Embed Size (px)
Citation preview
The fragmentation of production
amplifies systemic risk in supply
chains
IIASA
27 September 2018
Celian Colon (World Bank, Washington DC and
Ecole Normale Supérieure, Paris)
Joint work with Åke Brännström,
Elena Rovenskaya and Ulf Dieckmann (IIASA)
E&P Seminar, ENS Ulm, Paris — 12/05/2016IIASA — 27/09/2018 | 2
Disruptions can propagate through supply chainsComputer supply chain
Nikom Rojna industrial estateThailand, October 2011
Nikom Rojna industrial estateThailand, October 2011
▪ Production disruptions propagate through supply chains▪ Empirical evidence (e.g., Barrot & Sauvagnat 2016)
▪ Indirect losses of natural disasters often exceed direct loss (Hallegatte 2014)
▪ For businesses, a perception of rising systemic risk▪ Managers and insurers loosing track of risk propagation (e.g., Goldin 2010 & pers. commun.)
▪ A quest for supply chain resilience in the business management literature (e.g., Sheffi 2005)
Introduction
E&P Seminar, ENS Ulm, Paris — 12/05/2016IIASA — 27/09/2018 | 3
How fragmentation affect systemic risks?
▪ A trend towards global outsourcing1. Complex supply chains: more firms, more interconnected (Osadchiy et al. 2016)
▪ Models linking network structure and disruption propagation (e.g., Coluzzi et al. 2011)
2. Fragmented supply chains: production stages split between many firms (Hummels et al. 2001)
▪ Gap: How does fragmentation influence systemic risks?
▪ Risk-management decisions are interdependent in supply chains▪ Decisions taken by one firm modify the risk exposure of the other firms
▪ Operation-research models use game theory to elicit optimal strategies (Snyder et al. 2016)
▪ Method limited to very small supply chains
▪ A stylized model with evolutionary dynamics▪ Supply chains subject to random disruption (e.g., Weisbuch & Battiston 2009)
▪ Firms adapt their risk-mitigating strategy to the level of fragmentation
▪ Evolutionary game on networks (Szabo & Fath 2009) with coalitions
Introduction
E&P Seminar, ENS Ulm, Paris — 12/05/2016IIASA — 27/09/2018 | 4
Final consumers, demand d
Raw materials, unlimited
Model formulation
Connectivity matrix M
Technology: Linear production function with productivity z > 1:
Orders are equally split among suppliers (full substitutability).
Model formulation, I — Input–output network
Inventory with durability δ : A fraction δ ≥ 0 of unused inputs is stored:
z = 2
10 k€
5 k€
Risk mitigation: Overorder at rate η ≥ 0:
7.5 k€
2.5*0.8 = 2 k€η = 50%
δ = 80%
Shocks: At each time step, firms get perturbed with probability f, called the failure rate
E&P Seminar, ENS Ulm, Paris — 12/05/2016IIASA — 27/09/2018 | 5
Final consumers
Raw materials
Model formulation
Model formulation, II — Supply disruptions
0 20 400
0.5
1Firm 1
0 20 400
0.5
1Firm 2
0 20 400
0.5
1Firm 3
0 20 400
0.5
1Firm 4
0 20 400
0.5
1Firm 5
0 20 400
0.5
1Firm 6
0 20 400
0.5
1Firm 7
0 20 400
0.5
1Firm 8
0 20 400
0.5
1Firm 9
0 20 400
0.5
1Firm 10
0 20 400
0.5
1Firm 11
0 20 400
0.5
1Firm 12
0 20 400
0.5
1Firm 13
0 20 400
0.5
1Firm 14
0 20 400
0.5
1Firm 15
0 20 400
0.5
1Firm 16
0 20 400
0.5
1Firm 17
0 20 400
0.5
1Firm 18
0 20 400
0.5
1Firm 19
0 20 400
0.5
1Firm 20
0 20 400
0.5
1Firm 21
0 20 400
0.5
1Firm 22
0 20 400
0.5
1Firm 23
0 20 400
0.5
1Firm 24
0 20 400
0.5
1Firm 25
0 20 400
0.5
1Firm 26
0 20 400
0.5
1Firm 27
0 20 400
0.5
1Firm 28
0 20 400
0.5
1Firm 29
0 20 400
0.5
1Firm 30
0 20 400
0.5
1Firm 1
0 20 400
0.5
1Firm 2
0 20 400
0.5
1Firm 3
0 20 400
0.5
1Firm 4
0 20 400
0.5
1Firm 5
0 20 400
0.5
1Firm 6
0 20 400
0.5
1Firm 7
0 20 400
0.5
1Firm 8
0 20 400
0.5
1Firm 9
0 20 400
0.5
1Firm 10
0 20 400
0.5
1Firm 11
0 20 400
0.5
1Firm 12
0 20 400
0.5
1Firm 13
0 20 400
0.5
1Firm 14
0 20 400
0.5
1Firm 15
0 20 400
0.5
1Firm 16
0 20 400
0.5
1Firm 17
0 20 400
0.5
1Firm 18
0 20 400
0.5
1Firm 19
0 20 400
0.5
1Firm 20
0 20 400
0.5
1Firm 21
0 20 400
0.5
1Firm 22
0 20 400
0.5
1Firm 23
0 20 400
0.5
1Firm 24
0 20 400
0.5
1Firm 25
0 20 400
0.5
1Firm 26
0 20 400
0.5
1Firm 27
0 20 400
0.5
1Firm 28
0 20 400
0.5
1Firm 29
0 20 400
0.5
1Firm 30
Profit of firm 2
0 20 400
0.5
1Firm 1
0 20 400
0.5
1Firm 2
0 20 400
0.5
1Firm 3
0 20 400
0.5
1Firm 4
0 20 400
0.5
1Firm 5
0 20 400
0.5
1Firm 6
0 20 400
0.5
1Firm 7
0 20 400
0.5
1Firm 8
0 20 400
0.5
1Firm 9
0 20 400
0.5
1Firm 10
0 20 400
0.5
1Firm 11
0 20 400
0.5
1Firm 12
0 20 400
0.5
1Firm 13
0 20 400
0.5
1Firm 14
0 20 400
0.5
1Firm 15
0 20 400
0.5
1Firm 16
0 20 400
0.5
1Firm 17
0 20 400
0.5
1Firm 18
0 20 400
0.5
1Firm 19
0 20 400
0.5
1Firm 20
0 20 400
0.5
1Firm 21
0 20 400
0.5
1Firm 22
0 20 400
0.5
1Firm 23
0 20 400
0.5
1Firm 24
0 20 400
0.5
1Firm 25
0 20 400
0.5
1Firm 26
0 20 400
0.5
1Firm 27
0 20 400
0.5
1Firm 28
0 20 400
0.5
1Firm 29
0 20 400
0.5
1Firm 30
Firm level
Profit of firm 6
Profit of firm 8
Time
Time
Total profit
Supply-chain level
We evaluate the average profit loss L of the supply chain compared to a no-risk scenario
Loss incurred by firms that are externally perturbed
Loss due to the propagation of supply disruptions
L = Direct Loss + Indirect Loss
E&P Seminar, ENS Ulm, Paris — 12/05/2016IIASA — 27/09/2018 | 6
Model formulation, III — Evolution of strategy
▪ The n firms are allocated to g groups
▪ Fragmentation =
▪ Each firm adjust its overordering rate ηi to increase the profit of its group▪ Evolutionary process based on gradient ascent
▪ Each firm tries and tests different rates and picks the one that increases profits
▪ The process is iterated until a stationary state is reached
Final consumers
Raw materials
Fragmentation0% 100%
5/9 = 56%
Model formulation
Example of group configuration
E&P Seminar, ENS Ulm, Paris — 12/05/2016IIASA — 27/09/2018 | 7
Example of a fragmented chain, I — Differentiated strategies
Result
Productivity z = 2Durability of inventory δ = 50%Failure rate f = 10%
Individual firms
E&P Seminar, ENS Ulm, Paris — 12/05/2016IIASA — 27/09/2018 | 8
Example of a fragmented chain, II — Risk mitigation
Result
Relative reduction in indirect loss
Indirect loss≈ 2*(Direct loss)
Indirect loss≈ 1.2*(Direct loss)
Productivity z = 2Durability of inventory δ = 50%Failure rate f = 10%
E&P Seminar, ENS Ulm, Paris — 12/05/2016IIASA — 27/09/2018 | 9
Fragmentation amplifies systemic risks
Fragmentation diminishesrisk mitigation…
…by reducing incentives to overorder
Overorders
Higher sales
Reduced supply disruptions
Result
Averages over 20 networks of 30 firms, and 200 fragmentation scenarios
E&P Seminar, ENS Ulm, Paris — 12/05/2016IIASA — 27/09/2018 | 10
Supply chain mapping helps identify mitigation benchmarks
Suppose a decision-maker could impose the overordering rate based on objective criteria, what level of mitigation success could be reached?
Result
E&P Seminar, ENS Ulm, Paris — 12/05/2016IIASA — 27/09/2018 | 11
Supply chain mapping helps identify mitigation benchmarks
Suppose a decision-maker could impose the overordering rate based on objective criteria, what level of mitigation success could be reached?
Result
E&P Seminar, ENS Ulm, Paris — 12/05/2016IIASA — 27/09/2018 | 12
Concluding remarks
▪ Fragmentation, inventories and risks▪ More fragmented supply chains (Hummels et al. 2001) & lower inventories (Goldin
2010)
▪ Our model suggests that both trends may be linked: fragmentation disincentivises inventories.
▪ Risks are transferred from individual firms to the production system.
▪ A coming role for insurers?▪ There is a growing demand for supply chain insurance (Munsch 2013 & pers. com.)
▪ Insurers inherit the complexity of the system.
▪ Supply chain mapping helps provide benchmarks for mitigating risks.
Discussion
