Rethinking Economics Based on Complexity Theory The Economy Is an - - PowerPoint PPT Presentation

Rethinking Economics Based on Complexity Theory

Dirk Helbing (ETH Zurich)

„The Economy Is an Equilibrium System“ Going beyond Equilibrium Economics

As Coupling Gets Stronger, System Behavior Can Become Unstable: Crowd Disasters

At low densities: self-organized lane formation, like Adam Smith’s invisible hand At large densities: coordination breaks down Love Parade Disaster in Duisburg, 2010

As Coupling Gets Stronger, System Behavior Can Become Unstable: Traffic Breakdowns

Thanks to Yuki Sugiyama At high densities, free traffic flow is unstable: Despite best efforts, drivers fail to maintain speed Capacity drop, when capacity is most needed!

John D. Sterman’s Beer Game

Perturbations in demand amplify

Network structure Commodity flow (average of FRA, GER, JAP , UK, USA)

• D. H., U. Witt, S. Lämmer, T. Brenner, Physical Review E 70

70, 056118 (2004).

Supply Network of Economic Sectors

Business cycles because of the structure of production networks?

Input output matrix Related delivery network Resulting oscillations in the gross domestic product

Investigation of the effect of network structure:

§

Positive Positive and negative feedbacks negative feedbacks in production processes

§

Time lags ime lags in the information flow and adaptation process

Business Cycles as Result of Network Flows

„More Liquidity is Better“

The Need for Bubble Control

Mode of heat transfer depends on energy input: 1. diffusion 2. convection rolls 3. bubbles Does liquidity play a similar role?

„(Financial) Markets Are Efficient“ Going Beyond the Efficient Market Paradigm

The Flash Crash on May 6, 2010

The flash crash turned solid assets into penny stocks within minutes. Was an interaction effect, no criminal act, ‘fat finger’, or error. 600 billion dollars evaporated in 20 minutes

Google Data Reveal Stock Market Inefficiency

Time, t [Years]

2004 2005 2006 2007 2008 2009 2010 2011 A B

Cumulative Return, R [%]

−40 −20 20 40 60

16 16 ’culture’

100 200 300

Buy and Hold Strategy Google Trends Strategy μ + σ, μ, μ − σ of Random Strategy 312 16 ’debt’

US Search Volume Data

A

Cumulative Return, R [%]

’housing’ ’derivatives’ ’leverage’ ’credit’ ’investment’ ’marriage’ ’hedge’ ’nyse’ ’religion’ ’dow jones’ ’house’ ’color’ ’revenue’ ’consumption’ ’present’ ’trader’ ’office’ ’default’ ’unemployment’ ’stock market’ ’forex’ ’bubble’ ’car’ ’invest’ ’chance’ ’movie’ ’food’ ’return’ ’gains’ ’tourism’ ’lifestyle’ ’short selling’ ’gold’ ’metals’ BUY AND HOLD ’buy’ ’politics’ ’garden’ ’dividend’ ’rare earths’ ’water’ ’short sell’ ’economy’ ’home’ ’world’ ’conflict’ ’oil’ ’transaction’ ’labor’ ’ring’ −50 50 100 150 200 250 ’happy’ ’stocks’ ’profit’ ’cancer’ ’society’ ’portfolio’ ’markets’ ’money’ ’bonds’ ’greed’ ’holiday’ ’success’ ’economics’ ’banking’ ’fine’ ’restaurant’ ’war’ ’crisis’ ’inflation’ ’risk’ ’nasdaq’ ’opportunity’ ’returns’ ’financial markets’ ’crash’ ’growth’ ’sell’ ’consume’ ’finance’ ’cash’ ’health’ ’kitchen’ ’travel’ DOW JONES ’headlines’ ’gain’ ’culture’ ’loss’ ’earnings’ ’fun’ ’freedom’ ’fed’ ’train’ ’arts’ ’rich’ ’energy’ ’fond’ ’environment’ ’ore’ ’debt’

Entire Time Period

μ − σ of Random Strategy μ + σ of Random Strategy μ + 2σ μ + 3σ μ + 4σ

Random Time Periods

Cumulative Return, R [%]

’debt’ ’housing’ ’stocks’ ’credit’ ’markets’ ’portfolio’ ’cancer’ ’opportunity’ ’religion’ ’investment’ ’present’ ’revenue’ ’success’ ’house’ ’consumption’ ’returns’ ’holiday’ ’greed’ ’office’ ’movie’ ’return’ ’financial markets’ ’crisis’ ’finance’ ’trader’ ’chance’ ’invest’ ’cash’ ’food’ ’culture’ ’restaurant’ ’politics’ ’gains’ ’fun’ DOW JONES ’dividend’ ’sell’ ’buy’ ’travel’ ’loss’ ’health’ ’water’ ’short sell’ ’conflict’ ’energy’ ’fond’ ’arts’ ’rich’ ’environment’ ’labor’ ’ring’ −20 20 40 60 ’happy’ ’derivatives’ ’marriage’ ’profit’ ’leverage’ ’society’ ’nyse’ ’bonds’ ’banking’ ’hedge’ ’war’ ’stock market’ ’dow jones’ ’color’ ’risk’ ’money’ ’economics’ ’nasdaq’ ’unemployment’ ’ ’growth’ ’forex’ ’inflation’ ’crash’ ’bubble’ ’metals’ ’fine’ ’lifestyle’ ’default’ ’short selling’ ’gold’ ’headlines’ ’gain’ ’car’ ’earnings’ ’consume’ ’kitchen’ BUY AND HOLD ’rare earths’ ’tourism’ ’fed’ ’economy’ ’train’ ’freedom’ ’garden’ ’home’ ’oil’ ’world’ ’transaction’ ’ore’ μ − σ of Random Strategy μ + σ of Random Strategy

B

μ + 2σ

It would have been possible to make >300% profit during the financial crisis

Joint work with Tobias Preis, Suzy Moat, and H. Eugene Stanley

Is all relevant information in the price?

„Selfish Behavior Optimizes the Systemic Outcome“ When Does Adam Smith‘s Principle of the Invisible Hand Work?

Self-Organized Traffic Light Control

Measurement input Inspiration: Self-

• rganized oscillations

at bottlenecks Optimal compromise between coordination and local flexibility Published in JSTAT (2008)

Local Travel Time Minimization Does Not Optimize Traffic Flows

With constant arrival rates: The switching sequence adapts to the arrival patterns. We observe a flexible switching flexible switching regime with maximum r maximum red-times ed-times.

„Individuals and Companies Decide Rationally“ Going Beyond the Rational Agent Paradigm

Experimental Ultimatum Game with Power: Proposals Offered and Accepted

hyperfair proposal fair proposal rational proposal

Rational responder always rejects Fairness-oriented responder accepts Rational responder always accepts Fairness-oriented responder rejects Rational responder accepts Fairness-oriented responder accepts Rational responder rejects Fairness-oriented responder rejects

16% rejections 37% rejections 27% rejections

Fairness-oriented responder rejects Fairness-oriented responder accepts

Consensus and Overconfidence Undermine the Wisdom of Crowds

„The Mean Behavior Is Representative“ „Agents Can be Treated as if Acting Identically“

Something has changed…

People are linked and influence each other…

How Second-Order Free-Riders Are Eliminated +Punishment Spreads

D = Defectors (Free-Riders), M = Moralists, I=Immoralists C = Non-punishing Cooperators (Second-Order Free-Riders)

Overcoming the Tragedy of the Commons by Spatial Interactions

„Competition Eliminates Non-Selfish Behavior“ The Evolution of Moral Behavior

Evolution of Other-Regarding Preferences by Biological Selection

• R. Murphy et al.:

There is a broad spectrum of non- selfish behavior

Joint work with Thomas Grund

„More Networking is Good and Reduces Risks“

Different recipes, new solutions, and a paradigm shift in

• ur understanding of the world are needed.

Too Much Networking Can Cause Self- Destabilization: Breakdown of Cooperation

How the Banking Network Changed

From: Haldane

Bankruptcy Cascades of Banks

400 600 800 1000 1200 1400 1600 1800 2000

time

10

10

10

12

10

14

10

16

10

18

log of the aggregate output

Cascade size N u m b e r

• f
• c

c u r a n c e s x=1 x=5 x=10 x=49

1 2 5 10 49 50

interbank linkages

10 20 30 40 50

average number of surviving banks

Cascade size increases with interbank linkages

Joint work with G. Tedeschi, A. Mazloumian, and M. Gallegati

Cascading Effects During Financial Crises

Video by Frank Schweitzer, Stefano Battiston, et al.

„Regulation Can Fix the Imperfections of Economy“

Limits of Predictability

§ Large number of non-linearly interacting

system components leads to complex dynamics

§ Example: Weather forecast § Chaotic dynamics/butterfly effect/

sensitivity: Small initial differences can cause very different behavior

Source:

• J. D. Murray

§ Big changes may have small, no, adverse or unexpected effects § Irreducible randomness: A degree of uncertainty and perturbation that

cannot be eliminated

§ Principle of Le Chatelier/Goodhart’s law: A system tends to counteract

external control attempts

§ Delays may cause instabilities § Regime shifts („phase transitions“,

catastrophes): Sometimes small changes have a big impact

§ Unknown unknowns, structural instability

“

The Illusion of Control

34

2 1 2 2 2 2 2 2 2 2 1 1 1 1 1 1 1 1

) ( ) ( x x x x c x k dt dx x x x x c x k dt dx ! ! + " " " = " " " =

The Lotka-Volterra equations are used to describe, for example,

interactions in ecological systems

x1: number of species 1 x2: number of species 2 x0

1,2: attraction point

k1,2, c1,2, λ1,2: parameters The Challenge to Manage Complex Systems

Centralized Control and Its Limits

• Advantage of centralized control is large-

scale coordination

• Disadvantages are due to

– vulnerability of the network – information overload – wrong selection of control parameters – delays in adaptive feedback control

• Decentralized control can perform better in

complex systems with heterogeneous elements, large degree of fluctuations, and short-term predictability, because of greater flexibility to local conditions and greater robustness to perturbations

Logistic target achievement Level of autonomy of control (Windt, Böse, Philipp, 2006)

Managing Complexity: Modifying Interactions Allows to Promote Favorable Self-Organization

Some Grand Challenges (‚Hilbert Problems‘)

DH (2012) Accelerating scientific discovery by formulating grand scientific challenges

• 1. Socio-Economic and Real-World Challenges
• 2. Measurement and Methodological Issues
• 3. Understanding Underlying Principles
• 4. Modeling of Complex Adaptive Agents
• 5. Interactions, Networks, and Systemic

Interdependencies

• 6. Dynamics and Change
• 7. Self-Organization, Collective Behavior, Co-

Evolution and Emergence

• 8. Risks, Systemic Instability, and Resilience
• 9. Markets, Transaction and Exchange Systems
• 10. Institutions and Integrative Systems Design
• 11. Management of Complexity and Sustainability
• 12. Information, Innovation, and Knowledge
• 13. Social and Economic Capital
• 14. Values, Responsibility, and Ethics

„The Price is the Quantity that Matches Supply and Demand“

Price of hotel rooms as a function of room availability

Joint work with Aki-Hiro Sato

Note that there is no equilibrium price, when herding sets in Atypical price formation Price increases with demand, decreases with availability

Coordination Game with Network Interactions and Transaction Costs

Classical coordination game

Coordination failures are likely, but preserve diversity. However, diversity is more vulnerable than the representative agent approach suggests.

Emergence of Social Norms: Theoretical Results

Population 1 sets the norm Population 2 sets the norm Everyone tends to do what he/she likes (anomie) Strength of population 1 Reward of showing preferred behavior / Reward of conforming

Computer simulations: Red = individuals preferring behavior 1 Yellow = individuals adjusting to behavior 1 Blue = individuals preferring behavior 2 Green = individuals adjusting to behavior 2

Local cultures

Superexponential Bubbles or Crashes Instead of Stable Growth

Black Monday in 1987 was a 35 sigma event!) Credits to Didier Sornette

Further Reading

Thanks to all supporters!