Online Supplement: Advance Selling in a Supply Chain under Uncertain Supply and Demand

Similar documents
Minimum Spanning Trees

3-2-1 ANN Architecture

dy 1. If fx ( ) is continuous at x = 3, then 13. If y x ) for x 0, then f (g(x)) = g (f (x)) when x = a. ½ b. ½ c. 1 b. 4x a. 3 b. 3 c.

First order differential equation Linear equation; Method of integrating factors

Exponential Functions

10. EXTENDING TRACTABILITY

MAT 270 Test 3 Review (Spring 2012) Test on April 11 in PSA 21 Section 3.7 Implicit Derivative

First derivative analysis

Multiple Short Term Infusion Homework # 5 PHA 5127

1973 AP Calculus AB: Section I

AP Calculus BC AP Exam Problems Chapters 1 3

Finite Element Analysis

A New Binary Sequence Family with Low Correlation and Large Size

Exercise 1. Sketch the graph of the following function. (x 2

The second condition says that a node α of the tree has exactly n children if the arity of its label is n.

VLSI Testing Assignment 2

SECTION where P (cos θ, sin θ) and Q(cos θ, sin θ) are polynomials in cos θ and sin θ, provided Q is never equal to zero.

von Neumann-Wigner theorem: level s repulsion and degenerate eigenvalues.

Additional Math (4047) Paper 2 (100 marks) y x. 2 d. d d

Mathematics 1110H Calculus I: Limits, derivatives, and Integrals Trent University, Summer 2018 Solutions to the Actual Final Examination

3) Use the average steady-state equation to determine the dose. Note that only 100 mg tablets of aminophylline are available here.

Thomas Whitham Sixth Form

Case Study 1 PHA 5127 Fall 2006 Revised 9/19/06

CPSC 665 : An Algorithmist s Toolkit Lecture 4 : 21 Jan Linear Programming

Thomas Whitham Sixth Form

Linear-Phase FIR Transfer Functions. Functions. Functions. Functions. Functions. Functions. Let

Case Study 4 PHA 5127 Aminoglycosides Answers provided by Jeffrey Stark Graduate Student

AP Calculus Multiple-Choice Question Collection

6.1 Integration by Parts and Present Value. Copyright Cengage Learning. All rights reserved.

The Transfer Function. The Transfer Function. The Transfer Function. The Transfer Function. The Transfer Function. The Transfer Function

a 1and x is any real number.

Y 0. Standing Wave Interference between the incident & reflected waves Standing wave. A string with one end fixed on a wall

Supplementary Materials

Case Study Vancomycin Answers Provided by Jeffrey Stark, Graduate Student

Introduction to Arithmetic Geometry Fall 2013 Lecture #20 11/14/2013

The Matrix Exponential

PROBLEM SET Problem 1.

Edge-Triggered D Flip-flop. Formal Analysis. Fundamental-Mode Sequential Circuits. D latch: How do flip-flops work?

SOURCES OF ERROR AND LIMITS OF APPLYING THE ENERGETIC-BASED-REGULARIZATION METHOD

The Matrix Exponential

Grade 12 (MCV4UE) AP Calculus Page 1 of 5 Derivative of a Function & Differentiability

Network Congestion Games

Rikkert Frederix. Figure 2: As in fig. 1, for the transverse momentum of the 1 st jet.

LINEAR DELAY DIFFERENTIAL EQUATION WITH A POSITIVE AND A NEGATIVE TERM

Function Spaces. a x 3. (Letting x = 1 =)) a(0) + b + c (1) = 0. Row reducing the matrix. b 1. e 4 3. e 9. >: (x = 1 =)) a(0) + b + c (1) = 0

Chapter Finding Small Vertex Covers. Extending the Limits of Tractability. Coping With NP-Completeness. Vertex Cover

Schematic of a mixed flow reactor (both advection and dispersion must be accounted for)

Using Stochastic Approximation Methods to Compute Optimal Base-Stock Levels in Inventory Control Problems

A Propagating Wave Packet Group Velocity Dispersion

ENGR 7181 LECTURE NOTES WEEK 3 Dr. Amir G. Aghdam Concordia University

Atomic Physics. Final Mon. May 12, 12:25-2:25, Ingraham B10 Get prepared for the Final!

University of Washington Department of Chemistry Chemistry 453 Winter Quarter 2014 Lecture 20: Transition State Theory. ERD: 25.14

22/ Breakdown of the Born-Oppenheimer approximation. Selection rules for rotational-vibrational transitions. P, R branches.

Math 34A. Final Review

Calculus II (MAC )

Chapter 10. The singular integral Introducing S(n) and J(n)

Introduction to Condensed Matter Physics

is an appropriate single phase forced convection heat transfer coefficient (e.g. Weisman), and h

General Notes About 2007 AP Physics Scoring Guidelines

Abstract Interpretation: concrete and abstract semantics

Optimizing Product Launches in the Presence of Strategic Consumers Appendix

cycle that does not cross any edges (including its own), then it has at least

CS 491 G Combinatorial Optimization

BINOMIAL COEFFICIENTS INVOLVING INFINITE POWERS OF PRIMES. 1. Statement of results

COUNTING TAMELY RAMIFIED EXTENSIONS OF LOCAL FIELDS UP TO ISOMORPHISM

Revisiting Wiener s Attack New Weak Keys in RSA

Indeterminate Forms and L Hôpital s Rule. Indeterminate Forms

A Prey-Predator Model with an Alternative Food for the Predator, Harvesting of Both the Species and with A Gestation Period for Interaction

A. Limits and Horizontal Asymptotes ( ) f x f x. f x. x "±# ( ).

NEW APPLICATIONS OF THE ABEL-LIOUVILLE FORMULA

Partial Derivatives: Suppose that z = f(x, y) is a function of two variables.

2008 AP Calculus BC Multiple Choice Exam

MAT3707. Tutorial letter 201/1/2017 DISCRETE MATHEMATICS: COMBINATORICS. Semester 1. Department of Mathematical Sciences MAT3707/201/1/2017

2. Finite Impulse Response Filters (FIR)

Outlines: Graphs Part-4. Applications of Depth-First Search. Directed Acyclic Graph (DAG) Generic scheduling problem.

Chapter Taylor Theorem Revisited

On the irreducibility of some polynomials in two variables

λ = 2L n Electronic structure of metals = 3 = 2a Free electron model Many metals have an unpaired s-electron that is largely free

Thinking outside the (Edgeworth) Box

y cos x = cos xdx = sin x + c y = tan x + c sec x But, y = 1 when x = 0 giving c = 1. y = tan x + sec x (A1) (C4) OR y cos x = sin x + 1 [8]

Deift/Zhou Steepest descent, Part I

Physics 43 HW #9 Chapter 40 Key

10. The Discrete-Time Fourier Transform (DTFT)

1997 AP Calculus AB: Section I, Part A

CPS 616 W2017 MIDTERM SOLUTIONS 1

A RELATIVISTIC LAGRANGIAN FOR MULTIPLE CHARGED POINT-MASSES

A C 0 INTERIOR PENALTY METHOD FOR A FOURTH ORDER ELLIPTIC SINGULAR PERTURBATION PROBLEM

a g f 8 e 11 Also: Minimum degree, maximum degree, vertex of degree d 1 adjacent to vertex of degree d 2,...

UNIFIED ERROR ANALYSIS

Derangements and Applications

Walk Like a Mathematician Learning Task:

Appendix 2.3 General Solutions for the Step Response of Third- and Fourth-Order Systems (with some unpleasant surprises!)

BINOMIAL COEFFICIENTS INVOLVING INFINITE POWERS OF PRIMES

Section 11.6: Directional Derivatives and the Gradient Vector

EXST Regression Techniques Page 1

Examples and applications on SSSP and MST

(Upside-Down o Direct Rotation) β - Numbers

Einstein Equations for Tetrad Fields

Greenfield Wind Farm. Visual Simulation 1. Affinity Renewables Inc. Figure As viewed from Trans Canada Highway 104.

Problem Set 6 Solutions

Transcription:

Onlin Supplmnt Avanc Slling in a Supply Cain unr Uncrtain Supply an Dman. Proos o Analytical sults Proo o Lmma. Using a = minl 0 ; x g; w can rwrit () as ollows (x ; w ; x ; w ) = a +(m0 w )a +( +" x w )x Tus, i m0 w < l 0 ; tn x = m0 w is t uniqu optimal orr quantity tat maximizs ; otrwis, any quantity x ( l 0 ) incluing m0 w is optimal. Proo o Lmma. W n to consir only t cas wn w m 0 bcaus otrwis x AB = 0 rom Lmma. Lt us rst consir t cass (i) an (ii) in wic m 0 > 0. Obsrv rom () tat M is linar in w wn w m 0 l 0, an quaratic in w wit its maximum at w = m0 wn m 0 l 0 < w m 0 It is asy to sow tat M is unimoal an tat its maximum is attain at w AB = m 0 l 0 i m 0 l 0 > m0 (i.., l0 < m0 ), an at wab = m0 otrwis. In t cas (iii) wn m 0 0, or any w ; x AB = 0 rom Lmma. By substituting w AB in ac cas into x AB in Lmma, w obtain x AB By substituting (w AB ; x AB ) into () an (), w obtain t xprssions or M an ; rspctivly. Proo o Lmma 3. (a) W can rwrit (5) as ollows E ~"; ~ [ (x ; w )] = x + ( w )x g F ( x )g + x 0 () + ( w )()gf () Using t Libniz intgral rul, w obtain @E ~"; ~ [ (x ;w )] @x = x + ( w )g F ( x )g Sinc F ( x ) 0; E ~"; ~ [ (x ; w )] is unimoal an maximiz at x A (w ) = w (b) Sinc M (w ) in (6) is concav in w or any ~ ; E ~ [ M (w )] is also concav in w, nc a n uniqu w A xists. As illustrat in t blow gur, i ~ 05; arg min ~ oi w ; w w = 05; n w otrwis, arg min ~ oi w ; w w = ~ > 05 Tus, i ~ 05 8 ~ ; w A = 05; otrwis, w w A > 05 By substituting wa an xa (wa ) into (5) an (6), w obtain t rsults or EA an E A M mar. W can sow tat " 075 is su cint to guarant positiv pric in quilibrium unr avanc slling as ollows. (Tis conition is also su cint or positiv pric unr ynamic slling bcaus x AB x A ) In quilibrium, xa (wa ) = wa, so p A = + ~" min ~ ; x A (wa )g = + ~" min ~ ; wa g Dpning on t valu o wa ; tr ar tr cass (i) I wa <, tn min ~ ; wa g = wa 8 ~ ; an minp A j~" ["; "]; ~ [; ]g = " + +wa wa, tn min min ~ ; wa gj ~ [; ]g = wa, an minp A j~" ["; "]; ~ (ii) I [; ]g = " + +wa (iii) I wa >, tn min ~ ; wa g = ~ 8 ~, an minp A j~" ["; "]; ~ [; ]g = + " In (i) an (ii), minp A j~" ["; "]; ~ [; ]g > 0 i an only i " +wa +w, wr 075 bcaus w A 05. In (iii), minpa j~" ["; "]; ~ [; ]g > 0 i an only i ", wr < wa = +wa 075 Tror, wn " 075, p A > 0 or all ~" an ~

Figur Manuacturr s Ex-Post Pro t M unr Avanc Slling Proo o Torm. (a) W can rwrit E A M givn in (6) as E A M = max w w min ~ ; w F ( ~ ) = wa min ~ ; wa F ( ~ ) (8) W can also rwrit t manuacturr s x-ant pro t E B M unr rgular slling as E B M = " " B M( ~ ; ~")G(~")F ( ~ ) max w min ~ ; w F ( w ~ ); (9) wr t inquality is u to Jnsn s inquality an t proprty tat B M prsnt in Corollary is incrasing an convx in ~". Finally, obsrv tat t intgran o (9) is gratr tan or qual to tat o (8) or ~, nc E B M EA M EB M = EA M wn on o t ollowing two i conitions ar satis (i) Pr ~ = = Pr +~" ~ i = bcaus B M (~ ; ~") is linar in ~", an w tat maximizs t intgran o (9) is qual to w A = (s t gur abov), or (ii) Pr [~" = 0] = Pr +~" ~ i = bcaus E B M = EA M = 8 rom Lmma 3(i) an Corollary (i). (b) Suppos Pr +~" ~ i = Tis implis tat 05( +") 05, so E A = 6 by Lmma i 3(i). From Corollary (i), E B = E (+~") 6 in tis cas. By Jnsn s inquality, E B EA ; wr t quality ols wn Pr[~" = 0] = Wn Pr +~" ~ i <, Tabl sows tat E B E A can b itr positiv or ngativ. Proo o Torm. (a) T rsult ollows irctly rom Torm (a) an Lmma. (b) Suppos Pr +~" ~ i = Tn, rom Torm (b), E B EA. Sinc EAB E B 8; ~ ; ~", E AB E A, wr t quality ols wn EAB = E B = EA. Wn Pr +~" ~ i < ; Tabl 3 sows tat E AB E A can b itr positiv or ngativ. n Proo o Proposition. From t proo o Lmma 3, M (w ) = min ~ o w ; w w in (6) is concav in ~ or any givn w. By t wll-nown proprty o t scon-orr stocastic ominanc, E ~ u( ~ i ) E ~ u( ~ i ) or any concav unction u. Tus, E ~ [ M (w )] E ~ [ M (w )] or any givn w D n w 0 arg maxe ~ [ M (w )] an w 00 arg maxe ~ [ M (w )] Tn, E ~ [ A M ] = w w E ~ [ M (w 0 )] E ~ [ M (w 00)] E ~ [ M (w 00)] = E ~ [ A M ] T rst inquality ollows rom t

optimality o w 0 an t scon inquality ollows rom t scon-orr stocastic ominanc o ~ ovr ~ Proo o Proposition. (a) From Corollary, B M is incrasing an concav in ~ or ~ < +~" an it is constant or ~ +~". Tus, B M is non-crasing an concav in ~. By ollowing t sam argumnt as in t proo o Proposition, w can sow t sir rsult. Similarly, w can obtain t rsults or w B an xb by sowing tat wb is non-incrasing an convx in ~ an tat x B is non-crasing an concav in ~ (b) From Corollary, B M is incrasing an convx in ~" or ~" ~ an it is linarly incrasing or ~" > ~. Tus, B M is incrasing an an convx in ~". By ollowing t sam argumnt as in t proo o Proposition, w can sow t sir rsult. Similarly, w can sow tat w B is incrasing an convx in ~" an tat x B is non-crasing an concav in ~"; an obtain t sir rsults.. Computation unr Spci c Probability Distributions Tis sction prsnts t clos-orm xprssions o t quilibrium outcoms unr avanc or rgular slling, an xplains t procur to comput t quilibrium outcoms unr ynamic slling wn t spci c istributions o ~ an ~" ar us as scrib in 6. Unr t avanc slling stratgy, a A (w ) pns on t ranom yil ~ in t ollowing mannr i w ~ + r; a A (w ) = w an i r ~ w, aa (w ) = ~ Tus, w can xprss E ~ [ M (w )] in (6) as E ~ [ M (w )] = E ~ [w a A (w )] = +r w a w r + a r w r ; wr a min + r; max r; w. By noting tat a can ta on tr i rnt valus pning 8 on w, E ~ [ M (w )] can b rwrittn as w i 0 w ( + r) >< E ~ [ M (w )] = w 6r [w ( + r + p r)g] [w ( + r p r)g] i ( + r) < w < ( r) > w ( w ) i ( r) w Lmma 3 as sown tat i ( r) 05, w A = 05 I ( r) < 05, w can sow w A = 3 ( r) + 3p ( + r + r ) ( + r) + ( w 0 ) (0) T proo o t abov rsult is as ollows. Lt g (w ) w, g (w ) w 6r [w ( + r + p r)g] [w ( + r p r)g], an g 3 (w ) w ( w ) Not tat g is a cubic unction o w an its omain [ ( + r); ( r)] is contain in [ ( + r + p r); ( + r p r)]. Sinc 3 g w 3 < 0 an (+r p r) > 0; g is unimoal wit its maximum at a largr root o g (w ) w = 0; wic is qual to w 0 givn in (0). (Not tat g (w ) w = 0 is a quaratic quation o w.) Also, g 3 is concav in w wit its maximum at 05. Sinc g w w = (+r) = g w w = (+r) = > 0, 3

w A os not xist in t rst intrval o w. Suppos ( r) 05 Tn, g w w = ( r) = + ( r) 0; nc wa = 05 [ ( r); ]. Nxt, suppos ( r) < 05. Tn, g w w > 0; g w = (+r) w < 0 an g 3 w = ( r) w < 0; nc = ( r) wa = w0 is optimal. In tis cas, w A > 05 bcaus g w w = =05 6r ( r)g i > 0 By substituting w = w A into xa (w ), w can obtain t clos-orm xprssions or x A (wa ) an similarly or EA M an E A. For t rgular slling stratgy, using t x-post quilibrium outcoms givn in Corollary, w comput t x-ant quilibrium outcoms. Using Corollary, w can writ t x-ant xpct wolsal pric as E ~;~" w B = Pr(~" = ) Pr( ~ b ) + + i b r [ + ] r +Pr(~" = ) = r + 3r + ( + ) b + ( ) c g b + c ( r) ; Pr( ~ c ) + c wr b min + r; max r; + an c min + r; max r; Similarly, w can xprss otr x-ant quilibrium outcoms suc as E B M, ExB, an EB. For ac o six possibl cass o aving i rnt pairs o b or c ; w av riv t clos-orm xprssion o t quilibrium outcoms, wic ar availabl upon rqust. For t ynamic slling stratgy, w av prsnt in Lmma t clos-orm xprssions o t quilibrium outcoms in t scon-prio gam. As w sow blow, t trmination o t rtailr s quilibrium pr-boo quantity in t rst prio involvs t analysis o 39 cass. Tus, insta o ning t clos-orm xprssions or ac cas, w vis an cint procur to comput t quilibrium outcoms in t rst prio. First, consir t rtailr s problm in t rst prio. T conitions provi in Lmma can b r-writtn as ollows ~ l 0 ~m0 as ~ +~"+x trsol numbrs min + r; max r; ++x r, ~ l 0 > 0 as ~ > x ; an ~m 0 > 0 as x < +~". D n ; min + r; max r; +x, an min + r; max r; x ; an t inicator unction I(y) = i y is tru an I(y) = 0, otrwis. Tn E ~;~" [ (x ; w )] in (3) can b rwrittn as E ~;~" [ (x ; w )] = +r ( x w ) x r + r ( w ) r + Pr(~" = )I x < + + Pr(~" = )I x < 6 6 ( + x ) r + i ( x ) r +r ( x ) r + i ( x ) r, an I x < ; E~;~" [ (x ; w )] tas on +r Dpning on t valus o ;,, I x < + i rnt unctional orms o x. Not tat ; or is itr a constant or a linar unction o x ; an I x < + or I x < i rnt combination o ;,, I x < + is itr 0 or. Hnc, in ac intrval o x aving a, an I x <, E~;~" [ (x ; w )] is at most a cubic unction o x. T pr-boo orr quantity x AB (w ) tat maximizs E ~;~" [ (x ; w )] or any givn w is itr a bounary point btwn any two intrvals or an intrior point at wic t rst orr conition is satis. Tr ar potntially 39 caniats or x AB (w ) wic consist o 9 bounary points btwn two intrvals o x an 30 intrior optimal points witin any intrval o i [ ] r

x. T bounary points ar 0; +, ; (+r) (+) (at wic + r = ++x (at wic r = ++x ), (+r) ( ) (at wic + r = +x ), ), ( r) (+) ( r) ( ) (at wic r = +x ), ( + r) (at wic + r = x ), an ( r) (at wic r = x ). To n t intrior optimal points, w rst simpliy t xprssion o E ~;~" [ (x ; w )] or ac o t ollowing 30 cass = I x < = 0; so E~;~" [ (x ; w )] can av 3 i rnt xprs- (i) I x +, I x < + sions wn = + r; r or x ; (ii) I x < +, I x < + = an I x < = 0; so E~;~" [ (x ; w )] can av 9 i rnt xprssions wn = + r; r or x, an = + r; r or ++x ; (iii) I x <, I x < + = I x < = ; so E~;~" [ (x ; w )] can av 8 i rnt xprssions wn = + r; r or x, = + r; r or ++x, an = + r; r or +x (not 8 cass xist insta o 7 cass bcaus ). For ac o t abov 30 cass, w can asily obtain an intrior optimal point rom t rst orr conition (wic w omit r). By comparing E ~;~" [ (x ; w )] at ts 39 caniats, w can n t rtailr s bst rspons x AB (w ) or any givn w Nxt, consir t manuacturr s problm in t rst prio. Similar to E ~;~" [ (x ; w )], w can rwrit E ~;~" [ M (x ; w )] in () as E ~;~" [ M (x ; w )] = +r (w x ) r + r (w ) r + Pr(~" = )I x < + +r 8 ( + x ) r + i ( + ) ( x ) r + Pr(~" = )I x < +r 8 ( x ) r + ( ) ( x ) r W can cintly comput t pr-boo wolsal pric w AB tat maximizs E ~;~" M (x AB (w ); w ) as ollows. W rst comput t rtailr s pr-boo quantity x AB (w ) as a unction o w an intiy bounary points btwn any two ajacnt intrvals o w at wic x AB (w ) switcs rom on o t 39 caniat points to anotr. In ac intrval o w, E ~;~" M (x AB (w ); w ) is a continuous unction, nc its local maximum is attain at itr a bounary point or an intrior point at wic t rst orr conition is satis. By comparing local maxima, w can intiy a global optimal point w AB. 3. Invntory Holbac Consir t Staclbrg gam tat tas plac bor t supply an man uncrtaintis ar rsolv. In tis gam, t manuacturr rst sts is pr-boo wolsal pric w an tn t rtailr trmins r pr-boo orr quantity x. In aition, atr obsrving t raliz man an supply, t rtailr as an option o witoling t part o t pr-boo orr quantity tat s as rciv. W us suprscript AH to not quilibrium outcoms in tis gam. For gnral probability istributions o ~ an ~", u to t complxity o t mol, t problm is intractabl. Tus, w assum t sam istributions as in t prvious sction, an compar i 5

quilibrium outcoms in tis gam wit tos unr avanc or rgular slling. Suppos tat ~ is uniormly istribut btwn r an + r, wr r (0; ], an tat ~" = wit probability 0.5 an ~" = wit probability 0.5, wr (0; ]. Lt Q not t quantity tat t rtailr slls to t mart. Atr obsrving ~", t rtailr trmins Q ( x ) to maximiz r x-post pro t ( + ~" Q)Q w x Tus, t optimal Q AH quals +~" Clarly, t rtailr woul not pr-boo mor tan +, wic is QAH wn ~" = ; i.., x +. T x-ant xpct pro t ( o t rtailr in quation () o t bas mol is tn moi into w x + 05( + x )x + 05( x )x i x E ~" [ (x ; w )] = w x + 05( + x )x + 05( ) i x >. Tis can b intrprt as ollows. I t rtailr as rciv x rom r pr-boo orr, it is optimal or r to sll t ntir quantity tat s as rciv rom r pr-boo orr; i.., Q AH = x Otrwis, wn t man turns out to b low (i.., ~" = ), it is optimal or t rtailr to witol x (> 0) an sll only Q AH = ; an wn t man is ig (i.., ~" = ), it is optimal to sll Q AH = x ( ( ; + ]) From t abov quation or E ~" [ (x ; w )], givn t pr-boo ( wolsal pric w ; w obtain t ollowing optimal pr-boo quantity w x AH i x or quivalntly w (w ) = + w i x > or quivalntly w < Not tat + w > w i an only i w <. Compar wit x A (w ) = w or all w in our bas mol, x AH (w ) implis tat, givn t pr-boo wolsal pric w, t rtailr pr-boos mor wit t olbac option tan in t bas mol. In anticipation o t rtailr s bst rspons x AH (w ); t manuacturr trmins is prboo wolsal pric w to maximiz 8 is nxpct pro t < w min ~ o ; w E M (w ) = w a AH i w (w ) = n w min ~ o ; + w i w <. In t ollowing, w will rst analyz (Intrval ) w an (Intrval ) w < sparatly, an tn combin t rsults. W will sow tat E M (w ) is itr unimoal or bimoal in w. (Intrval ) In tis intrval, t rtailr os not witol t invntory. Tus, w can us t rsults o t bas mol wit no olbac option prsnt in t prvious sction. Sinc E M (w ) is unimoal, it is optimal or t manuacturr to st w = max05; g i ( w = maxw 0 ; g otrwis, wr w 0 is givn in (0). r) 05; an (Intrval ) In tis intrval, t rtailr witols t invntory atr obsrving raliz ~ an ~". Dpning on t ranom yil ~, a AH (w ) can b xprss as ollows i + w ~ + r; a AH (w ) = + in (6) as w an i r ~ w, aah (w ) = ~ Tus, w can xprss E ~ [ M (w )] E ~ [ M (w )] = E ~ [w a AH (w )] = +r + w w w r + w r w r ; wr w min + r; max r; + w. By noting tat w can ta on tr i rnt valus pning on w, E ~ [ M (w )] can b rwrittn as 6

8 w i 0 w + ( + r) >< w + r w ( + r + p r) E ~ [ M (w )] = w + ( + r p + i ( + r) < w < + ( r) r) > + (w ) + ( + ) i + ( r) w By ollowing t sam mto as prsnt in Sction o tis onlin supplmnt, w can sow tat E M (w ) is unimoal in tis intrval an tat t optimal wolsal pric in tis intrval is i ( r) + ; w = min + ; g; otrwis, w = minw 00 ; g, wr w 00 = + 3 ( + r) + 6p ( + r + r ) ( + )( + r) + ( + ) (Intrvals & ) Bcaus E M (w ) is unimoal in ac o Intrval an Intrval, E M (w ) is unimoal or bimoal in w across t two intrvals. Tror, a global optimum w AH can b oun by comparing t maximum valu o E M (w ) in ac o t two intrvals. By substituting w AH into x AH (w ), w can obtain x AH (w AH ) an t rsulting xpct pro ts o bot rms. Using t mto scrib abov, w av comput t quilibrium outcoms o 80 scnarios by varying rom 0 to an varying an r rom 0 to 09 wit an incrmnt o 0 W tn compar t quilibrium outcoms wit tos unr avanc or rgular slling. To illustrat, w prsnt t rsults o 7 scnarios in Tabls A an A. From our numrical xprimnts, w raw t ollowing obsrvations (a) In all 80 scnarios, E AH M EA M an xah x A. Howvr, w av obsrv bot wah w A > 0 an wah w A < 0; an bot EAH > EA an EAH < EA. (b) In all 80 scnarios, E B M EAH M. Howvr, w av obsrv bot wb w AH > 0 an w B w AH < 0; bot x B > xah an x B < xah ; an bot E B > EAH an EB < EAH. Obsrvation (a) sows tat t rtailr pr-boos a largr quantity wit t olbac option. Howvr, tis os not ncssarily bn t t rtailr bcaus t manuacturr can anticipat t rtailr s bst rspons an trmin is wolsal pric accoringly. As a rsult, it is only t manuacturr wo will always bn t rom tis aitional option. Obsrvation (b) con rms tat our rsults obtain in t bas mol continu to ol in t olbac mol t manuacturr is always bttr o unr rgular slling tan unr avanc slling, wras t rtailr may prr itr stratgy. T comparison btwn avanc slling wit t olbac option an ynamic slling is similar to t comparison btwn avanc slling witout tis option an ynamic slling. Tabl A3 summarizs t comparativ statics o tis gam. It sows tat t cts o capacity an supply uncrtainty r on t quilibrium outcoms ar t sam as tos unr avanc slling witout t olbac option. Wil man uncrtainty os not a ct t quilibrium outcoms witout t olbac option, it os a ct t quilibrium outcoms o tis gam. T cts o on t xpct pro ts o bot rms ar t sam as tos unr rgular slling man uncrtainty bn ts t manuacturr but os not always bn t t rtailr. In aition, w av obsrv tat t rtailr s pr-booing quantity is incrasing in man uncrtainty, but t 7

manuacturr s wolsal pric is not monotonic in. Tabl A. Equilibrium Outcoms unr Avanc Slling E cts o t Holbac Option. Paramtrs Equilibrium Outcom Comparisons r E AH M EA M EAH E A wah w A x AH x A 0. 0. 0. 0.000 0.000 0.000 0.000 0. 0. 0.5 0.000 0.000 0.000 0.000 0. 0. 0.9 0.000 0.000 0.000 0.000 0. 0.5 0. 0.000 0.000 0.000 0.000 0. 0.5 0.5 0.000 0.000 0.000 0.000 0. 0.5 0.9 0.000 0.000 0.000 0.000 0. 0.9 0. 0.006-0.00 0.05 0.003 0. 0.9 0.5 0.008-0.003 0.06 0.03 0. 0.9 0.9 0.00-0.00 0.068 0.05 0. 0. 0. 0.000 0.000 0.000 0.000 0. 0. 0.5 0.000 0.000 0.000 0.000 0. 0. 0.9 0.000 0.000 0.000 0.000 0. 0.5 0. 0.05 0.035-0. 0. 0. 0.5 0.5 0.006 0.06-0.093 0.08 0. 0.5 0.9 0.003 0.03-0.5 0.09 0. 0.9 0. 0.090 0.03 0.06 0.36 0. 0.9 0.5 0.070 0.03 0.057 0.3 0. 0.9 0.9 0.060 0.05 0.03 0.5 0.7 0. 0. 0.000 0.000 0.000 0.000 0.7 0. 0.5 0.000 0.000 0.000 0.000 0.7 0. 0.9 0.000 0.000 0.000 0.000 0.7 0.5 0. 0.06 0.039-0.5 0.5 0.7 0.5 0.5 0.05 0.037-0.0 0.0 0.7 0.5 0.9 0.009 0.036-0.0 0.08 0.7 0.9 0. 0.0 0.05-0.05 0.5 0.7 0.9 0.5 0.096 0.037 0.006 0.9 0.7 0.9 0.9 0.078 0.037 0.000 0.88 Tabl A. Equilibrium Outcoms gular Slling vs. Avanc Slling wit t Holbac Option Paramtrs Equilibrium Outcom Comparisons r E B M EAH M E B EAH Ew B w AH Ex B x AH 0. 0. 0. 0.00-0.00 0.00-0.005 0. 0. 0.5 0.006-0.003 0.09-0.0 0. 0. 0.9 0.008-0.003 0.085-0.03 0. 0.5 0. 0.00-0.00 0.00-0.005 0. 0.5 0.5 0.006-0.00 0.05-0.06 0. 0.5 0.9 0.008-0.005 0.097-0.09 0. 0.9 0. 0.00-0.00 0.033-0.05 0. 0.9 0.5 0.00-0.005 0.063-0.075 0. 0.9 0.9 0.006-0.009 0.093-0.05 0. 0. 0. 0.00 0.00 0.000 0.000 0. 0. 0.5 0.00 0.00-0.05 0.008 0. 0. 0.9 0.007 0.003 0.05-0.008 0. 0.5 0. 0.06-0.00 0.6-0.5 0. 0.5 0.5 0.0-0.06 0.08-0.089 0. 0.5 0.9 0.0-0.05 0.5-0.0 0. 0.9 0. 0.005 0.005 0.0-0.73 0. 0.9 0.5 0.05 0.007 0.0-0.67 0. 0.9 0.9 0.07-0.00 0.06-0.96 0.7 0. 0. 0.00 0.00 0.000 0.000 0.7 0. 0.5 0.00 0.00 0.000 0.000 0.7 0. 0.9 0.00 0.003 0.00-0.00 0.7 0.5 0. 0.06-0.03 0.5-0.5 0.7 0.5 0.5 0.06-0.0 0.0-0.0 0.7 0.5 0.9 0.0-0.03 0.3-0.5 0.7 0.9 0. 0.00-0.00 0.05-0.5 0.7 0.9 0.5 0.00 0.008 0.005-0.00 0.7 0.9 0.9 0.0-0.00 0.0-0.09 Tabl A3. Comparativ Statics o t Gam unr Avanc Slling wit Holbac " # " r # " # x AH 56 0 3 88 35 07 0 309 w AH 0 56 3 35 88 07 39 8 309 E AH 56 0 3 59 35 07 5 96 309 E AH M 56 0 3 0 509 07 0 309 8