SlideShare uma empresa Scribd logo
1 de 14
Stability of the World Trade
       Web over time.


 Scott Pauls
 Department of Mathematics
 Dartmouth College

           Conference on Emergent Risk,
               Princeton University
              September 28-29,2012
Stability of the World Trade
       Web over time.


 Scott Pauls
 Department of Mathematics
 Dartmouth College

           Conference on Emergent Risk,
               Princeton University
              September 28-29,2012
Systemic and emergent risk
“Whereas systemic risk is the threat that
individual failures or accidents
represent to a system through the
process of contagion, emergent risk is
the threat to the individual parts
produced by their participation in and
interaction with the system itself.”
(Centeno and Tham)

              Conference on Emergent Risk,
                  Princeton University
                 September 28-29,2012
Stability and robustness
Structure
                      We use network models:
             actors are nodes, relationships are edges.


Dynamics
                 We construct dynamics to model
                  exchanges between the actors.

Robustness

                 We define robustness in terms of
                      responses to shocks.

                    Conference on Emergent Risk,
                        Princeton University
                       September 28-29,2012
World Trade Web
 nodes are countries.

 edges are directed and
 weighted, giving the
 dollars that flow from
 country i to country j
 for traded goods.

 dynamics are given by
 the Income-Expenditure
 model.


Barbieri, Katherine, Omar M. G. Keshk, and Brian Pollins. 2009. “TRADING DATA: Evaluating our
Assumptions and Coding Rules.” Conflict Management and Peace Science. 26(5): 471-491.

                                  Conference on Emergent Risk,
                                      Princeton University
                                     September 28-29,2012
Income-Expenditure model

                                              propensity to spend

                                                            debt


Markov model




               Conference on Emergent Risk,
                   Princeton University
                  September 28-29,2012
Attacks on the system
Edge deformation: policy decisions, sharp trade
              evolution.

Bilateral edge deletion: war, collapse of trade
                      agreement.

Node deformation: internal collapse (e.g. bhat
                  collapse in the 1980s)

Node deletion: unrealistic but useful as a type of worst
               case scenario

Maximal Extinction Analysis (MEA): really a worst case
                                   scenario!
                      Conference on Emergent Risk,
                          Princeton University
                         September 28-29,2012
Power and robustness


                                       Income after
                                       rebalancing




                                     Total initial income




      Conference on Emergent Risk,
          Princeton University
         September 28-29,2012
WTWs are
“robust yet fragile”

Left hand side:
TARGETED ATTACK
The strength of maximal
attacks of each type.
Colored bars (and circles)
indicate significance.


Right hand side:
RANDOM ATTACK
Circles indicate the
proportion of all possible
attacks which are not
significant.


                             Conference on Emergent Risk,
                                 Princeton University
                                September 28-29,2012
The role of connectance




       Conference on Emergent Risk,
           Princeton University
          September 28-29,2012
The role of connectance




       Conference on Emergent Risk,
           Princeton University
          September 28-29,2012
A closer look

                                 U.S./Canada
                                     link




                                     U.S.
                                 deformation




  Conference on Emergent Risk,
      Princeton University
     September 28-29,2012
Conclusions and the big picture
We see evidence that increased connectance has two effects related to
systemic risk.

1.   On one hand, denser connections allow for more paths through
     which shocks may be mitigated.
2.   But, on the other, denser connection patterns provide more paths
     along which collapse can spread.

These two are in tension.

With regard to emergent risk, we see an additional wrinkle related to
connectance coupled with the topology of the network.

3.   Denser connections allow for propagation of shocks which, while
     possibly mitigated overall, can have adverse impact on individual
     countries.


                            Conference on Emergent Risk,
                                Princeton University
                               September 28-29,2012
Emergent and Systemic risk
In our model, the tension is resolved in different ways depending on the size of the
shock.

Systemic risk

a.   Smaller shocks are easily absorbed into the system (and sometimes result in
     income increases!).
b.   But, there is a tipping point above which the larger shocks spark a substantial
     contagion effect.

Emergent Risk

c.   Even with smaller shocks, we see evidence that mere participation in the WTW
     brings new risk.
d.   Large shocks amplify this risk.


We need a new lexicon to describe these types of networks.




                               Conference on Emergent Risk,
                                   Princeton University
                                  September 28-29,2012

Mais conteúdo relacionado

Semelhante a Stability of the world trade web over time

HOW REVERSIBILITY DIFFERENTIATES CYBER FROM KINETIC WARFARE: A CASE STUDY IN ...
HOW REVERSIBILITY DIFFERENTIATES CYBER FROM KINETIC WARFARE: A CASE STUDY IN ...HOW REVERSIBILITY DIFFERENTIATES CYBER FROM KINETIC WARFARE: A CASE STUDY IN ...
HOW REVERSIBILITY DIFFERENTIATES CYBER FROM KINETIC WARFARE: A CASE STUDY IN ...ijsptm
 
CT IRP Grid Security 9 20 11rev1
CT IRP Grid Security 9 20 11rev1CT IRP Grid Security 9 20 11rev1
CT IRP Grid Security 9 20 11rev1jgordes
 
Society for Risk Analysis On Transnational Risk & Terrorism
Society for Risk Analysis  On Transnational Risk & TerrorismSociety for Risk Analysis  On Transnational Risk & Terrorism
Society for Risk Analysis On Transnational Risk & TerrorismJohn Marke
 
Datashop Alchemy
Datashop  AlchemyDatashop  Alchemy
Datashop AlchemyInnovAccer
 
Capstone Final Draft Rev 2 - The Cyber-Security Dilemma_ The ’Cyber-Army’ Bui...
Capstone Final Draft Rev 2 - The Cyber-Security Dilemma_ The ’Cyber-Army’ Bui...Capstone Final Draft Rev 2 - The Cyber-Security Dilemma_ The ’Cyber-Army’ Bui...
Capstone Final Draft Rev 2 - The Cyber-Security Dilemma_ The ’Cyber-Army’ Bui...James Creamer III
 
Differentiated social risk Rebound dynamics and sustainability performance in...
Differentiated social risk Rebound dynamics and sustainability performance in...Differentiated social risk Rebound dynamics and sustainability performance in...
Differentiated social risk Rebound dynamics and sustainability performance in...Sandy Worden
 
Whose Global Pulse: Diagnosing Sickness or Health?
Whose Global Pulse: Diagnosing Sickness or Health?Whose Global Pulse: Diagnosing Sickness or Health?
Whose Global Pulse: Diagnosing Sickness or Health?Global Pulse
 
Innovat Increas Risks Cbr Disast Resp Mjd29sept11
Innovat Increas Risks Cbr Disast Resp Mjd29sept11Innovat Increas Risks Cbr Disast Resp Mjd29sept11
Innovat Increas Risks Cbr Disast Resp Mjd29sept11martindudziak
 
Scott 校外口試
Scott 校外口試Scott 校外口試
Scott 校外口試jilung hsieh
 
Disaster resilient societies
Disaster resilient societiesDisaster resilient societies
Disaster resilient societiesMaricica Botescu
 
WORK & STRESS, 1998, VOL. 12, NO. 3 293-306 Achieving a sa.docx
WORK & STRESS, 1998, VOL. 12, NO. 3 293-306 Achieving a sa.docxWORK & STRESS, 1998, VOL. 12, NO. 3 293-306 Achieving a sa.docx
WORK & STRESS, 1998, VOL. 12, NO. 3 293-306 Achieving a sa.docxambersalomon88660
 
Supply Chain Risk
Supply Chain RiskSupply Chain Risk
Supply Chain RiskJan Husdal
 
Perfect Storm T Mc Kenna Pmoz 2010 Final
Perfect Storm   T Mc Kenna Pmoz 2010 FinalPerfect Storm   T Mc Kenna Pmoz 2010 Final
Perfect Storm T Mc Kenna Pmoz 2010 FinalSTARTPM
 
Financial Network Analysis and Visualisation - Seminar at NYU/Stern 5 April 2011
Financial Network Analysis and Visualisation - Seminar at NYU/Stern 5 April 2011Financial Network Analysis and Visualisation - Seminar at NYU/Stern 5 April 2011
Financial Network Analysis and Visualisation - Seminar at NYU/Stern 5 April 2011Kimmo Soramaki
 

Semelhante a Stability of the world trade web over time (20)

HOW REVERSIBILITY DIFFERENTIATES CYBER FROM KINETIC WARFARE: A CASE STUDY IN ...
HOW REVERSIBILITY DIFFERENTIATES CYBER FROM KINETIC WARFARE: A CASE STUDY IN ...HOW REVERSIBILITY DIFFERENTIATES CYBER FROM KINETIC WARFARE: A CASE STUDY IN ...
HOW REVERSIBILITY DIFFERENTIATES CYBER FROM KINETIC WARFARE: A CASE STUDY IN ...
 
CT IRP Grid Security 9 20 11rev1
CT IRP Grid Security 9 20 11rev1CT IRP Grid Security 9 20 11rev1
CT IRP Grid Security 9 20 11rev1
 
CSIAC_V1N4_FINAL_2
CSIAC_V1N4_FINAL_2CSIAC_V1N4_FINAL_2
CSIAC_V1N4_FINAL_2
 
Living Resiliently on a Crowding, Turbulent Planet
Living Resiliently on a Crowding, Turbulent PlanetLiving Resiliently on a Crowding, Turbulent Planet
Living Resiliently on a Crowding, Turbulent Planet
 
GDRR Opening Workshop - Network Connectivity and Implications for Systemic Ri...
GDRR Opening Workshop - Network Connectivity and Implications for Systemic Ri...GDRR Opening Workshop - Network Connectivity and Implications for Systemic Ri...
GDRR Opening Workshop - Network Connectivity and Implications for Systemic Ri...
 
PSA - Lezione 28 ottobre 2014 - RISK
PSA - Lezione 28 ottobre 2014 - RISKPSA - Lezione 28 ottobre 2014 - RISK
PSA - Lezione 28 ottobre 2014 - RISK
 
Society for Risk Analysis On Transnational Risk & Terrorism
Society for Risk Analysis  On Transnational Risk & TerrorismSociety for Risk Analysis  On Transnational Risk & Terrorism
Society for Risk Analysis On Transnational Risk & Terrorism
 
Datashop Alchemy
Datashop  AlchemyDatashop  Alchemy
Datashop Alchemy
 
Capstone Final Draft Rev 2 - The Cyber-Security Dilemma_ The ’Cyber-Army’ Bui...
Capstone Final Draft Rev 2 - The Cyber-Security Dilemma_ The ’Cyber-Army’ Bui...Capstone Final Draft Rev 2 - The Cyber-Security Dilemma_ The ’Cyber-Army’ Bui...
Capstone Final Draft Rev 2 - The Cyber-Security Dilemma_ The ’Cyber-Army’ Bui...
 
Differentiated social risk Rebound dynamics and sustainability performance in...
Differentiated social risk Rebound dynamics and sustainability performance in...Differentiated social risk Rebound dynamics and sustainability performance in...
Differentiated social risk Rebound dynamics and sustainability performance in...
 
Whose Global Pulse: Diagnosing Sickness or Health?
Whose Global Pulse: Diagnosing Sickness or Health?Whose Global Pulse: Diagnosing Sickness or Health?
Whose Global Pulse: Diagnosing Sickness or Health?
 
Innovat Increas Risks Cbr Disast Resp Mjd29sept11
Innovat Increas Risks Cbr Disast Resp Mjd29sept11Innovat Increas Risks Cbr Disast Resp Mjd29sept11
Innovat Increas Risks Cbr Disast Resp Mjd29sept11
 
Scott 校外口試
Scott 校外口試Scott 校外口試
Scott 校外口試
 
Disaster resilient societies
Disaster resilient societiesDisaster resilient societies
Disaster resilient societies
 
Systemic Risks and Emerging Challenges.pdf
Systemic Risks and Emerging Challenges.pdfSystemic Risks and Emerging Challenges.pdf
Systemic Risks and Emerging Challenges.pdf
 
WORK & STRESS, 1998, VOL. 12, NO. 3 293-306 Achieving a sa.docx
WORK & STRESS, 1998, VOL. 12, NO. 3 293-306 Achieving a sa.docxWORK & STRESS, 1998, VOL. 12, NO. 3 293-306 Achieving a sa.docx
WORK & STRESS, 1998, VOL. 12, NO. 3 293-306 Achieving a sa.docx
 
Supply Chain Risk
Supply Chain RiskSupply Chain Risk
Supply Chain Risk
 
Perfect Storm T Mc Kenna Pmoz 2010 Final
Perfect Storm   T Mc Kenna Pmoz 2010 FinalPerfect Storm   T Mc Kenna Pmoz 2010 Final
Perfect Storm T Mc Kenna Pmoz 2010 Final
 
Cascading Disasters
Cascading DisastersCascading Disasters
Cascading Disasters
 
Financial Network Analysis and Visualisation - Seminar at NYU/Stern 5 April 2011
Financial Network Analysis and Visualisation - Seminar at NYU/Stern 5 April 2011Financial Network Analysis and Visualisation - Seminar at NYU/Stern 5 April 2011
Financial Network Analysis and Visualisation - Seminar at NYU/Stern 5 April 2011
 

Stability of the world trade web over time

  • 1. Stability of the World Trade Web over time. Scott Pauls Department of Mathematics Dartmouth College Conference on Emergent Risk, Princeton University September 28-29,2012
  • 2. Stability of the World Trade Web over time. Scott Pauls Department of Mathematics Dartmouth College Conference on Emergent Risk, Princeton University September 28-29,2012
  • 3. Systemic and emergent risk “Whereas systemic risk is the threat that individual failures or accidents represent to a system through the process of contagion, emergent risk is the threat to the individual parts produced by their participation in and interaction with the system itself.” (Centeno and Tham) Conference on Emergent Risk, Princeton University September 28-29,2012
  • 4. Stability and robustness Structure We use network models: actors are nodes, relationships are edges. Dynamics We construct dynamics to model exchanges between the actors. Robustness We define robustness in terms of responses to shocks. Conference on Emergent Risk, Princeton University September 28-29,2012
  • 5. World Trade Web nodes are countries. edges are directed and weighted, giving the dollars that flow from country i to country j for traded goods. dynamics are given by the Income-Expenditure model. Barbieri, Katherine, Omar M. G. Keshk, and Brian Pollins. 2009. “TRADING DATA: Evaluating our Assumptions and Coding Rules.” Conflict Management and Peace Science. 26(5): 471-491. Conference on Emergent Risk, Princeton University September 28-29,2012
  • 6. Income-Expenditure model propensity to spend debt Markov model Conference on Emergent Risk, Princeton University September 28-29,2012
  • 7. Attacks on the system Edge deformation: policy decisions, sharp trade evolution. Bilateral edge deletion: war, collapse of trade agreement. Node deformation: internal collapse (e.g. bhat collapse in the 1980s) Node deletion: unrealistic but useful as a type of worst case scenario Maximal Extinction Analysis (MEA): really a worst case scenario! Conference on Emergent Risk, Princeton University September 28-29,2012
  • 8. Power and robustness Income after rebalancing Total initial income Conference on Emergent Risk, Princeton University September 28-29,2012
  • 9. WTWs are “robust yet fragile” Left hand side: TARGETED ATTACK The strength of maximal attacks of each type. Colored bars (and circles) indicate significance. Right hand side: RANDOM ATTACK Circles indicate the proportion of all possible attacks which are not significant. Conference on Emergent Risk, Princeton University September 28-29,2012
  • 10. The role of connectance Conference on Emergent Risk, Princeton University September 28-29,2012
  • 11. The role of connectance Conference on Emergent Risk, Princeton University September 28-29,2012
  • 12. A closer look U.S./Canada link U.S. deformation Conference on Emergent Risk, Princeton University September 28-29,2012
  • 13. Conclusions and the big picture We see evidence that increased connectance has two effects related to systemic risk. 1. On one hand, denser connections allow for more paths through which shocks may be mitigated. 2. But, on the other, denser connection patterns provide more paths along which collapse can spread. These two are in tension. With regard to emergent risk, we see an additional wrinkle related to connectance coupled with the topology of the network. 3. Denser connections allow for propagation of shocks which, while possibly mitigated overall, can have adverse impact on individual countries. Conference on Emergent Risk, Princeton University September 28-29,2012
  • 14. Emergent and Systemic risk In our model, the tension is resolved in different ways depending on the size of the shock. Systemic risk a. Smaller shocks are easily absorbed into the system (and sometimes result in income increases!). b. But, there is a tipping point above which the larger shocks spark a substantial contagion effect. Emergent Risk c. Even with smaller shocks, we see evidence that mere participation in the WTW brings new risk. d. Large shocks amplify this risk. We need a new lexicon to describe these types of networks. Conference on Emergent Risk, Princeton University September 28-29,2012

Notas do Editor

  1. This is from the proposal describing the conference. In the work I’ll describe, our system is a network of trade relationships – measurements of the import/export business that countries do with one another. In that context, we are concerned with both of these types of risk and, in many ways, view them as aspects of a broader measure of risk inherent to the system. Robust participation in trade has obvious benefits in terms of income, availability of goods, mobility, etc. But that same participation creates vulnerability. Distant, and perhaps small, economic events can have disproportionate effects on other members of the trade network, even those not directly impacted by the event. So, in our work, emergent risk (as defined here) has a component of simple exposure to systemic risk but also a component of risk associated to the pattern of interactions across the network. We consider this in the context of globalization, which both pushes countries into the trade network and encourages them to trade more widely. This has structural consequences for the WTW. To what extent does this impact the risk (whether emergent or systemic) of the system?
  2. So, how do we assess these types of risk – our paradigm rests on the analysis of the response of the network to shocks.Given a (well defined) shock to the system, how does the system respond? What is the functional damage to the network?The interaction of the dynamics and the network topology are crucial.Earlier work in this direction: percolation, contagion, robust-yet-fragile categorization of router networks.
  3. The method we use is based on extinction analyses in ecology, primarily those done on food webs. Basic result – increasing connectance (i.e. network density) corresponds to increasing robustness.
  4. Our adjacency matrix is denoted by M - we use the import matrices from the Correlates of War project. 𝑀𝑖𝑗 is the $ amount of imports into country i from country j.Our adjacency matrix is denoted by M - we use the import matrices from the Correlates of War project. 𝑀_𝑖𝑗 is the $ amount of imports into country i from country j.
  5. In strength = $ of income from selling goods, Out strength = $ expended buying goods from others𝛼 is the percentage of income that you will spend,𝛽 is money spend over and above the income𝑚 is the Markov chain associated withM. If everything is computed from a fixed data matrix, then 𝐸𝑡 is constant over time.In strength = $ of income from selling goods, Out strength = $ expended buying goods from others𝛼 is the percentage of income that you will spend,𝛽 is money spend over and above the income𝑚 is the Markov chain associated withM. If everything is computed from a fixed data matrix, then 𝐸_𝑡 is constant over time.
  6. MEA: delete node of highest power, rebalance and repeat until 50% of the income has been removed.
  7. Power measures the strength of a given attack. Suppose an attack changes in total $ from 100,000 to 90,000. Then Power = 1-90000/100000=1-9/10=1/10.Robustness takes the maximum power over all attacks of a given type. Using 1-Power transforms this into a number which is low if power is high and vice versa.
  8. Take-away: for all attacks, targeted attacks are usually significant (i.e. the damage spreads) and for all but node deformations, random attacks are generally not significant. Hence RYF. But there is some nuance, some years buck this trend and node deformations show fragility for both types of attacks.
  9. For the smaller shocks, we see a similar trend to the food web results – as connectance increases, we see increased robustness as well. But, for the edge shocks, there seems to be a decay after a critical juncture at roughly 0.4.The weakest result is (again) for the node deformation.
  10. However, in the face of the maximal extinction analysis we see a different trend – increasing connectance corresponds to decreasing robustness.Notes:There is a structural transition in the 1970s which causes a temporary jump in robustness.the jagged black line is the mean robustness over a family of bootstrap null models. We gauge the significance of the results in terms of different null models. The black circles here are years when 𝑅𝑀𝐸𝐴 is outside of the 5-95% range for the same statistic over 100 bootstrap nulls.However, in the face of the maximal extinction analysis we see a different trend – increasing connectance corresponds to decreasing robustness.Notes:There is a structural transition in the 1970s which causes a temporary jump in robustness.the jagged black line is the mean robustness over a family of bootstrap null models. We gauge the significance of the results in terms of different null models. The black circles here are years when 𝑅_𝑀𝐸𝐴 is outside of the 5-95% range for the same statistic over 100 bootstrap nulls.
  11. Impact ratio = ∑𝐸5−∑𝐸0𝑡𝑜𝑡𝑎𝑙 𝑒𝑑𝑔𝑒 𝑤𝑒𝑖𝑔h𝑡 (RHS)Impact ratio = 𝐸5(𝑖)𝐸0(𝑖) (LHS)Bilateral link deletion: (a) 75% of the attacks do less damage than their edge weight (and some, the negative ones), create more income. 25% of the attacks do more damage. (b) 2006, US/Can is largest impact link, so we look at it closely. These are the same ratios, but country by country after US/CAN is deleted. They are ordered, from smallest to largest, by the mean network distance between the country and the US and Canada. Note that there is little association between these two. This is one indication of emergent risk. No matter how far you are from the epicenter of a crisis, you can have substantial risk just from your participation and the pattern of network interactions in the global network.Node deformation:Different picture, percentages flipped – 25% do less damage, 75% do more. Now, geographical association is more apparent. Impact ratio = (∑𝐸_5−∑𝐸_0)/(𝑡𝑜𝑡𝑎𝑙 𝑒𝑑𝑔𝑒 𝑤𝑒𝑖𝑔ℎ𝑡) (RHS)Impact ratio = (𝐸_5 (𝑖))/(𝐸_0 (𝑖)) (LHS)Bilateral link deletion: (a) 75% of the attacks do less damage than their edge weight (and some, the negative ones), create more income. 25% of the attacks do more damage. (b) 2006, US/Can is largest impact link, so we look at it closely. These are the same ratios, but country by country after US/CAN is deleted. They are ordered, from smallest to largest, by the mean network distance between the country and the US and Canada. Note that there is little association between these two. This is one indication of emergent risk. No matter how far you are from the epicenter of a crisis, you can have substantial risk just from your participation and the pattern of network interactions in the global network.Node deformation:Different picture, percentages flipped – 25% do less damage, 75% do more. Now, geographical association is more apparent.
  12. On b): Substantial questions here – is the tipping point so large that we should never expect shocks of that magnitude (i.e. “too big to fail”)? Do we really have a framework in which to properly evaluate this question? (see, 2008 financial crisis)