
productivity of workers. The parameter can be
varied between 0 (no additional investment in
skills) and 5 (investment in skills increased by
500%)
Interventions affecting productivity
labor intensity intervention
-1 1
Intervention increases investment into equipment -
Maintains woker and increases stock of equipment.
The parameter can be varied between 0 (no
additional investment in equipment) and 5
(investment in equipment increased by 500%)
Labor intensity
-0.4
-0.7
-1
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
labor intensity intervention : 8 Liberalisation response
Other Interventions
Represents a campaign that stimulates local
demand - whats inserted is the assumed increase of
local demand by the campaign. The parameter can
be varied between 0 (additional increase in
demand) and 5 (local demand increased by 500%)
Represents an upgrading of wages - the
value inserted represents the increase of the
wages by promoting better pabor standards,
higher wages have an effect on the skill per
worker
Access to credit lifts the overall investment
by the value inserted. The parameter can be
varied between 0 (no additional investment)
and 5 (investment increased by 500%)
NQS upgrading represents investment in NQS
facilities. The parameter canbe varied between
0 (no additional investment into NQS
upgrading) and 5 (investment into NQS
upgrading increased by 500%)
Represents investments to improve logistics
which reduces fluctuations in the delivery delay.
The parameter can be varied between 0 (no
additional investment into logistics) and 5
(investment into logistics increased by 500%)
Simulation 2:
Interventions of a possible development program
Set parameters
for year n
Observe effects
for year n+1
Adapt parameters
for year n+1
Observe effects
for year n+2
Adapt parameters
for year n+3
20. 20
Simulation 2: How the interventions of the response
program overcome the effects of the external change
Costs
2
1
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".expected costs." : 8 Liberalisation response
Production
20
15
10
5
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".production." : 8 Liberalisation response
Price
2
1.4
0.8
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".expected price." : 8 Liberalisation response
International Demand
20
10
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".international demand." : 8 Liberalisation response
Local Demand
10
5
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".local demand." : 8 Liberalisation response
Equipment
6
3
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".equipment." : 8 Liberalisation response
Productivity
4
3
2
1
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
relative productivity : 8 Liberalisation response
Quality
8
4
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
score quality : 8 Liberalisation response
Jobs
2
1.4
0.8
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
".jobs." : 8 Liberalisation response
Wages
2
1.4
0.8
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
wages : 8 Liberalisation response
Skill per worker
8
6
4
2
0
2010 2015 2020 2025 2030 2035
Time (Year)
dmnl
relative average skills per worker : 8 Liberalisation response
21. Work in progress - some preliminary conclusions
• SD: an appropriate approach to cope with complexity
• A “meta language” - alternative to linear causal chains
• “Qualitative modeling” through “Group Model Workshops”:
– Consensus building on parameters & dynamics of complex settings
– Making implicit assumptions explicit
– Feeding “lessons learned” from evaluation into model structure
• Computer simulation (“quantitative modeling”)
– Not a must - “qualitative modeling” is useful exercise in itself
– Requires experienced programmer
– Rather time consuming
• Useful for generic types of interventions (project families)
• Towards an “Artificial Intelligence” tool for program design?
22. 22
Omitting structures or variables
known to be important because
numerical data are unavailable is
actually less scientific and less
accurate than using your best
judgment to estimate their
values.
To omit such variables is
equivalent to saying they have
zero effect - probably the only
value that is known to be wrong!
John Sterman (2002), All models are wrong:
reflections on becoming a systems scientist, System
Dynamics Review, Vol. 18, p. 523
The simple is false
- but the complex
is unusable
Paul Valéry
Also known as
“Bonini’s paradox”
The complexity of our mental models
vastly exceeds our capacity to
understand their implications. …
Formalizing qualitative models and
testing them via simulation often leads
to radical changes in the way we
understand reality.
John Sterman (2000), Business Dynamics, p. 29
„All
models
are
wrong“