Supply Chain Network Structure and Risk Propagation John R. Birge 1 1 University of Chicago Booth School of Business (joint work with Jing Wu, Chicago Booth) Kellogg School of Management, Northwestern U. Birge (Chicago Booth) Supply Chain Structure October 1, 2014 1 / 27
73
Embed
Supply Chain Network Structure and Risk Propagationfaculty.chicagobooth.edu/.../supplychain_finance_20141… · · 2014-10-30Supply Chain Network Structure and Risk Propagation
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Supply Chain Network Structure and Risk Propagation
John R. Birge 1
1University of Chicago Booth School of Business (joint work with Jing Wu, Chicago Booth)
Supply chain relationships have direct and indirect effects on firmperformance
Correlation and reliability issues in particular create nonlinear effects on thevalue of supplier connections
Effects of centrality on risk propagation can vary depending on position inthe supply chain
New databases (e.g., Bloomberg SPLC) provide opportunities to investigatefinancial and operational supply chain network interactions
Empirical results on firm returns show significant supplier and customereffects, lagged effects from supplier shocks, and differences in thesecond-order (systematic) risk impact of centrality depending on the firm’schain level
Customer to Supplier: Calloway Golf/Coastcast Corporation (Cohen andFrazzini (2008))
Calloway reports earnings to be ∼half of expectations ($0.36 from $0.70)Calloway’s stock price drops 30%Coastcast share price (50% of sales to Calloway) virtually unchanged for overone monthAfter Coastcast announces lower earnings, price drop mirrors Calloway’s
Supplier to Customer: Philips/Sony/Ericsson v. Nokia
Fire in Philips plant, key supplier for Nokia and Ericsson, in March 2000Initially, Philips states 1-week shutdown, then revises 2 weeks later (to 6weeks)Philips stock price drops then recoversNokia reacts quickly - earnings and share price riseEricsson reacts slowly - earnings and share price drop (with some delay)
Observations:Customer shocks are transmitted to suppliersSupplier shocks are also transmitted to customers (e.g., Hendricks and Singhal(2003))Historically, transmission has often been delayedAdaptive supply chains (e.g., alternate suppliers) can dampen shocks (andmay reduce risk exposures)
Share price effects:Model of price at time t:
pt =∑s
e−(rs+δ)sds
Expected dividends, ds , depend partly on supply chain partners (first-ordereffect)Risk premium, δ, depends on position in network and relationship ofconnections to risk transmission (second-order effect)
Impact of connection visibility:Changes in expectations may be delayed due to inattention or lack oftransparency into supply chain connections.
1 Previous work: Jackson and Wolinsky (1996): Firm i maximizes utility ui bycreating connections ij in graph G (distance dij) where
ui = wii +∑j 6=i
δdijwij −∑(ij)∈G
cij .
2 Results: For symmetric networks, either G is complete, a star graph, orlinkless.
3 Innovation: allow nonlinear effects wi,jkxijxik where xij = 1 for (ij) ∈ G
wi,jk may be positive or negative depending on correlation of j and k interestsand their reliabilityHypothesis: negative correlation yields negative wi,jk .
1 Previous work: Jackson and Wolinsky (1996): Firm i maximizes utility ui bycreating connections ij in graph G (distance dij) where
ui = wii +∑j 6=i
δdijwij −∑(ij)∈G
cij .
2 Results: For symmetric networks, either G is complete, a star graph, orlinkless.
3 Innovation: allow nonlinear effects wi,jkxijxik where xij = 1 for (ij) ∈ G
wi,jk may be positive or negative depending on correlation of j and k interestsand their reliabilityHypothesis: negative correlation yields negative wi,jk .
1 Previous work: Jackson and Wolinsky (1996): Firm i maximizes utility ui bycreating connections ij in graph G (distance dij) where
ui = wii +∑j 6=i
δdijwij −∑(ij)∈G
cij .
2 Results: For symmetric networks, either G is complete, a star graph, orlinkless.
3 Innovation: allow nonlinear effects wi,jkxijxik where xij = 1 for (ij) ∈ G
wi,jk may be positive or negative depending on correlation of j and k interestsand their reliabilityHypothesis: negative correlation yields negative wi,jk .
1 Previous work: Jackson and Wolinsky (1996): Firm i maximizes utility ui bycreating connections ij in graph G (distance dij) where
ui = wii +∑j 6=i
δdijwij −∑(ij)∈G
cij .
2 Results: For symmetric networks, either G is complete, a star graph, orlinkless.
3 Innovation: allow nonlinear effects wi,jkxijxik where xij = 1 for (ij) ∈ G
wi,jk may be positive or negative depending on correlation of j and k interestsand their reliabilityHypothesis: negative correlation yields negative wi,jk .
Even symmetric networks with this cost structure may have widely varyingequilibrium structureWith a single parameter α for the correlation, a full range of degreedistributions existExamples:
Even symmetric networks with this cost structure may have widely varyingequilibrium structureWith a single parameter α for the correlation, a full range of degreedistributions existExamples:
Even symmetric networks with this cost structure may have widely varyingequilibrium structureWith a single parameter α for the correlation, a full range of degreedistributions existExamples:
Even symmetric networks with this cost structure may have widely varyingequilibrium structureWith a single parameter α for the correlation, a full range of degreedistributions existExamples:
Even symmetric networks with this cost structure may have widely varyingequilibrium structureWith a single parameter α for the correlation, a full range of degreedistributions existExamples:
Manufacturer M1 can survive if one of the wholesaler nodes is wiped out;Wholesaler W2 cannot survive either manufacturer’s loss (due to lowermargin/market power).Implications:
Performance metric: stock return (proxy for direct shock effects andsystematic risk)
First-order effects (changes in future expectations of a firm)
Suppliers’ and customers’ concurrent performance relates to the firmSupplier momentum (one-month lag) may be related to firm performanceCustomer momentum (following Cohen and Frazzini (2008)) not related tofirm performance due to greater awareness
Second-order (systematic risk) effects (relationship to network connections)
Centrality influences firm risk and return performanceMore central manufacturing firms have lower returns (lower risk)More central logistics (transportation, wholesale, retail) firms have higherreturns (higher risk)
Characterize centrality by eigenvector centrality and in- and out-degreecentralityUse average of industry if no relationship in datasetSplit by NAICS code (3 for manufacturing, 4 for logistics)
Split into quintiles of centrality.
Observe trends and significance in returns across quintiles.