Managing risk in the supply chain

Transcription

1 Managing risk in he supply chain David Simchi-Levi Engineering Sysems Division Massachuses Insiue of Technology Gérard P. Cachon The Wharon School Universiy of Pennsylvania Supply Chain Roundable July 2004 Agenda Sources of and ools for supply chain managemen risk Recen research on supply chain risk managemen Risk neural decision makers Risk averse decision makers More discussion 1

2 Sources of supply chain managemen risk Demand risk Supply risk Price risk Sraegic risk Environmenal risk Unknown risks Qualiy risk Tools for managing risk in he supply chain Locaion pooling Lead ime pooling Produc pooling Delayed differeniaion Capaciy pooling / flexible manufacuring Dynamic pricing / capaciy conrols Assemble-o-order Ousourcing / offsourcing CPFR, VMI Conracs (buy-backs, quaniy flexibiliy, ec.) Markes/exchanges/aucions Financial engineering: Deviaion measures, real opions, porfolio opimizaion, ec. 2

3 Lieraure Managing risk in a risk neural world Qualiy risk: Baiman, Neessine and Kunreuher (2004), Baiman and Neessine (2004) Delivery lead ime risk: Cachon and Zhang (2004a,b) Supplier qualiy/performance: Debo (2004) Forecas qualiy: Lariviere (2004) Forecas sharing: Anand and Goyal (2004) Spo price volailiy: Wu, Kleindorfer and Zhang (2002); Wu and Kleindorfer (2004) Allocaion of invenory risk Supplier Supplier Push Sell o he newsvendor Reailer Reailer Pull Buy from a newsvendor Everyone in he supply chain can be beer off by swiching from one exreme risk allocaion o he oher (i.e., from push o pull or from pull o push) Smar allocaion of risk can reduce he need for complex conracs. 3

4 Is supply chain risk reducion always Pareo improving? Iyer and Bergen (1997): Quick response does no always benefi he supplier. Anupindi and Bassok (1999): Locaion pooling a he reail level does no always benefi he supplier. Lee and Whang (2002): A secondary marke does no always benefi he supplier. Dong and Rudi (2004): Invenory ransshipmen among reailers does no always benefi he supplier. Is supply chain risk reducion beneficial in a compeiive seing? Roller and Tombak (1993): Manufacuring flexibiliy can be harmful. Carr, Duenyas, Lovejoy (1999): Less demand or supply risk can be harmful. Anand and Girora (2004): Delayed differeniaion can be harmful. Bu Cachon and Harker (2002): Capaciy pooling wih a conrac manufacurer benefis compeing firms because price compeiion is reduced. 4

5 Approaches o Risk Managemen Economics lieraure Von Neumann-Morgensern uiliies Expeced uiliy E{U(Π)} Finance lieraure Capial Asse Pricing Model (CAPM) Markowiz Mean-Variance radeoff E{Π}- kvar{π} Porfolio Approach Linking Economics wih Finance When The uiliy funcion is quadraic OR The uiliy is CARA* [U(Π)=-exp(-r Π)] and Π is normally disribued Then Expeced uiliy maximizaion is equivalen o mean-variance objecive maximizaion *CARA: Consan Absolue Risk Averse 5

6 Supply Chain Managemen Academia: Tradiional models focus on maximizing expeced profi Pracice: Significan increase in he level of risk faced by many companies Examples: Cisco, Apple, Sony Lieraure Managing risk in a risk averse world Single Period Lau (80): Tradeoff beween profi mean and sandard deviaion Eeckhoud, Gollier and Schlesinger (95): Exponenial uiliy funcion Chen and Federgruen (00): Mean-variance radeoff Schweizer and Cachon (00): Empirical work Muli Period Bourakiz and Sobel (92): Exponenial uiliy funcion, base-sock is opimal Infinie Horizon Bourakiz and Sobel (92): Exponenial uiliy funcion, base-sock is opimal Chen and Federgruen (00): Mean-variance radeoff for invenory level or cusomer waiing ime 6

7 Risk Measures Mean-Variance Limiaion Equally penalizes desirable upside and undesirable downside oucomes Oher Risk Measures Uiliy funcions VaR CVaR Risk Measure: uiliy funcion 7

8 Risk Measure: Value a Risk η q η Pr ob( f ( µ, d ~ ) < qη ) = η Problems wih Value a Risk Does no preserve subaddiiviy A porfolio wih wo insrumens may have a larger VaR han he sum of he VaRs of he wo insrumens The VaR risk measure is indifferen o he exen of which he profi falls below he qη 8

9 Risk Measure: Condiional Value a Risk η Muli-period Invenory Model Single produc, periodic review Finie horizon: T Fixed ordering cos: K Variable ordering cos: c Zero lead ime Convex invenory and backorder cos: Demand funcion: d = α p + β Objecive: Maximize expeced uiliy or condiional value a risk of he oal discouned profi (x) h 9

10 Muli-period model: sequence of evens y Kδ ( y x) + c( y x) h( y d% ) x d % y d % +1 p p d % Finie Horizon Model: Main Resuls Chen, Sim, Simchi-Levi and Sun (2004) Fixed Price K=0 K>0 Price Conrol K=0 K>0 Risk Neural Model Base sock (s,s) Base sock lis price (s,s,a,p) Exac Exponenial Uiliy Increasing & Concave Uiliy or CVaR Base sock Wealh dependen Base sock (s,s)? Base sock Wealh dependen Base sock (s,s,a,p)? Heuri sics CVaR Bass sock (s,s) Base sock (s,s,a,p) 10

11 Quesions Is here a disconnec beween academic research and indusry needs? Wha is an appropriae risk measure? Wha are appropriae risk models? Models ha combine operaional and financial hedging sraegies Is here anyhing we can learn from oher indusries? Teaching cases? Mehods o deal wih unknown unknown 11