Online Supplement for Dynamic Multi-Technology. Production-Inventory Problem with Emissions Trading

Similar documents
Epistemic Game Theory: Online Appendix

UNIVERSITAT AUTÒNOMA DE BARCELONA MARCH 2017 EXAMINATION

Solution in semi infinite diffusion couples (error function analysis)

( ) () we define the interaction representation by the unitary transformation () = ()

Graduate Macroeconomics 2 Problem set 5. - Solutions

arxiv: v1 [cs.sy] 2 Sep 2014

V.Abramov - FURTHER ANALYSIS OF CONFIDENCE INTERVALS FOR LARGE CLIENT/SERVER COMPUTER NETWORKS

F-Tests and Analysis of Variance (ANOVA) in the Simple Linear Regression Model. 1. Introduction

(,,, ) (,,, ). In addition, there are three other consumers, -2, -1, and 0. Consumer -2 has the utility function

. The geometric multiplicity is dim[ker( λi. number of linearly independent eigenvectors associated with this eigenvalue.

Comb Filters. Comb Filters

Dynamic Team Decision Theory. EECS 558 Project Shrutivandana Sharma and David Shuman December 10, 2005

In the complete model, these slopes are ANALYSIS OF VARIANCE FOR THE COMPLETE TWO-WAY MODEL. (! i+1 -! i ) + [(!") i+1,q - [(!

Tight results for Next Fit and Worst Fit with resource augmentation

Online Appendix for. Strategic safety stocks in supply chains with evolving forecasts

CH.3. COMPATIBILITY EQUATIONS. Continuum Mechanics Course (MMC) - ETSECCPB - UPC

Robust and Accurate Cancer Classification with Gene Expression Profiling

Testing a new idea to solve the P = NP problem with mathematical induction

. The geometric multiplicity is dim[ker( λi. A )], i.e. the number of linearly independent eigenvectors associated with this eigenvalue.

GENERATING CERTAIN QUINTIC IRREDUCIBLE POLYNOMIALS OVER FINITE FIELDS. Youngwoo Ahn and Kitae Kim

FI 3103 Quantum Physics

John Geweke a and Gianni Amisano b a Departments of Economics and Statistics, University of Iowa, USA b European Central Bank, Frankfurt, Germany

Modeling and Solving of Multi-Product Inventory Lot-Sizing with Supplier Selection under Quantity Discounts

Lecture 18: The Laplace Transform (See Sections and 14.7 in Boas)

Comparison of Differences between Power Means 1

CS286.2 Lecture 14: Quantum de Finetti Theorems II

Appendix H: Rarefaction and extrapolation of Hill numbers for incidence data

( ) [ ] MAP Decision Rule

Existence and Uniqueness Results for Random Impulsive Integro-Differential Equation

Density Matrix Description of NMR BCMB/CHEM 8190

ECON 8105 FALL 2017 ANSWERS TO MIDTERM EXAMINATION

Advanced Machine Learning & Perception

Volatility Interpolation

Linear Response Theory: The connection between QFT and experiments

Should Exact Index Numbers have Standard Errors? Theory and Application to Asian Growth

2/20/2013. EE 101 Midterm 2 Review

Advanced Macroeconomics II: Exchange economy

Modern Dynamic Asset Pricing Models

On elements with index of the form 2 a 3 b in a parametric family of biquadratic elds

Teaching Notes #2 Equilibrium with Complete Markets 1

Lecture 2 M/G/1 queues. M/G/1-queue

Econ107 Applied Econometrics Topic 5: Specification: Choosing Independent Variables (Studenmund, Chapter 6)

Lecture 6: Learning for Control (Generalised Linear Regression)

Chapter Lagrangian Interpolation

Cubic Bezier Homotopy Function for Solving Exponential Equations

Density Matrix Description of NMR BCMB/CHEM 8190

Example: MOSFET Amplifier Distortion

TSS = SST + SSE An orthogonal partition of the total SS

Variants of Pegasos. December 11, 2009

First-order piecewise-linear dynamic circuits

Reactive Methods to Solve the Berth AllocationProblem with Stochastic Arrival and Handling Times

Let s treat the problem of the response of a system to an applied external force. Again,

Knowing What Others Know: Coordination Motives in Information Acquisition Additional Notes

Supporting information How to concatenate the local attractors of subnetworks in the HPFP

PHYS 705: Classical Mechanics. Canonical Transformation

Political Economy of Institutions and Development: Problem Set 2 Due Date: Thursday, March 15, 2019.

( t) Outline of program: BGC1: Survival and event history analysis Oslo, March-May Recapitulation. The additive regression model

Chapters 2 Kinematics. Position, Distance, Displacement

Oligopoly with exhaustible resource input

Optimal environmental charges under imperfect compliance

The safety stock and inventory cost paradox in a stochastic lead time setting

On One Analytic Method of. Constructing Program Controls

MALAYSIAN JOURNAL OF MATHEMATICAL SCIENCES. Institute for Mathematical Research, Universiti Putra Malaysia, UPM Serdang, Selangor, Malaysia

Solutions to exam in SF1811 Optimization, Jan 14, 2015

Midterm Exam. Thursday, April hour, 15 minutes

2 Aggregate demand in partial equilibrium static framework

Notes on the stability of dynamic systems and the use of Eigen Values.

ON THE WEAK LIMITS OF SMOOTH MAPS FOR THE DIRICHLET ENERGY BETWEEN MANIFOLDS

Dual Approximate Dynamic Programming for Large Scale Hydro Valleys

[ ] 2. [ ]3 + (Δx i + Δx i 1 ) / 2. Δx i-1 Δx i Δx i+1. TPG4160 Reservoir Simulation 2018 Lecture note 3. page 1 of 5

A Deza Frankl type theorem for set partitions

Li An-Ping. Beijing , P.R.China

Lecture VI Regression

CS434a/541a: Pattern Recognition Prof. Olga Veksler. Lecture 4

The Dynamic Programming Models for Inventory Control System with Time-varying Demand

Efficient Asynchronous Channel Hopping Design for Cognitive Radio Networks

EXECUTION COSTS IN FINANCIAL MARKETS WITH SEVERAL INSTITUTIONAL INVESTORS

Part II CONTINUOUS TIME STOCHASTIC PROCESSES

Technical Appendix to The Equivalence of Wage and Price Staggering in Monetary Business Cycle Models

Chapter 6 DETECTION AND ESTIMATION: Model of digital communication system. Fundamental issues in digital communications are

HEAT CONDUCTION PROBLEM IN A TWO-LAYERED HOLLOW CYLINDER BY USING THE GREEN S FUNCTION METHOD

Chapter 2 Linear dynamic analysis of a structural system

10. A.C CIRCUITS. Theoretically current grows to maximum value after infinite time. But practically it grows to maximum after 5τ. Decay of current :

How about the more general "linear" scalar functions of scalars (i.e., a 1st degree polynomial of the following form with a constant term )?

A Tour of Modeling Techniques

CONSISTENT ESTIMATION OF THE NUMBER OF DYNAMIC FACTORS IN A LARGE N AND T PANEL. Detailed Appendix

On the numerical treatment ofthenonlinear partial differentialequation of fractional order

Existence of Time Periodic Solutions for the Ginzburg-Landau Equations. model of superconductivity

Ordinary Differential Equations in Neuroscience with Matlab examples. Aim 1- Gain understanding of how to set up and solve ODE s

FTCS Solution to the Heat Equation

# c i. INFERENCE FOR CONTRASTS (Chapter 4) It's unbiased: Recall: A contrast is a linear combination of effects with coefficients summing to zero:

Demographics in Dynamic Heckscher-Ohlin Models: Overlapping Generations versus Infinitely Lived Consumers*

Multi-priority Online Scheduling with Cancellations

Approximate Analytic Solution of (2+1) - Dimensional Zakharov-Kuznetsov(Zk) Equations Using Homotopy

Mechanics Physics 151

EEL 6266 Power System Operation and Control. Chapter 5 Unit Commitment

General viscosity iterative method for a sequence of quasi-nonexpansive mappings

J i-1 i. J i i+1. Numerical integration of the diffusion equation (I) Finite difference method. Spatial Discretization. Internal nodes.

Mechanics Physics 151

Mechanics Physics 151

Transcription:

Onlne Supplemen for Dynamc Mul-Technology Producon-Invenory Problem wh Emssons Tradng by We Zhang Zhongsheng Hua Yu Xa and Baofeng Huo Proof of Lemma For any ( qr ) Θ s easy o verfy ha he lnear programmng problem n equaon (5) s feasble and does no have he unbound opmal value. I s clear ha he rank of he coeffcen marx for (5) s. In oher words each basc feasble soluon has no more han non-zero elemens. Therefore s opmal o produce wh no more han wo echnologes for any ( qr ) Θ. Proof of Lemma From he dscusson n subsecon 4. and he algorhm ha deermnes he effecve echnology se S we know ha f r μ q r μ ( = L m ) hen s opmal o produce wh echnologes and n S. The producon quanes of echnologes and are q = ( r μq) ( μ μ) and q = ( μ q r) ( μ μ) respecvely and he mnmum producon cos s ha F( qr ) = ( μ q r)( c c ) ( μ μ ) + c q= ω q vr = L m. When fndng echnology + n S Sep 3 n he algorhm ha deermnes S mples Consequenly for = L m v ( c c ) ( μ μ ) > ( c c ) ( μ μ ) = L m. + + + + c c c c ( c c ) ( c c ) c c + + + + + + + v+ = = > μ+ μ+ μ μ+ ( μ μ+ ) ( μ μ+ ) μ μ+ I follows ha ω+ ω+ = ( c+ + μ+ v+ ) ( c+ + μ+ v+ ) = μ+ v+ μ+ v+ >. Snce μ > L > μ m and < c < L < cm we have < v < L < vm and < ω < L < ω m.

Proof of Lemma 3 We frs show he convexy of F( qr ). For any r and r le q [ r μ r μ ] and q [ r μ r μ ]. Lemma mples ha where F( q r ) = c q + cq and F( q r ) = c q + c q q + q = q μ q + μq = r q + q = q and μ q + μq = r Therefore for any λ []. λf( q r ) + ( λ) F( q r ) = c ( λq ) + c ( λq ) + c [( λ) q ] + c [( λ) q ] F[ λq + ( λ) q λr + ( λ) r ]. The nequaly holds because λq λ q ( λ) q and ( ) ha mnmze he producon cos under gven ( λq + ( λ) q λr + ( λ) r ). Consequenly F( qr ) s convex. r λ q are he feasble soluons Nex we prove he submodulary of F( qr ). We only need o show ha for all r q q and F( q r ) F( q r ) F( q r ) F( q r ). (A.) We verfy wheher equaon (A.) holds or no n wo cases: () here exss such ha r μ q q r μ r μ q r μ < r μ ; and () here exs and such ha +. q r μ + + Case In hs case r μ q q r μ ( ) ( ) ω ω ω( ). r μ q q r μ ; r μ. q r μ < r μ + q r μ + + F q r F q r = q vr q + vr = q q We dscuss wo subcases: () here exss such ha such ha Snce Subcase In hs subcase r r r μ q q r μ and () here exs and F( q r ) F( q r ) = ω q v r ω q + v r = ω ( q q ). s clear ha ω ω. Consequenly Subcase I s clear ha μ μ or equvalenly. From Lemma hs mples ha F( q r) Fq ( r) = ω ( q q) ω ( q q) = Fq ( r) Fq ( r). r μ q r μ < r μ q r μ + + + r q r < L < r + q r μ μ μ μ. From Subcase we have F( r μ r) Fq ( r) Fr ( μ r) Fq ( r)

F( r μ r ) F( r μ r ) F( r μ r ) F( r μ r ) = L + + + F( q r ) F( r μ r ) F( q r ) F( r μ r ). Combnng he above equaons we have Case + + F q r F q r F q r F q r r μ q r μ < r μ q r μ ( ) ( ) ( ) ( ). + + + From Case we know ha F( r μ r ) F( q r ) F( r μ r ) F( q r ) F( r μ r ) F( r μ r ) F( r μ r ) F( r μ r ) = L + + + Combnng he above equaons we have F( q r ) F( r μ r ) F( q r ) F( r μ r ). + + F q r F q r F q r F q r ( ) ( ) ( ) ( ). Equaon (A.) holds n boh Case and Case. Therefore F( qr ) s submodular. In summary F( qr ) s convex and submodular. Proof of Lemma 5 Le W % ( x z) = W( x z) and H% ( y w ) = H ( y w ). Then W% ( x z) = mn { F( q r) + H% ( x q z r)} ( qr ) Θ = mn q+ y= x r+ w= z ( q r y w) Ω{ F( qr ) + H % ( yw )} where Ω= {( qryw ) : r rμ q rμm yw }. I can be verfed ha Ω s a nonempy closed convex sublace. The supermodulary and convexy of H mply ha H % s submodular and convex. From Theorem n Chen e al. (3) we have ha W% ( x z) s submodular and convex n. I follows ha W( x z ) s supermodular and convex. Proof of Proposon The supermodulary and convexy of H ( y w K ) W ( x z K ) and V ( x z K ) can be proved by nducon. I s clear ha V ( ) T + x z s supermodular and convex. Suppose ha V ( ) + x z K s supermodular and convex. Ths mples ha H ( y w K) = G ( y) + E [ V ( y D w K% )] s also supermodular and convex because () γ + + G ( y) s convex and () boh supermodulary and convexy are preserved under expecaon and 3

lnear operaon. As shown n Lemma F( qr ) s submodular and convex. Lemma 5 mples ha W ( x z K) = mn{ F( q r) + H ( x+ q z r K ):( q r) Θ} s supermodular and convex so W ( x z K ) s submodular and convex n ( x z ). Snce C s a convex funcon C ( z z) s submodular and convex n ( zz ) (Topks 998). I s well known ha submodulary and convexy are preserved under mnmzaon. Accordngly we have he submodulary and convexy of V ( x z K) = mn{ C ( z z K) + W ( x z K): z } n ( x z ). Ths mples ha V ( x z K ) s supermodular and convex. In summary H ( y w K ) W ( x z K ) and V ( x z K ) are boh supermodular and convex under any gven K for all perods. Proof of Theorem Before provng Theorem we rewre problem (3) based on he formula of F( qr ). For = L m we defne W ( x z K) mn μ μ { ω y+ vw+ H ( y w K)} ω x v z w z( z w) y x ( z w) V ( x z K) mn{mn { K ( z z) + W ( x z K)}mn { k ( z z) + W ( x z K)}}. z z z z Then problem (3) can be rewren as V ( x z K) = mn{ V ( x z K) L V ( x z K)}. m Here V ( x z K ) s he opmal cos for he problem usng us echnologes and. Based on Lemma 5 s easy o verfy he supermodulary and convexy of W ( x z K ) and V ( x z K ). Defne he followng hresholds for he allowance level: L ( x K) argmn { K z W ( x z K)} z + K z + z + z + U ( x ) argmn { k z W ( x z K)} = L m l ( x K) sup{ z :( W ( x z K)) < v} u ( x K) nf{ z :( W ( x z K)) > v} = L m. I s easy o verfy ha () L ( x K) U ( x K ) ( k K ) and () L ( x K) l ( x K) u ( x K) U ( x K ) f k < v < K from he convexy of W ( x z K ). Smlar o he proof of Theorems -4 n Gong and Zhou (3) can be proved from s supermodulary and convexy ha he opmal soluons o V ( x z K ) have he followng properes leng y and z be he opmal soluons o V ( x z K ): 4

LEMMA A.. The followng resuls hold: () If v K hen produce exclusvely wh echnology : when z U ( x K ) sell allowances down o z = U ( x K ) and hen produce up o ˆU y = max{ S ( K ) x}; when L ( x K) < z< U ( x K ) do no rade allowances and hen produce up o y = max{ s ( μ x+ z K ) x}; and when z L ( x K ) purchase allowances up o z = L ( x K ) and hen produce up o ˆ L y = max{ S ( K ) x}. () If v k hen produce exclusvely wh echnology : when z U ( x K ) sell allowances down o z = U ( x K ) and hen produce up o max{ ˆU y = S ( K ) x}; when L ( x K) < z< U ( x K ) do no rade allowances and hen produce up o y = max{ s ( μ x+ z K ) x}; and when z L ( x K ) purchase allowances up o z = L ( x K ) and hen produce up o y = max{ S ( K ) x}. 5 ˆ L () If k < v < K hen when z U ( x K ) sell allowances down o z = U ( x K ) and hen produce exclusvely wh echnology up o ˆU y = max{ S ( K ) x}; when u ( x K) z< U ( x K ) do no rade allowances and hen produce exclusvely wh echnology up o y = max{ s ( μ x+ z K ) x}; when l ( x K) < z< u ( x K ) do no rade allowances and hen produce [ μx+ z μs( K) w( K )] ( μ μ) uns wh echnology and [ μ S( K) + w( K ) μ x z] ( μ μ) uns wh echnology up o y = S ( K ) ; when L ( x K) < z l ( x K ) do no rade allowances and hen produce exclusvely wh echnology up o purchase allowances up o o y = max{ S ( K ) x}. ˆ L y = max{ s ( μ x+ z K ) x}; and when z L ( x K ) z = L ( x K ) and hen produce exclusvely wh echnology up The defnons of ˆ U S ( K ) ˆ L S ( K ) s ( μ x+ z K ) S ( K ) and w ( K ) can be found n Secon 5. In Lemma A. L ( x K ) U ( x K ) l ( x K ) and u ( x K ) have he followng specfc formulas: ˆL + L ( ) ( ( ) ) (max{ ˆL L x K = μ S K x + w S ( K) x}) ˆU + U ( ) ( ( ) ) (max{ ˆU U x K = μ S K x + w S ( K ) x}) = L m + l ( x K) = μ ( S ( K) x) + w (max{ S ( K) x}) u ( x K) = μ ( S ( K) x) + w (max{ S ( K ) x}) = L m. + Noe ha x< S ( K ) when l ( x K) < z< u ( x K ). From Lemma A. we know ha he opmal soluons o V ( x z K ) have he followng properes: () If v K hen produce

exclusvely wh echnology () If v k hen produce exclusvely wh echnology () If k < v < K hen (a) produce exclusvely wh echnology when z u ( x K ) (b) produce wh echnologes and smulaneously when l ( x K) < z< u ( x K ) and (c) produce exclusvely wh echnology when z l ( x K ). Lemma A. shows he opmal soluons y z and he correspondng echnology selecon for V ( x z K ). Please noe ha producng wh echnology ( = L m ) s feasble o boh V ( x z K ) and V+ ( x z K ). Consequenly he opmal soluons o V ( x z K) = mn{ V ( x z K ) L V ( x z K )} can be characerzed by dscussng he opmal m soluons o V ( x z K) L V ( x z K) as shown n Lemma A.. For example f he opmal m soluon o V ( x z K ) s also feasble o V+ ( x z K ) hen mn{ V ( x z K) V + ( x z K)} = V+ ( x z K ). We nex characerze he srucures of he opmal polces when v K vm k and v k K v + respecvely. Case K v < L < vm From Lemma A.() we know ha when v K he opmal soluon o V ( x z K ) s producng exclusvely wh echnology. Snce producng exclusvely wh echnology ( = 3 L m ) s feasble o boh V ( x z K ) and V ( x z K ) we have V ( x z K ) = mn{ V ( x z K) L V ( x z K)} = V ( x z K). Consequenly he opmal polces have he m followng srucure: when z U ( ) x K sell allowances down o z ( ) = U x K and hen produce exclusvely wh echnology up o ˆ U y = max{ S ( K ) x}; when L ( x K ) < z < U ( ) x K do no rade allowances and hen produce exclusvely wh echnology up o y = max{ s ( μx+ z K ) x}; and when z L ( ) x K purchase allowances up o ( ) z = L x K and hen produce exclusvely wh echnology up o ˆ L y = max{ S ( K ) x}. Case v < L < vm k From Lemma A.() we know ha when v k he opmal soluon o V ( x z K ) s producng exclusvely wh echnology. Snce producng exclusvely wh echnology ( = L m ) s feasble o boh V ( x z K ) and V+ ( x z K ) we have m m V ( x z K) = mn{ V ( x z K) L V ( x z K)} = V ( x z K). Consequenly he opmal polces have he followng srucure: when z U ( x K ) sell allowances down o z = U ( x K ) and hen produce exclusvely wh echnology m up o m m 6

y = max{ S ( K ) x}; when L ( x K) < z< U ( x K ) do no rade allowances and hen ˆU m m m produce exclusvely wh echnology m up o y = max{ s ( μ x+ z K ) x}; and when m m z Lm ( x K ) purchase allowances up o z = Lm( x K ) and hen produce exclusvely wh echnology m up o max{ ˆ L y = S ( K ) x}. Case 3 v < L< v k K v < L < v + m Accordng o he analyses n Cases and we have m V( xz K) = mn{ V ( xz K) L V ( xz K)} = mn{ V ( xz K) V ( xz K)}. m + From Lemma A.( ) we know ha he opmal soluons o V ( ) x z K and V + ( x z K ) are he same. Thus he srucure of he opmal polces s: when z U ( ) x K sell allowances down o z = U ( x K ) and hen produce exclusvely wh echnology up o y = max{ S ( K ) x}; when ˆU L ( x K) < z< U ( x K ) do no rade allowances and hen produce exclusvely wh echnology up o y = max{ s ( ) }; μ x+ z K x and when ( ) z L x K purchase allowances up o z ( ) = L x K and hen produce exclusvely wh echnology up o ˆ L y = max{ S ( ) }. K x Proof of Theorem Recall ha v L v k v L v K v L v Smlar o Cases and n he < < < + < < < + < < m. proof of Theorem we have We frs consder v mn{ V ( x z K) L V ( x z K)} = V ( x z K) (A.9) mn{ V ( x z K) L V ( x z K)} = V ( x z K). (A.) + m + mn{ V ( x z K) V ( x z K )} and + mn{ V ( x z K) V ( x z K )}. Snce + k Lemma A.() ndcaes ha he opmal soluon o V ( x z K ) s producng exclusvely wh echnology. I s clear ha hs soluon s also feasble o Consequenly Smlarly from + + V ( ). + x z K mn{ V ( x z K) V ( x z K)} = V ( x z K ). (A.) v < K v + and Lemma A.() we have Concludng equaons (A.9-) we have mn{ V ( x z K) V ( x z K)} V ( x z K ). (A.) + = V ( x z K) = mn{ V ( x z K) L V ( x z K)} = mn{ V ( x z K) L V ( x z K)}. (A.3) m + If = + Lemma A.() ndcaes ha he opmal polces have he same srucure as 7

shown n Theorem. We nex consder he case wh = +. Snce k < v + ( v + < K) we have c + + kμ + > c + k μ ( c + + μ + K > c + + μ K + ). From he defnons of ˆ U S ( ) K and ˆ L S ( K ) we have ˆU ˆU S + ( K) S ( K ) and ˆL ( ) ˆL S K S ( K ). Therefore + + Defne U + ( x K) U ( ) x K and L + ( x K) L + ( x K ) for all x. (A.4) y% max{ s + ( μ + x+ z K ) x}. Now we dscuss he relaonshp beween u + ( x K ) and l + ( x K ). From Lemma A. we have W + ( x z K) = c ( y% + x) + H( y% μ + x+ z μ + y% K) when u + ( x K) z< U + ( x K ) W + ( x z K) = c ( y% + x) + H( y% μ + x+ z μ + y% K) when L + ( x K) < z l + ( x K ). I s clear ha u + ( x K ) l ( ) + x K and L ( ) + x K can be rewren as: u ( x K) = nf{ z :( W ( x z K)) > v u ( x K) z< U ( x K)} Snce + + + z + + + l ( x K) = sup{ z :( W ( x z K)) < v L ( x K) < z l ( x K)} + + + z + + + L ( x K) = nf{ z : ( W ( x z K)) > K L ( x K) < z l ( x K)}. + + + z + + + < + < he convexes of W + ( x z K ) and W ( x z ) v v K L ( x K) u ( x K) l ( x K ) for all x. + + + + K mply ha Recall ha Ln ( x K) ln ( x K) un ( x K) Un ( x K ) n= + +. We nex characerze he srucure of he opmal polces when = + n fve cases. (A.5) Case z U ( ) x K Lemma A.() ndcaes ha when z U ( x K ) s opmal o produce exclusvely wh echnology ( = + + ) and sell allowances. From (A.4)) we have mn{ V ( x z K) V ( x z K)} = V ( x z K ) when + + + U ( x ) U ( x K ) (equaon + K z U ( ) x K (A.6) Because producng exclusvely wh echnology + wh sellng allowances s feasble o V ( ) + x z K when z U ( ). x K Equaon (A.6) ndcaes ha when z U ( ) x K s opmal o sell allowances down o level z = U ( x K ) and hen produce exclusvely wh echnology up o U y = max{ S ( ) } K x. Case l + ( x K) < z< U ( ) x K Snce z l ( x K) u ( x K ) (equaon (A.5)) Lemma A.() ndcaes ha he ˆ + + opmal soluon o V + ( x z K ) s producng exclusvely wh echnology +. Ths soluon s clearly feasble o V ( ) + x z K. Hence 8

mn{ V + ( x z K) V + ( x z K)} = V + ( x z K ) when l + ( x K) < z< U ( x K ). Ths mples ha: () when u ( ) ( ) + x K z< U x K do no rade allowances and hen produce exclusvely wh echnology up o y = max{ s ( ) }; μ x+ z K x and () when l ( x K) < z< u ( x K) do no rade allowances and hen produce + + ( μ + x+ z μ + S+ ( K) w + ( K )) ( μ μ + ) uns wh echnology and ( μ S ( K) + w ( K ) μ x z) ( μ μ ) uns wh echnology +. + + + Case 3 u ( x K) z l ( x K ) + + Snce L + ( x K) u + ( x K ) (equaon (A.5)) and l + ( x K) U + ( x K ) Lemma A.() mples ha V + ( x z K ) and V + ( x z K ) have he same opmal soluon when u ( x K) z l ( x K ). Therefore when u ( x K) z l ( x K ) do no rade + + + + allowances and hen produce exclusvely wh echnology + up o y = max{ s ( μ x+ z K ) x}. + + Case 4 Snce L ( x K) < z< u ( x K ) + + z u ( x K) l ( x K ) (equaon (A.5)) Lemma A.() suggess ha he + + opmal soluon o V ( ) + x z K s producng exclusvely wh echnology +. Ths soluon s also feasble o V + ( x z K ). Consequenly mn{ V + ( x z K) V + ( x z K)} = V + ( x z K ) when L + ( x K) < z< u + ( x K ). Ths mples ha: () when L + ( x K) z< l + ( x K ) do no rade allowances and hen produce exclusvely wh echnology + up o y = max{ s ( μ x+ z K ) x}; and () when + + + + l ( x K) < z< u ( x K) do no rade allowances and hen produce ( μ x+ z μ S ( K) w ( K)) ( μ μ ) uns wh echnology + and + + + + + + ( μ S ( K) + w ( K ) μ x z) ( μ μ ) uns wh echnology +. + + + + + + Case 5 z L + ( x) Lemma A.() ndcaes ha s opmal o produce exclusvely wh echnology ( = + + ) and purchase allowances when z L ( x K ). From L + ( x K) L + ( x K ) (equaon (A.4)) we have mn{ V ( x z K) V ( x z K)} = V ( x z K ) when + + + z L + ( x K ) (A.7) because producng exclusvely wh echnology + wh purchasng allowances s feasble o V + ( x z K ) when z L + ( x K ). Equaon (A.7) ndcaes ha: when z L + ( x K ) s opmal o purchase allowances up o z = L ( x K ) and hen produce exclusvely wh + 9

echnology + up o y = max{ Sˆ ( K ) x}. L + So far we have fully characerzed he opmal polces for he case wh = + whch has he same srucure as shown n Theorem. Smlarly for > + can be proved ha he opmal polcy has he same srucure as shown n Theorem. Deals are omed for brevy.