ON THE TIMING OPTION IN A FUTURES CONTRACT. FRANCESCA BIAGINI Dipartimento di Matematica, Università dibologna

Similar documents
LIDSTONE IN THE CONTINUOUS CASE by. Ragnar Norberg

The Mathematics Of Stock Option Valuation - Part Four Deriving The Black-Scholes Model Via Partial Differential Equations

Matematisk statistik Tentamen: kl FMS170/MASM19 Prissättning av Derivattillgångar, 9 hp Lunds tekniska högskola. Solution.

UCLA Department of Economics Fall PhD. Qualifying Exam in Macroeconomic Theory

INSTITUTE OF ACTUARIES OF INDIA

Black-Scholes Model and Risk Neutral Pricing

May 2007 Exam MFE Solutions 1. Answer = (B)

Proceedings of the 48th European Study Group Mathematics with Industry 1

Pricing formula for power quanto options with each type of payoffs at maturity

Models of Default Risk

Introduction to Black-Scholes Model

Pricing FX Target Redemption Forward under. Regime Switching Model

MA Advanced Macro, 2016 (Karl Whelan) 1

Dual Valuation and Hedging of Bermudan Options

Bond Prices and Interest Rates

Macroeconomics II A dynamic approach to short run economic fluctuations. The DAD/DAS model.

FINAL EXAM EC26102: MONEY, BANKING AND FINANCIAL MARKETS MAY 11, 2004

Option pricing and hedging in jump diffusion models

You should turn in (at least) FOUR bluebooks, one (or more, if needed) bluebook(s) for each question.

Economic Growth Continued: From Solow to Ramsey

MAFS Quantitative Modeling of Derivative Securities

Modeling of Tradeable Securities with Dividends

Modeling of Tradeable Securities with Dividends

Computations in the Hull-White Model

Equivalent Martingale Measure in Asian Geometric Average Option Pricing

Market Models. Practitioner Course: Interest Rate Models. John Dodson. March 29, 2009

Supplement to Models for Quantifying Risk, 5 th Edition Cunningham, Herzog, and London

A pricing model for the Guaranteed Lifelong Withdrawal Benefit Option

ANSWER ALL QUESTIONS. CHAPTERS 6-9; (Blanchard)

Problem 1 / 25 Problem 2 / 25 Problem 3 / 11 Problem 4 / 15 Problem 5 / 24 TOTAL / 100

INSTITUTE OF ACTUARIES OF INDIA

VALUATION OF THE AMERICAN-STYLE OF ASIAN OPTION BY A SOLUTION TO AN INTEGRAL EQUATION

Tentamen i 5B1575 Finansiella Derivat. Torsdag 25 augusti 2005 kl

Option Valuation of Oil & Gas E&P Projects by Futures Term Structure Approach. Hidetaka (Hugh) Nakaoka

ECON Lecture 5 (OB), Sept. 21, 2010

Brownian motion. Since σ is not random, we can conclude from Example sheet 3, Problem 1, that

Change of measure and Girsanov theorem

Erratic Price, Smooth Dividend. Variance Bounds. Present Value. Ex Post Rational Price. Standard and Poor s Composite Stock-Price Index

CHRISTOPH MÖHR ABSTRACT

EXPONENTIAL MARTINGALES AND TIME INTEGRALS OF BROWNIAN MOTION

IJRSS Volume 2, Issue 2 ISSN:

Fundamental Basic. Fundamentals. Fundamental PV Principle. Time Value of Money. Fundamental. Chapter 2. How to Calculate Present Values

Continuous-time term structure models: Forward measure approach

Macroeconomics. Part 3 Macroeconomics of Financial Markets. Lecture 8 Investment: basic concepts

PARAMETER ESTIMATION IN A BLACK SCHOLES

An Incentive-Based, Multi-Period Decision Model for Hierarchical Systems

arxiv:math/ v2 [math.pr] 26 Jan 2007

A Theory of Tax Effects on Economic Damages. Scott Gilbert Southern Illinois University Carbondale. Comments? Please send to

Dynamic Programming Applications. Capacity Expansion

A True Buyer s Risk and Classification of Options

Interest Rate Products

t=1 C t e δt, and the tc t v t i t=1 C t (1 + i) t = n tc t (1 + i) t C t (1 + i) t = C t vi

Tentamen i 5B1575 Finansiella Derivat. Måndag 27 augusti 2007 kl Answers and suggestions for solutions.

Completeness of a General Semimartingale Market under Constrained Trading

Problem Set 1 Answers. a. The computer is a final good produced and sold in Hence, 2006 GDP increases by $2,000.

Chapter Outline CHAPTER

1 Purpose of the paper

PDE APPROACH TO VALUATION AND HEDGING OF CREDIT DERIVATIVES

2. Quantity and price measures in macroeconomic statistics 2.1. Long-run deflation? As typical price indexes, Figure 2-1 depicts the GDP deflator,

Completeness of a General Semimartingale Market under Constrained Trading

Final Exam Answers Exchange Rate Economics

4452 Mathematical Modeling Lecture 17: Modeling of Data: Linear Regression

Appendix B: DETAILS ABOUT THE SIMULATION MODEL. contained in lookup tables that are all calculated on an auxiliary spreadsheet.

San Francisco State University ECON 560 Summer 2018 Problem set 3 Due Monday, July 23

DYNAMIC ECONOMETRIC MODELS Vol. 7 Nicolaus Copernicus University Toruń Krzysztof Jajuga Wrocław University of Economics

Alexander L. Baranovski, Carsten von Lieres and André Wilch 18. May 2009/Eurobanking 2009

An Analytical Implementation of the Hull and White Model

Forward Price models and implied Volatility Term Structures

THE HURST INDEX OF LONG-RANGE DEPENDENT RENEWAL PROCESSES. By D. J. Daley Australian National University

AN EASY METHOD TO PRICE QUANTO FORWARD CONTRACTS IN THE HJM MODEL WITH STOCHASTIC INTEREST RATES

A Note on Missing Data Effects on the Hausman (1978) Simultaneity Test:

Systemic Risk Illustrated

Spring 2011 Social Sciences 7418 University of Wisconsin-Madison

Risk-Neutral Probabilities Explained

Available online at ScienceDirect

FIXED INCOME MICHAEL MONOYIOS

Foreign Exchange, ADR s and Quanto-Securities

(c) Suppose X UF (2, 2), with density f(x) = 1/(1 + x) 2 for x 0 and 0 otherwise. Then. 0 (1 + x) 2 dx (5) { 1, if t = 0,

A dual approach to some multiple exercise option problems

Description of the CBOE S&P 500 2% OTM BuyWrite Index (BXY SM )

Lecture Notes to Finansiella Derivat (5B1575) VT Note 1: No Arbitrage Pricing

Unemployment and Phillips curve

Market and Information Economics

Optimal Early Exercise of Vulnerable American Options

A UNIFIED PDE MODELLING FOR CVA AND FVA

A BSDE approach to the Skorokhod embedding problem for the Brownian motion with drift

CURRENCY TRANSLATED OPTIONS

Reconciling Gross Output TFP Growth with Value Added TFP Growth

The Binomial Model and Risk Neutrality: Some Important Details

Area Yield Futures and Futures Options: Risk Management and Hedging.

a. If Y is 1,000, M is 100, and the growth rate of nominal money is 1 percent, what must i and P be?

Description of the CBOE Russell 2000 BuyWrite Index (BXR SM )

If You Are No Longer Able to Work

The Relationship between Money Demand and Interest Rates: An Empirical Investigation in Sri Lanka

Technological progress breakthrough inventions. Dr hab. Joanna Siwińska-Gorzelak

Valuation and Hedging of Correlation Swaps. Mats Draijer

Some Remarks on Derivatives Markets (third edition, 2013)

Lecture: Autonomous Financing and Financing Based on Market Values I

New Acceleration Schemes with the Asymptotic Expansion in Monte Carlo Simulation

OPTIMUM FISCAL AND MONETARY POLICY USING THE MONETARY OVERLAPPING GENERATION MODELS

Transcription:

Mahemaical Finance, Vol. 17, No. 2 (April 2007), 267 283 ON THE TIMING OPTION IN A FUTURES CONTRACT FRANCESCA BIAGINI Diparimeno di Maemaica, Universià dibologna TOMAS BJÖRK Deparmen of Finance, Sockholm School of Economics The iming opion embedded in a fuures conrac allows he shor posiion o decide when o deliver he underlying asse during he las monh of he conrac period. In his paper we derive, wihin a very general incomplee marke framework, an explici model independen formula for he fuures price process in he presence of a iming opion. We also provide a characerizaion of he opimal delivery sraegy, and we analyze some concree examples. KEY WORDS: fuures conrac, iming opion, opimal sopping 1. INTRODUCTION In sandard exbook reamens, a fuures conrac is ypically defined by he properies of zero spo price and coninuous (or discree) reselemen, plus a simple no arbirage condiion a he las delivery day. If he underlying price process is denoed by X and he fuures price process for delivery a T is denoed by F(, T) his leads o he well known formula (1.1) F(, T) = E Q X T F, 0 T, where Q denoes he (no necessarily unique) risk neural maringale measure. In pracice, however, here are a number of complicaing facors which are ignored in he exbook reamen, and in paricular i is ypically he case ha a sandard fuures conrac has several embedded opion elemens. The mos common of hese opions are he iming opion, and he end-of-he-monh opion, he qualiy opion, and he wild card opion. All hese opions are opions for he shor end of he conrac, and hey work roughly as follows. The iming opion is he opion o deliver a any ime during he las monh of he conrac. The end of he monh opion is he opion o deliver a any day during he las week of he conrac, despie he fac ha he fuures price for he las week is fixed on he firs day of ha week and hen held consan. Suppor from he Tom Hedelius and Jan Wallander Foundaion is graefully acknowledged. Boh auhors are graeful o B. Näslund, J. Kallsen, C. Kuehn, an anonymous associae edior, and an anonymous referee for anumber of very helpful commens and suggesions. Manuscrip received March 2005; final revision received November 2005. Address correspondence o Tomas Björk, Deparmen of Finance, Sockholm School of Economics, Box 6501, SE-113 83 Sockholm, Sweden; e-mail: omas.bjork@hhs.se. C 2007 The Auhors. Journal compilaion C 2007 Blackwell Publishing Inc., 350 Main S., Malden, MA 02148, USA, and 9600 Garsingon Road, Oxford OX4 2DQ, UK. 267

268 F. BIAGINI AND T. BJÖRK The qualiy opion is he opion o choose, ou of a prespecified baske of asses, which asse o deliver. The wild card opion is, for example for bond fuures, he opion o iniiae delivery beween 2 P.M. and 8 P.M. inhe afernoon during he delivery monh of he conrac. The poin here is ha he fuures price is seled a 2 P.M. buhe rade in he underlying bonds goes on unil 8 P.M. The purpose of he presen paper is o sudy he iming opion wihin a very general framework, allowing for incomplee markes, and our goal is o invesigae how he general formula (1.1) has o be modified when we inroduce a iming opion elemen. Our main resul is given in Theorems 4.2 and 4.3 where i is shown ha, independenly of any model assumpions, he fuures price in he presence of he iming opion is given by he formula (1.2) F(, T) = inf T EQ X F, where varies over he class of opional sopping imes, and inf denoes he essenial infimum. This formula is of course very similar o he pricing formula for an American opion. Noe, however, ha equaion (1.2) does no follow direcly from sandard heory for American conracs, he reason being ha he fuures price is no a price in he echnical sense. The fuures price process insead plays he role of he cumulaive dividend process for he fuures conrac, which in urn can be viewed as a price-dividend pair, wih spo price idenically equal o zero. Furhermore, we prove ha he opimal delivery policy ˆ(), for a shor conrac enered a,isgivenby (1.3) ˆ = inf{ 0; F(, T) = X }. We also sudy some special cases and show he following. If he underlying X is he price of a raded financial asse wihou dividends, hen i is opimal o deliver immediaely, so ˆ() = and hus (1.4) F(, T) = X. If he underlying X has a convenience yield which is greaer han he shor rae, hen he opimal delivery sraegy is o wai unil he las day. In his case we hus have ˆ() = T and (1.5) F(, T) = E Q X T F, which we recognize from (1.1) as he classical formula for fuures conracs wihou a iming opion. Opion elemens of fuures conrac have also been sudied earlier. The qualiy opion is discussed in deail in Gay and Manaser (1984), and he wild card opion is analyzed in Cohen (1995) and Gay and Manaser (1986). The iming opion is (among oher opics) reaed in Boyle (1989) bu heoreical resuls are only obained for he special case when X is he price process of a raded underlying asse. In his seing, and under he added assumpion of a consan shor rae, he formula (1.4) is derived. The organizaion of he paper is as follows. In Secion 2, we se he scene for he financial marke. Noe ha we make no specific model assumpions a all abou marke compleeness or he naure of he underlying process, and our seup allows for discree

ON THE TIMING OPTION IN A FUTURES CONTRACT 269 as well as coninuous ime models. In Secion 3, we derive a fundamenal equaion, he soluion of which will deermine he fuures price process. We aack he fundamenal equaion by firs sudying he discree ime case in Secion 4.1, and prove he main formula (1.2). In Secion 4.2, we prove he parallel resul in he echnically more demanding coninuous ime case. We finish he main paper by some concree financial applicaions, and in paricular we clarify compleely under which condiions he fuures price process, including an embedded iming opion, coincides wih he classical formula (1.1). A he oher end of he specrum, we also invesigae under which condiions immediae delivery is opimal. 2. SETUP We consider a financial marke living on a sochasic basis (,F, F, Q), where he filraion F ={F } 0 T saisfies he usual condiions. We allow for boh discree and coninuous ime, so he conrac period is eiher he inerval 0, T orhe se {0, 1,...,T}. To se he financial scene we need some basic assumpions, so for he res of he paper we assume ha here exiss a predicable shor rae process r, and a corresponding money accoun process B. Inconinuous ime B has he dynamics (2.1) db = r B d. In he discree ime case, he shor rae a ime will be denoed by r +1 so he bank accoun B has he dynamics (2.2) B +1 = (1 + r +1 )B. In his case he shor rae is assumed o be predicable, i.e., r is F 1 -measurable (r is known already a 1) for all, wih he convenion F 1 = F 0. The marke is assumed o be free of arbirage in he sense he measure Q above is a maringale measure w.r. he money accoun B for he given ime horizon. Noe ha we do no assume marke compleeness. Obviously, if he marke is incomplee, he maringale measure Q will no be unique, so in an incomplee seing he pricing formulas derived below will depend upon he paricular maringale measure chosen. We discuss his in more deail in Secion 5. We will need a weak boundedness assumpion on he shor rae. ASSUMPTION 2.1. For he res of he paper we assume he following. In he coninuous ime case we assume ha he ineres rae process is predicable, and ha here exiss a posiive real number c such ha (2.3) r c, wih probabiliy one, for all. Defining he money accoun as usual by B by B = exp( 0 r s ds) we assume ha (2.4) E Q B T <. In he discree ime case we assume he ineres rae process is predicable, and ha here exiss a posiive real number c such ha (2.5) 1 + r n c, wih probabiliy one, for all n.

270 F. BIAGINI AND T. BJÖRK (2.6) REMARK 2.1. We noe ha he if we define C by C = sup E Q B F, T where varies over he class of sopping imes, hen he inequaliy E Q B T < easily implies (2.7) C < Q a.s. for all 0, T. Wihin his framework we now wan o consider a fuures conrac wih an embedded iming opion. ASSUMPTION 2.2. We assume he exisence of an exogenously specified nonnegaive adaped cadlag process X. The process X will henceforh be referred o as he index process, and we assume ha (2.8) E Q X <, 0 T. The inerpreaion of his assumpion is ha he index process X is he underlying process on which he fuures conrac is wrien. For obvious reasons we wan o include conracs like commodiy fuures, index fuures, fuures wih an embedded qualiy opion, and also fuures on a nonfinancial index like aweaher fuures conrac. For his reason we do no assume ha X is he price process of a raded financial asse in an idealized fricionless marke. Typical choices of X could hus be one of he following. X is he price a ime of a commodiy, wih a nonrivial convenience yield. X is he price a ime of a, possibly dividend paying, financial asse. X = min{s 1,...,Sn } where S1,...,Sn are price processes of financial asses (for example socks or bonds). This seup would be naural if we have an embedded qualiy opion. X is a nonfinancial process, like he emperaure a some prespecified locaion. We now wan o define a fuures conrac, wih an embedded iming opion, on he underlying index process X over he ime inerval 0, T. If, for example, we are considering au.s. ineres rae fuure, his means ha he inerval 0, T corresponds o he las monh of he conrac period. Noe ha we hus assume ha he iming opion is valid for he enire inerval 0, T. The analysis of he fuures price process for imes prior o he iming opion period, is rivial and given by sandard heory. If, for example, we le he iming opion be acive only in he inerval T 0, T, hen we immediaely obain (2.9) F(, T) = E Q F(T 0, T) F, 0 T 0, where F(T 0, T) isgiven by he heory developed in he presen paper. We can now give formal definiion of he (coninuous ime) conrac. See below for he discree ime modificaion. DEFINITION 2.1. A fuures conrac on X wih final delivery dae T, including an embedded iming opion, on he inerval 0, T, wih coninuous reselemen, is a financial conrac saisfying he following clauses.

ON THE TIMING OPTION IN A FUTURES CONTRACT 271 A each ime 0, T here exiss on he marke a fuures price quoaion denoed by F(, T). Furhermore, for each fixed T, he process F(, T) isa semimaringale w.r.. he filraion F. Since T will be fixed in he discussion below, we will ofen denoe F(, T)byF. The holder of he shor end of he fuures conrac can, a any ime 0, T, decide wheher o deliver or no. The decision wheher o deliver a or no is allowed o be based upon he informaion conained in F. If he holder of he shor end decides o deliver a ime, she will pay he amoun X and receive he quoed fuures price F(, T). If delivery has no been made prior o he final delivery dae T, he holder of he shor end will pay X T and receive F(T, T). During he enire inerval 0, T here is coninuous reselemen as for a sandard fuures conrac. More precisely; over he infiniesimal inerval, + d he holder of he shor end will pay he amoun df(, T) = F( + d, T) F(, T). The spo price of he fuures conrac is always equal o zero, i.e., you can a any ime ener or leave he conrac a zero cos. The cash flow for he holder of he long end is he negaive of he cash flow for he shor end. The imporan poin o noice here is ha he iming opion is only an opion for he holder of he shor end of he conrac. For discree ime models, he only difference is he reselemen clause which hen says ha if you hold a shor fuure beween and + 1, you will pay he amoun F( + 1, T) F(, T) a ime + 1. Our main problem is he following. PROBLEM 2.1. Given an exogenous specificaion of he index process X, wha can be said abou he exisence and srucure of he fuures price process F(, T)? 3. THE FUNDAMENTAL PRICING EQUATION We now go on o reformulae Problem 2.1 in more precise mahemaical erms, and his will lead us o a fairly complicaed infinie dimensional sysem of equaions for he deerminaion of he fuures price process (if ha objec exiss). We focus on he coninuous ime case, he discree ime case being very similar. 3.1. The Pricing Equaion in Coninuous Time For he given final delivery dae T, le us consider a fixed poin in ime T and discuss he (coninuous ime) fuures conrac from he poin of view of he shor end of he conrac. From he definiion above, i is obvious ha he holder of he shor end has o decide on a delivery sraegy, and we formalize such a sraegy as a sopping ime, wih T, Q a.s. If he holder of he shor end uses he paricular delivery sraegy, hen he arbirage free value of her cash flows is given by he expression (3.1) E Q e r s ds (F X ) r s ds df u F.

272 F. BIAGINI AND T. BJÖRK The firs erm in he expecaion corresponds o he cash flow for he acual delivery, i.e., he shor end delivers X and receives he quoed fuures price F, and he inegral erm corresponds o he cash flow of he coninuous reselemen. Since he iming opion is an opion for he holder of he shor end, she will ry o choose he sopping ime so as o maximize he arbirage free value. Thus, he value of he shor end of he fuures conrac a ime is given by sup E Q e r s ds (F X ) r (3.2) s ds df u F, T where, for shor sup denoes he essenial supremum. We now recall ha, by definiion, he spo price of he fuures conrac is always equal o zero, and we have hus derived our fundamenal pricing equaion, which is in fac an equilibrium condiion for each. PROPOSITION 3.1. The fuures price process F, if i exiss, will saisfy for each 0, T, he fundamenal pricing equaion sup E Q e r s ds (F X ) r (3.3) s ds df u F = 0, T where varies over he class of sopping imes. Some remarks are now in order. REMARK 3.1. A firs sigh, equaion (3.3) may look like a sandard opimal sopping problem, bu i is in fac more complicaed han ha. Obviously, for a given fuures price process F, he lef-hand side of (3.3) represens a sandard opimal sopping problem, bu he poin here is ha he fuures process F is no an a priori given objec. Insead we have o find aprocess F such ha he opimal sopping problem defined by he lef-hand side of (1.3) has he opimal value zero for each T. I is no a all obvious ha here exiss a soluion process F o he fundamenal equaion (3.1), and i is even less obvious ha a soluion will be unique. These quesions will be reaed below. I may seem ha we are only considering he fuures price process from he perspecive of he seller of he conrac. However, he oal cash flows sum o zero, so if he fundamenal pricing equaion above is saisfied, he (spo) value of he conrac is zero also o he buyer (and if exercised in a nonopimal fashion, he value would be posiive for he buyer and negaive for he seller). The main problems o be sudied are he following PROBLEM 3.1. Consider an exogenously given index process X. Our primary problem is o find a process {F ; 0 T} such ha (3.3) is saisfied for all 0, T. If we manage o find a process F wih he above properies, we would also like o find, for each fixed 0, T, he opimal sopping ime ˆ realizing he supremum in sup E Q e r s ds (F X ) r (3.4) s ds df u F. T

ON THE TIMING OPTION IN A FUTURES CONTRACT 273 We also noe ha even if we manage o prove he exisence of a soluion process F, here is no guaranee of he exisence of an opimal sopping ime ˆ, since in he general case we can (as usual) only be sure of he exisence of ɛ-opimal sopping imes. 3.2. Some Preliminary Observaions A complee reamen of he pricing equaion will be given in he nex wo secions, bu wemay already a his sage draw some preliminary conclusions. (3.5) LEMMA 3.1. The fuures price process has o saisfy he condiion F(, T) X, T, (3.6) F(T, T) = X T. Proof. The economic reason for (3.5) is obvious. If, for some,wehavef(, T) > X hen we ener ino a shor posiion (a zero cos) and immediaely decide o deliver. We pay X and receive F(, T) hus making an arbirage profi, and immediaely close he posiion (again a zero cos). A more formal proof is obained by noing ha he fundamenal equaion (3.3) implies ha E Q e r s ds (F X ) r (3.7) s ds df u F 0, for all sopping imes wih T.Inparicular, he (3.7) holds for = which gives us (3.8) E Q F X F 0, and since boh F and X are adaped, he inequaliy (3.5) follows. The boundary condiion (3.6) is an immediae consequence of no arbirage. We finish his secion by proving ha for he very special case of zero shor rae, we can easily obain an explici formula for he fuures price process. Noe ha, for simpliciy of noaion, he symbol inf henceforh denoes he essenial infimum. (3.9) PROPOSITION 3.2. If r 0, hen F(, T) = inf T EQ X F. Proof. Wih zero shor rae he fundamenal equaion reads as (3.10) sup E (F Q X ) df u F = 0. T Using he fac ha df u = F F we hus obain (3.11) sup E Q F X F = 0. T

274 F. BIAGINI AND T. BJÖRK Since F is adaped his implies (3.12) F = sup E Q X F = inf T T EQ X F. In he Secion 4, we will prove ha he formula (3.9) is in fac valid also in he general case wihou he assumpion of zero shor rae. 3.3. The Pricing Equaion in Discree Time Case By going hrough a compleely parallel argumen as above, i is easy o see ha in a discree ime model he fundamenal equaion (3.3) will have he form ( ) ( ) sup E Q 1 n 1 (3.13) (F X ) F n F = 0, T 1 + r n=+1 n 1 + r n=+1 u=+1 u where F n = F n F n 1. 4. DETERMINING THE FUTURES PRICE PROCESS In his secion, we will solve he fundamenal pricing equaions (3.3) and (3.13), hus obaining an explici represenaion for he fuures price process. We sar wih he discree ime case, since his is echnically less complicaed. 4.1. The Discree Time Case We will no analyze equaion (3.13) direcly, bu raher use a sandard dynamic programming argumen as a way of aacking he problem. To do his we consider he decision problem of he holder of he shor end of he fuures conrac. Suppose ha a ime n youhave enered ino he shor conrac. Then you have he following wo alernaives: 1. You can decide o deliver immediaely, in which case you will receive he amoun (4.1) F n X n. 2. You can decide o wai unil n + 1. This implies ha a ime n + 1you will obain he amoun F n F n+1. The arbirage free value, a n,ofhis cash flow is given by he expression (4.2) E Q Fn F n+1 1 + r n+1 F n = 1 1 + r n+1 E Q F n F n+1 F n, where we have used he fac ha r is predicable. The value of your conrac, afer having received he cash flow above, is by definiion zero. Obviously you would like o make he bes possible decision, so he value a ime n of a shor posiion is given by 1 (4.3) max (F n X n ), E Q F n F n+1 F n. 1 + r n+1

ON THE TIMING OPTION IN A FUTURES CONTRACT 275 On he oher hand, he spo price of he fuures conrac is by definiion always equal o zero,soweconclude ha 1 max (F n X n ), E Q (4.4) F n F Fn n+1 = 0, 1 + r n+1 for n = 1,...,T 1. We now recall he following basic resul from opimal sopping heory (see Snell 1952). THEOREM 4.1 (Snell Envelope Theorem). value process V by Wih noaions as above, define he opimal (4.5) V = inf T EQ X F, where varies of he class of sopping imes. Then V is characerized by he propery of being he larges submaringale dominaed by X. The process V aboveisreferred o as he (lower) Snell Envelope of X wih horizon T, and we may now sae and prove our main resul in discree ime. THEOREM 4.2. Given he index process X, and a final delivery dae T, he fuures price process F(, T) exiss uniquely and coincides wih he lower Snell envelope of X wih horizon T, i.e., (4.6) F(, T) = inf T EQ X F, where varies over he se of sopping imes. Furhermore, if he shor posiion is enered a ime, hen he opimal delivery ime is given by (4.7) ˆ() = inf {k ; F k = X k }. Proof. We will show ha here exiss a unique fuures price process F and ha i is in fac he larges submaringale dominaed by X. The resul hen follows direcly from he Snell Envelope Theorem. We sar by noing ha, since by Assumpion 2.11+ r n > 0, we can wrie (4.4) as (4.8) and since F is adaped his implies (4.9) This gives us he recursive sysem (4.10) max(f n X n ), E Q F n F n+1 F n = 0, F n = max X n, E Q F n+1 F n. F n = minx n, E Q F n+1 F n, n = 0,...,T 1 (4.11) F T = X T, where he boundary condiions follows direcly from no arbirage. This recursive formula for F proves exisence and uniqueness. We now goonoprovehaf is a submaringale dominaed by X. From (4.10) we immediaely have F n E Q F n+1 F n,

276 F. BIAGINI AND T. BJÖRK which proves he submaringale propery, and we also have F n X n, which in fac was already proved in Lemma 3.1. I remains o prove he maximaliy propery of F and for his we use backwards inducion. Assume hus ha Z is a submaringale dominaed by X. Inparicular his implies ha Z T X T,bu since F T = X T we obain Z T F T.For he inducion sep, assume ha Z n+1 F n+1.wehen wan o prove ha his implies he inequaliy Z n F n. To do his we observe ha he submaringale propery of Z ogeher wih he inducion assumpion implies Z n E Q Z n+1 F n E Q F n+1 F n. By assumpion we also have Z n X n,sowehaveinfac Z n minx n, E Q F n+1 F n, and from his inequaliy and (4.10) we obain Z n F n. 4.2. The Coninuous Time Case and Some Examples We now go on o find a formula for he fuures price process in coninuous ime and, based on he discree ime resuls of he previous secion, we of course conjecure ha also in coninuous ime we have he formula F(, T) = inf T E Q X F. Happily enough, his also urns ou o be correc, bu a echnical problem is ha in coninuous ime i is impossible o jus mimic he discree ime argumens above, since we can no longer use inducion. Thus, we have o use oher mehods, and we will rely on some very nonrivial resuls from coninuous ime opimal sopping heory. All hese resuls can be found in he highly readable appendix D in Karazas and Shreve (1998). From now on, we assume he following furher inegrabiliy condiion on he underlying process X: (4.12) E Q sup X <. 0 T Before proving our main resul, we need he following echnical resul. LEMMA 4.1. Suppose ha he index process X saisfies condiion (4.12) and consider is Snell Envelope F(, T) = inf T E Q X F. Le H = e 0 rudu,where r saisfies he weak boundness Assumpion 2.1. If Tisasopping ime such ha he sopped submaringale F is a maringale, hen he sochasic inegral s 0 e 0 rudu df s is a maringale. Proof. Werecall ha by Assumpion 2.2, he index process X is supposed o be a nonnegaive adaped càdlàg process. Hence, if is a sopping ime such ha he sopped submaringale F is a maringale, i is indeed a càdlàg maringale and consequenly we need only o verify ha ( T ) 1 Hs 2 df 2 (4.13) s < E Q 0 in order o guaranee ha he sochasic inegral s 0 e 0 rudu df s is a maringale by using he Burkholder Davis Gundy inequaliies (see Revuz and Yor 1994, p. 151, and Proer

ON THE TIMING OPTION IN A FUTURES CONTRACT 277 2004, p. 193). Here, he process F is he quadraic variaion of F (for furher deails, see Proer 2004, p. 66). Since H = e 0 rudu and r is uniformly bounded from below (Assumpion 2.1), we obain he following esimaes ( T ) 1 ( Hs 2 df 2 T ) s = E Q e 2 1 s 0 rudu df 2 s E Q 0 0 ( T E Q 0 ) 1 e 2cs df 2 s e ct E Q ( F ) 1 2 T. Since he process F is given by he Snell Envelope of X,iisanonnegaive submaringale dominaed by X, hence we can use he Burkholder Davis Gundy inequaliies and ge E Q ( F ) 1 2 T ke Q sup F 0 T ke Q sup F ke Q sup X, 0 T 0 T where k is a suiable consan. Since X saisfies (4.12), he las erm of he inequaliy is finie. Hence, we can conclude ha he sochasic inegral s 0 e 0 rudu df s is a maringale. We may now sae our main resul in coninuous ime. THEOREM 4.3. Under Assumpion 2.1 and if (4.12) holds, here exiss, for each fixed T, a unique fuures price process F(, T) solving he he fundamenal equaion (3.3). The fuures price process is given by he expression (4.14) F(, T) = inf T EQ X F. Furhermore, if X has coninuous rajecories, hen he opimal delivery ime ˆ(), forhe holder of a shor posiion a ime is given by (4.15) ˆ() = inf {u ; F(u, T) = X u }. Proof. We firs show ha if we define F by (4.14) hen F solves he pricing equaion (3.3). Having proved his we will hen go on o prove ha if F solves (3.3), hen F mus necessarily have he form (4.14). We hus sar by defining aprocess F as he lower Snell envelope of X, i.e., (4.16) F = inf T EQ X F, and we wan o show ha for his choice of F, he fundamenal pricing equaion (3.3) is saisfied. From he (coninuous ime version of ) Snell Envelope Theorem, we know ha F is a submaringale. Thus (for fixed ) he sochasic differenial r s ds df u, is a submaringale differenial, and since F X we see ha he inequaliy E Q e r s ds (F X ) r (4.17) s ds df u F 0,

278 F. BIAGINI AND T. BJÖRK will hold for every sopping ime wih T. Toshow ha F defined as above saisfies (3.3) i is herefore enough o show ha for some sopping ime we have E Q e r s ds (F X ) r (4.18) s ds df u F = 0. For simpliciy of exposiion we now assume ha, for each, he infimum in he opimal sopping problem (4.19) inf T EQ X F, is realized by some (no necessarily unique) sopping ime ˆ. The proof of he general case is more complicaed and herefore relegaed o he Appendix. From general heory (see Karazas and Shreve 1998, p. 355, heorem D9) we cie he following facs. 1. Wih F defined by (4.16) we have (4.20) Fˆ = Xˆ. 2. The sopped process F ˆ defined by F ˆ (4.21) s = F s ˆ, where denoes he minimum, is a maringale on he inerval, T. Choosing = ˆ, equaion (4.18) hus reduces o he equaion (4.22) which we can wrie as (4.23) E Q ˆ E Q T r s ds df u F = 0, r s ds df ˆ u F = 0, and since, by a small variaion of Lemma 4.1, he process s r v dv df ˆ u is a maringale, (4.23) is indeed saisfied. This proves exisence. In order o prove uniqueness le us assume ha, for a fixed T,aprocess F solves (3.3). We now wan o prove ha F is in fac he lower Snell envelope of X, i.e., we have o prove ha F is he larges submaringale dominaed by X. We firs noe ha, afer premuliplicaion wih he exponenial facor e 0 rsds, he fundamenal equaion (3.3) can be rewrien as 0 rsds df u = inf 0 T EQ 0 rsds df u + e (4.24) 0 rsds (X F ) 0 Defining he process V by (4.25) V = 0 0 r sds df u, we hus see ha V is he lower Snell envelope of he process Z, defined by (4.26) i.e., Z = 0 0 r sds df u + e 0 r sds (X F ), F.

ON THE TIMING OPTION IN A FUTURES CONTRACT 279 (4.27) V = inf T EQ Z F. From he Snell Theorem i now follows ha V is a submaringale, and since he exponenial inegrand in (4.25) is posiive, his implies ha also F is a submaringale. We have already proved in Proposiion 3.1 ha F X so i only remains o prove maximaliy. To his end, le us assume ha G is a submaringale dominaed by X.Wenowwan o prove ha G F for every T.Tohis end we choose a fixed bu arbirary.for simpliciy of exposiion we now assume ha, for a fixed, here exiss and opimal sopping ime aaining he infimum in (4.27), and he denoe his sopping ime by. The proof in he general case is found in he Appendix. We obain from (3.3) E Q e r s ds (F X ) r (4.28) s ds df u F = 0. Since F X and F is a submaringale, his implies ha (4.29) and ha (4.30) E Q F = X, r s ds df u F = 0, which in urn (afer premuliplicaion by an exponenial facor) implies ha (4.31) E Q V F = V. Since V is a submaringale, his implies ha he sopped process V is in fac a maringale on he ime inerval, T, which in urn implies ha he sopped process F is a maringale on, T. In paricular we hen have (4.32) F = E Q F F = E Q X F, where we have used (4.29). On he oher hand, from he assumpions on G we have (4.33) G E Q G F E Q X F = F, which proves he maximaliy of F. The second saemen in he heorem formulaion follows direcly from heorem D.12 in Karazas and Shreve (1998). As a more or less rivial consequence, we immediaely have he following resul for fuures on underlying sub- and supermaringales. PROPOSITION 4.1. (4.34) IfXis a submaringale under Q, hen F(, T) = X, and i is always opimal o deliver a once, i.e., (4.35) If X is a supermaringale under Q, hen (4.36) and i is always opimal o wai, i.e., (4.37) ˆ() =. F(, T) = E Q X T F, ˆ() = T.

280 F. BIAGINI AND T. BJÖRK Proof. Follows a once from he represenaion (4.14). From his resul we immediaely have some simple financial implicaions. PROPOSITION 4.2. Assume ha one of he following condiions hold 1. X is he price process of a raded financial asse wihou dividends, and he shor rae process r is nonnegaive wih probabiliy one. 2. X is he price process of a raded asse wih a coninuous dividend yield rae process δ such ha δ r for all wih probabiliy one. 3. X is an exchange rae process (quoed as unis of domesic currency per uni of foreign currency) and he foreign shor rae r f has he propery ha r f r for all wih probabiliy one. Then he fuures price is given by (4.38) F(, T) = X, and i is always opimal o deliver a once, i.e., (4.39) Proof. ˆ() =. The Q dynamics of X are as follows in he hree cases above dx = r X d + dm, dx = X r δ d + dm, dx = X r r f d + dm, where M is he generic noaion for a maringale. The assumpions guaranee, in each case, ha X is a Q-submaringale and we may hus apply Proposiion 4.1. Wih an almos idenical proof we have he following parallel resul, which shows ha under cerain condiions he fuures price process is no changed by he inroducion of a iming opion. PROPOSITION 4.3. Assume ha one of he following condiions hold 1. X is he price process of an asse wih a convenience yield rae process γ such ha γ r for all wih probabiliy one. 2. X is an exchange rae process (quoed as unis of domesic currency per uni of foreign currency) and he foreign shor rae r f has he propery ha r f r for all wih probabiliy one. Then he fuures price is given by (4.40) F(, T) = E Q X T F, and i is always opimal o wai unil T o deliver, i.e., (4.41) ˆ() = T.

ON THE TIMING OPTION IN A FUTURES CONTRACT 281 5. CONCLUSIONS AND DISCUSSION The main resul of he presen paper is given in Theorems 4.2 and 4.3 where we provide he formula (5.1) F(, T) = inf T EQ X F which gives us he arbirage free fuures price process in erms of he underlying index X and he maringale measure Q. InSecion 4.2, we also gave some immediae implicaions of he general formula, bu hese resuls are of secondary imporance. We now have a number of commens on he main resul (5.1). We see ha he formula (5.1) for he fuures price in he presence of a iming opion looks very much like he sandard pricing formula for an American opion. Therefore, one may perhaps expec ha (5.1) is a direc consequence of he well known pricing formula for American conracs. As far as we can undersand, his is no he case. As noed above, he fuures price process F(, T) isno a price process a all, since is economic role is ha of a cumulaive dividend process for he fuures conrac (which always has spo price zero). From a more echnical poin of view, we also see ha he deerminaion of he F process is quie inricae, since F has o solve he infinie dimensional fundamenal equaion (3.3) (which is in fac an equilibrium condiion for each )orhe corresponding discree ime equaion (4.4). We assumed absence of arbirage bu we did no make any assumpions concerning marke compleeness. In an incomplee marke, he maringale measure Q is no unique, so in his case formula (5.1) does no provide us wih a unique arbirage free fuures price process. In an incomplee seing, he inerpreaion of Theorems 4.2 and 4.3 is hen ha, given absence of arbirage, he fuures price process has o be given by formula (5.1) for some choice of a maringale measure Q. This is of course compleely parallel o he sandard risk neural pricing formula which, in he incomplee seing, gives us a price of a coningen claim which depends upon he maringale measure chosen. Noe however, ha some of he resuls above are independen of he choice of he maringale measure. In paricular, his is rue for Proposiion 4.2. APPENDIX: A PROOF OF THEOREM 4.3 IN THE GENERAL CASE In his Appendix, we provide he proof of Theorem 4.3 for he general case, i.e., wihou assuming ha he infima in (4.19) and (4.27) are aained. We sar wih he exisence proof and o his end we define he process F (as before) by (A.1) F = inf T EQ X F, and we have o show ha F hus defined saisfies he fundamenal pricing equaion sup E Q e r s ds (F X ) r (A.2) s ds df u F = 0. T As in he simplified proof above i is easy o see ha (A.3) E Q e r s ds (F X ) r s ds df u F 0,

282 F. BIAGINI AND T. BJÖRK for all sopping imes wih T. Thus, o prove ha F saisfies (A.2) i is enough o prove ha here exiss a sequence of sopping imes { n } n=1 such ha E Q e n n r s ds (F n X n ) r (A.4) s ds df u F 1 n, for all n.todohis we consider a fixed and define n by (A.5) n = inf {s ; F s X s (1 1/n)}. Like in he earlier proof we can rewrie he sochasic inegral in (A.4) as n r s ds df u = T r s ds df n u, and i can be shown (see Karazas and Shreve 1998) ha he sopped process F n is a maringale. Thus, by Lemma 4.1 we ge ha he sochasic differenial r s ds df n u is a maringale differenial, and we obain E Q e n n r s ds (F n X n ) r s ds df u F = E Q e n r s ds (F n X n ) F. From he definiion of n we hen have E Q e n r s ds (F n X n ) F E Q e n ( r s ds 1 1 ) X n X n F n = 1 n EQ e n r s ds X n F. Furhermore, we have E Q e n r s ds X n F e c(t ) E Q 1 X n F 1 1/n EQ F n F = e c(t ) F 1 1/n, where we again have used he maringale propery of he sopped process F n.wehus have E Q e n n r s ds (F n X n ) r s ds df u F e c(t ) F n(1 1/n) which ends o zero as n. We now urn o he uniquenesss proof and for his we consider again a fixed and define for each n he sopping ime n by { n = inf s ; V s Z s 1 } (A.6). n Since V n is a maringale on, T and since V is given by (4.25) i now follows ha F n is a maringale on he same inerval. By definiion of n i follows ha e n r (A.7) s ds (X n F n ) 1/n,

ON THE TIMING OPTION IN A FUTURES CONTRACT 283 so we have X n (A.8) F n + 1 n B n. B Now assume ha G is a submaringale dominaed by X. Wehen obain (A.9) G E Q G n F E Q X n F (A.10) (A.11) E Q F n F + 1 n Bn EQ F = F + 1 n EQ Bn B F B F + 1 n C B where C is given by (2.6). Leing n gives us G F and we are done. REFERENCES BOYLE, P.(1989): The Qualiy Opion and Timing Opion in Fuures Conracs, J. Finance 44, 101 113. COHEN, H.(1995): Isolaing he Wild Card Opion, Mah. Finance 5, 155 165. GAY, G., and S. MANASTER (1984): The Qualiy Opion Implici in Fuures Conracs, J. Financial Econ. 13, 353 370. GAY, G., and S. MANASTER (1986): Implici Delivery Opions and Opimal Delivery Sraegies for Financial Fuures Conracs, J. Financial Econ. 16, 41 72. KARATZAS, I., and S. SHREVE (1998): Mehods of Mahemaical Finance. New York: Springer. PROTTER, P. E.(2004): Sochasic Inegraion and Differenial Equaions, 2nd ediion. Berlin: Springer. REVUZ, D.,and M. YOR (1994): Coninuous Maringales and Brownian Moion. Berlin: Springer. SNELL, J.(1952): Applicaion of Maringale Sysem Theorems, Trans. Am. Mah. Soc. 73, 293 312.