J. Michael Harrison

Martingales and stochastic integrals in the theory of continuous trading

This paper develops a general stochastic model of a frictionless security market with continuous trading. The vector price process is given by a semimartingale of a certain class, and the general stochastic integral is used to represent capital gains. Within the framework of this model, we discuss the modern theory of contingent claim valuation, including the celebrated option pricing formula of Black and Scholes. It is shown that the security market is complete if and only if its vector price process has a certain martingale representation property. A multidimensional generalization of the Black-Scholes model is examined in some detail, and some other examples are discussed briefly.

Speculative Investor Behavior in a Stock Market with Heterogeneous Expectations

I. Introduction, 323.—II. Formulation, 325.—III. An example, 326.—IV. Consistent price schemes, 328.—V. Back to the example, 332.—VI. Relaxing the assumptions, 333.—VII. Concluding remarks, 334.

A stochastic calculus model of continuous trading: Complete markets

A paper by the same authors in the 1981 volume of Stochastic Processes and Their Applications presented a general model, based on martingales and stochastic integrals, for the economic problem of investing in a portfolio of securities. In particular, and using the terminology developed therein, that paper stated that every integrable contingent claim is attainable (i.e., the model is complete) if and only if every martingale can be represented as a stochastic integral with respect to the discounted price process. This paper provides a detailed proof of that result as well as the following: The model is complete if and only if there exists a unique martingale measure.

Brownian Models of Queueing Networks with Heterogeneous Customer Populations

Consider an open queueing network with I single-server stations and K customer classes. Each customer class requires service at a specified station, and customers change class after service in a Markovian fashion. (With K allowed to be arbitrary, this routing structure is almost perfectly general.) There is a renewal input process and general service time distribution for each class. The correspondence between customer classes and service stations is in general many to one, and the service discipline (or scheduling protocol) at each station is left as a matter for dynamic decision making.

Instantaneous Control of Brownian Motion

A controller continuously monitors a storage system, such as an inventory or bank account, whose content Z = { Z t , t (ge)0} fluctuates as a ((mu), (sigma) 2 ) Brownian motion in the absence of control. Holding costs are incurred continuously at rate h ( Z t ). At any time, the controller may instantaneously increase the content of the system, incurring a proportional cost of r times the size of the increase, or decrease the content at a cost of l times the size of the decrease. We consider the case where h is convex on a finite interval [(alpha), (beta)] and h = (infinity) outside this interval. The objective is to minimize the expected discounted sum of holding costs and control costs over an infinite planning horizon.It is shown that there exists an optimal control limit policy, characterized by two parameters a and b ((alpha) (le) a b (le) (beta)). Roughly speaking, this policy exerts the minimum amounts of control sufficient to keep Z t (in) [ a , b ] for all t (ge) 0. Put another way, the optimal control limit policy imposes on Z a lower reflecting barrier at a and an upper reflecting barrier at b . The optimality of a particular control limit policy is proved directly, with heavy reliance on the change of variable formula for semimartingales. We do not give a full-blown algorithm for construction of the optimal control limits, but a computational scheme could easily be developed from our constructive proof of existence.

Impulse Control of Brownian Motion

Consider a storage system, such as an inventory or cash fund, whose content fluctuates as a ((mu), (sigma) 2 ) Brownian motion in the absence of control. Holding costs are continuously incurred at a rate proportional to the storage level and we may cause the storage level to jump by any desired amount at any time except that the content must be kept nonnegative. Both positive and negative jumps entail fixed plus proportional costs, and our objective is to minimize expected discounted costs over an infinite planning horizon. A control band policy is one that enforces an upward jump to q whenever level zero is hit, and enforces a downward jump to Q whenever level S is hit (0 q Q S ). We prove the existence of an optimal control band policy and calculate explicitly the optimal values of the critical numbers ( q , Q , S ).

Heavy traffic resource pooling in parallel-server systems

We consider a queueing system with r non‐identical servers working in parallel, exogenous arrivals into m different job classes, and linear holding costs for each class. Each arrival requires a single service, which may be provided by any of several different servers in our general formulation; the service time distribution depends on both the job class being processed and the server selected. The system manager seeks to minimize holding costs by dynamically scheduling waiting jobs onto available servers. A linear program involving only first‐moment data (average arrival rates and mean service times) is used to define heavy traffic for a system of this form, and also to articulate a condition of overlapping server capabilities which leads to resource pooling in the heavy traffic limit. Assuming that the latter condition holds, we rescale time and state space in standard fashion, then identify a Brownian control problem that is the formal heavy traffic limit of our rescaled scheduling problem. Because of the assumed overlap in server capabilities, the limiting Brownian control problem is effectively one‐dimensional, and it admits a pathwise optimal solution. That is, in the limiting Brownian control problem the multiple servers of our original model merge to form a single pool of service capacity, and there exists a dynamic control policy which minimizes cumulative cost incurred up to any time t with probability one. Interpreted in our original problem context, the Brownian solution suggests the following: virtually all backlogged work should be held in one particular job class, and all servers can and should be productively employed except when the total backlog is small. It is conjectured that such ideal system behavior can be approached using a family of relatively simple scheduling policies related to the cμ rule.

Dynamic Control of a Queue with Adjustable Service Rate

We consider a single-server queue with Poisson arrivals, where holding costs are continuously incurred as a nondecreasing function of the queue length. The queue length evolves as a birth-and-death process with constant arrival rate ? = 1 and with state-dependent service rates µ nthat can be chosen from a fixed subset A of [0, 8). Finally, there is a nondecreasing cost-of-effort functionc(·) on A, and service costs are incurred at ratec(µ n ) when the queue length isn. The objective is to minimize average cost per time unit over an infinite planning horizon. The standard optimality equation of average-cost dynamic programming allows one to write out the optimal service rates in terms of the minimum achievable average costz*. Here we present a method for computingz* that is so fast and so transparent it may be reasonably described as an explicit solution for the problem of service rate control. The optimal service rates are nondecreasing as a function of queue length and are bounded if the holding cost function is bounded. From a managerial standpoint it is natural to comparez*, the minimum average cost achievable with state-dependent service rates, against the minimum average cost achievable with a single fixed service rate. The difference between those two minima represents the economic value of a responsive service mechanism, and numerical examples are presented that show it can be substantial.

Brownian models of multiclass queueing networks: Current status and open problems

This paper is concerned with Brownian system models that arise as heavy traffic approximations for open queueing networks. The focus is on model formulation, and more specifically, on the formulation of Brownian models for networks with complex routing. We survey the current state of knowledge in this dynamic area of research, including important open problems. Brownian approximations culminate in estimates of complete distributions; we present numerical examples for which complete sojourn time distributions are estimated, and those estimates are compared against simulation.

Multi-Resource Investment Strategies: Operational Hedging Under Demand Uncertainty

Consider a firm that markets multiple products, each manufactured using several resources representing various types of capital and labor, and a linear production technology. The firm faces uncertain product demand and has the option to dynamically readjust its resource investment levels, thereby changing the capacities of its linear manufacturing process. The cost to adjust a resource level either up or down is assumed to be linear. The model developed here explicitly incorporates both capacity investment decisions and production decisions, and is general enough to include reversible and irreversible investment. The product demand vectors for successive periods are assumed to be independent and identically distributed. The optimal investment strategy is determined with a multi-dimensional newsvendor model using demand distributions, a technology matrix, prices (product contribution margins), and marginal investment costs. Our analysis highlights an important conceptual distinction between deterministic and stochastic environments: the optimal investment strategy in our stochastic model typically involves some degree of capacity imbalance which can never be optimal when demand is known.