Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

A Model of Supply of Storage

Author(s): Eugenio S. A. Bobenrieth H., Juan R. A. Bobenrieth H. and Brian D. Wright


Source: Economic Development and Cultural Change, Vol. 52, No. 3 (April 2004), pp. 605-616
Published by: University of Chicago Press
Stable URL: http://www.jstor.org/stable/10.1086/383331
Accessed: 09-02-2016 18:10 UTC

Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at http://www.jstor.org/page/
info/about/policies/terms.jsp

JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of content
in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms of scholarship.
For more information about JSTOR, please contact support@jstor.org.

University of Chicago Press is collaborating with JSTOR to digitize, preserve and extend access to Economic Development and
Cultural Change.

http://www.jstor.org

This content downloaded from 130.240.43.43 on Tue, 09 Feb 2016 18:10:17 UTC
All use subject to JSTOR Terms and Conditions
A Model of Supply of Storage

Eugenio S. A. Bobenrieth H.
University of Concepción

Juan R. A. Bobenrieth H.
University of Bı́o-Bı́o

Brian D. Wright
University of California, Berkeley, and the Giannini Foundation

I. Introduction
The relation between the returns to holding stocks of commodities and the
amount of stocks held has long been a challenge to economists. In the standard
storage model in the tradition of Gustafson (1958) and Johnson and Sumner
(1976) with positive marginal cost of storage and zero marketing cost, the
expected financial return to holding stocks from one period to the next is
always positive when stocks are positive. But an empirical regularity of mar-
kets for storable commodities is that the expected return to holding stocks
appears to be negative when stocks are low. Given the results of recent ad-
vances in empirical tests of this model, the need for a richer model of the
relation between stocks and returns—the “supply of storage”—has become
obvious.
In this article we include, in a model similar to that of Scheinkman and
Schechtman (1983), a marketing cost function consistent with some of the
informal arguments in the literature that relates convenience yield to the cost
of sales. Our model formalizes intuition about the role of marketing that dates
back to the time of Kaldor (1939) and Working (1934, 1948, 1949).
In this model, though additions to stocks facilitate merchandising in the
spirit of Kaldor’s original notion, there is no extramarket dividend to the
holders of stocks as assumed in a number of studies of returns to commodity-
linked assets.1 In a simple setting, with independently and identically distrib-
uted (i.i.d.) production, we show that the marginal social value of stocks is
equal to the price net of marginal marketing costs at all times. Some stocks
are held even when prices are very high, and expected to fall, but such stocks

䉷 2004 by The University of Chicago. All rights reserved.


0013-0079/2004/5203-0006$10.00

This content downloaded from 130.240.43.43 on Tue, 09 Feb 2016 18:10:17 UTC
All use subject to JSTOR Terms and Conditions
606 Economic Development and Cultural Change

Fig. 1

yield no special unobserved dividend. They are held because they are so
expensive to sell, not because they are so profitable to keep. The response of
price spreads to changes in stocks is qualitatively consistent with observed
supply of storage behavior.
In the next section, we provide some background on the issue of supply
of storage and convenience yield. Then we discuss our representation of the
cost of marketing in Section III. The model is presented in Section IV, which
also contains the result that the marginal social value of stocks is less than
the market price. In Section V we derive restrictions on the supply of storage
consistent with observed market behavior, including the occurrence of “back-
wardation.” Conclusions follow in Section VI.

II. Supply of Storage and Convenience Yield


The consensus of numerous empirical studies is that, for a wide array of
commodities (including grains, petroleum, timber, and minerals), the dif-
ference between the nearest forward or futures price and the current (“spot”)
price falls and becomes negative—that is, goes into “backwardation,” in the
British terminology (Keynes 1930, p. 143), or, in American usage, “inverse
carry”—as aggregate stocks fall toward their bound of zero.2 In the typical
supply of storage relation illustrated in figure 1 (redrawn from Working

This content downloaded from 130.240.43.43 on Tue, 09 Feb 2016 18:10:17 UTC
All use subject to JSTOR Terms and Conditions
E. Bobenrieth H., J. Bobenrieth H., and B. Wright 607

[1934]), when supplies held over in storage to be available next period


(“carryout stocks”) are low, observed returns do not cover observed costs,
including interest and physical storage costs. Thus, stocks appear to be
carried at a loss.
Following Kaldor (1939), Working (1948, 1949) and subsequent writers
reconciled theory and observation by positing that actual holders of stocks
earn an implicit convenience yield that does not accrue to holders of long
positions in forward or futures markets, and this yield was defined as the
difference between the current cost of holding stocks (including interest and
storage costs) and the current expected rate of change of the spot price. As
Leuthold, Junkus, and Cordier (1989, p. 47) put it, “the smaller the level
of stocks, the greater is the convenience yield from holding an additional
unit. The total convenience yield of stocks increases with quantities held
until marginal convenience yield becomes zero at some large level of
stocks.”
The interpretation of convenience yield as a negative storage cost anal-
ogous to a flow of dividends on financial assets has been used in studies
of the prices of commodity-linked assets, including forward and futures
contracts, commodity bonds, and natural resource deposits. In these works,
the convenience yield is an extra return added to apparent financial returns
in determining asset evaluations. This return is rationalized variously as the
return in customer goodwill or in profits for meeting unexpected orders, for
avoidance of costly short-run market purchases, and for other advantages
plausibly related to the notion of convenience of operation of a business.
However, these purported advantages with storage have not yet been sat-
isfactorily modeled.3
As noted by Scheinkman and Schechtman (1983, p. 429), in the standard
storage model, “backwardation could appear, however, if we assumed a neg-
ative marginal cost of storage at least for some levels of storage. Our opinion
is that if such benefits from storage are present they should be modeled from
more basic principles.” Holbrook Working himself had reservations about
appealing to negative storage costs to rationalize convenience yield:

There remains, however, the question whether it is good theory to treat these
negative values as negative prices. . . . Other possible treatments come to mind,
but none which seem to me to have merits which warrant advancing it as
preferable to recognition of the existence of negative prices of storage. . . .
The main reason . . . for adopting cost-of-storage theory, or some alternative
which provides direct explanation of inter-temporal price relations, is that some
such explanation is necessary to account for observed price behavior. (Working
1949, pp. 1260–62)

But negative marginal storage costs are not necessary to rationalize the
empirical supply of storage relation. Wright and Williams (1989) established
that the essential features of the supply of storage relation can be generated
as an aggregation phenomenon in which price of a good with zero stocks is

This content downloaded from 130.240.43.43 on Tue, 09 Feb 2016 18:10:17 UTC
All use subject to JSTOR Terms and Conditions
608 Economic Development and Cultural Change

associated with positive quantities of similar goods differentiated by location


or some other characteristics, even though the correct shadow prices of the
latter display no backwardation. The empirical relevance of this possibility
has been supported by Benirshka and Binkley (1995) and Brennan, Williams,
and Wright (1997).4 In this article we show more formally that the qualitative
features of the typical supply of storage relation are found in a microeconomic
model in which marketing costs are a decreasing function of stocks and an
increasing function of sales.

III. The Cost of Marketing


Previous models of commodity markets have, in general, assumed that the
relation between the spot-future price spread and stocks has been driven by
the marginal cost of storage. Thus negative spreads imply that this cost be-
comes negative at low levels of stocks.
In this article we draw a distinction between the cost of storing a com-
modity and the cost of merchandising or bringing it to market, which we
denote by the shorthand term “marketing cost.”5 The latter is modeled as a
decreasing function of carryout stocks, given initial available supplies. The
marginal cost of marketing rises steeply as carryout stocks decline toward
what Working calls “minimum working stocks,” below which a reduction
“seems very difficult” (1948, p. 22).
The specific details of marketing infrastructures are highly complex,
commodity specific, and differentiated by spatial and qualitative character-
istics. We choose a simple stylized model of marketing costs that captures
the essentials of the relation between stocks, sales, and price spreads. The
representative atomistic producer-storer must pass its output through a storage
and merchandising facility on the way to market. We recognize an asymmetry
in the cost of placing the commodity into storage, and the cost of withdrawing
it for sale to the consumer, by assuming that dumping the harvest in the
storage facility has zero cost but that the cost of removing it is a decreasing
function of the amount that remains in the facility.
Given available supply z and stocks x, sales are c { z ⫺ x, and total
marketing cost of sales is G(z, x) p ∫ x h(u)du, with h : ⺢⫹ r ⺢⫹ continuous,
z

convex, and strictly decreasing, lim ur0 h(u) p ⫹⬁ , and lim zr⬁ ∫ 0 h(u)du ! ⫹
z

⬁. As figure 2 shows, for a given x, marketing cost is a strictly increasing


concave function of available supplies, and, for a given z, marketing cost is
a strictly decreasing convex function of carryout stocks.6
To fix ideas about this stylization, think of the marketing infrastructure
of the representative firm as an underground storage facility into which all
harvests are dumped and from which sales are withdrawn. Input into the tank
is costless, but the marginal cost of withdrawals from the tank is positive and
depends on the distance from the surface of the contents to the top of the
tank. When the facility is nearly full, withdrawal requires little effort, but
when the tank is less full, withdrawals become more costly. When it is almost
empty, the cost of further “clean-out” might be very high indeed.7

This content downloaded from 130.240.43.43 on Tue, 09 Feb 2016 18:10:17 UTC
All use subject to JSTOR Terms and Conditions
Fig. 2

609

This content downloaded from 130.240.43.43 on Tue, 09 Feb 2016 18:10:17 UTC
All use subject to JSTOR Terms and Conditions
610 Economic Development and Cultural Change

If one thinks of the marketing cost as the cost of transportation to market,


given a certain distribution of storage facilities, the above example is a stylized
analogue to the empirically based numerical model of Brennan et al. (1997).
They show, in a spatial-temporal framework, that as near-term wheat market
deliveries are increased in response to a temporary relative increase in current
price, stocks held close to the market are depleted, transport shadow prices
increase, and thus the marginal cost of merchandising increases. Current stor-
age remains positive (and may even increase) at distant locations as back-
wardation becomes more severe.
Our marketing cost function implies increasing marginal cost of trans-
formation with respect to a characteristic of a marginal unit (e.g., distance or
impurity), which itself is decreasing in available supply. The examples we
have used to motivate our stylized marketing cost function embody the notion
that lies at the heart of discussions of convenience yield: the presence of
existing stocks reduces the marginal cost of marketing. Our emphasis on a
relation between stocks and the cost of marketing is anticipated in many
rationalizations of the observed nature of the supply of storage relation. Telser
(1958, p. 235), for example, has argued: “The availability of stocks permits
a processor or producer to maintain a given level of output at a lower cost
than would be required without stocks.” Earlier, Working (1948, 1949) made
much the same point by observing that storage is an adjunct to firms’ main
processing or merchandising business and may save them enough there to
justify the loss from backwardation.

IV. The Model


As in previous studies of commodity storage, we model a competitive market
for a single storable consumption commodity. Time is discrete. All agents
have rational expectations.
Production is one common exogenous i.i.d. multiplicative disturbance
with density function that is continuous when restricted to its support K {
[0, m],¯ 0 ! m¯ ! ⫹⬁.
Assume that there is a continuum of identical storers, and a continuum
of identical consumers, both of total measure one. The storer can store output
from one period to the next. The amount stored is x ≥ 0 . Storage cost is given
by a continuously differentiable function f : ⺢⫹ r ⺢⫹, with f(0) p 0,
f (0) ≥ 0, f (x) 1 0, and f (x) ≥ 0 for all x 1 0. Given storage x, the next
period’s total available supply is z  { x ⫹ q , where q  is the next period’s
production. Storers are risk neutral and have a constant discount factor d,
0 ! d ! 1.
The utility function of the representative consumer U: ⺢⫹ r ⺢⫹ is con-
tinuous, once continuously differentiable, strictly increasing and strictly con-
cave. It satisfies U(0) p 0, U (0) p ⫹⬁, and lim cr⬁ U(c) p B ! ⫹⬁. The
inverse demand curve for the consumer is then f p U . Note that U(c) p
∫ 0 f (y)dy and that total revenue cf (c) is bounded (by B).
c

This content downloaded from 130.240.43.43 on Tue, 09 Feb 2016 18:10:17 UTC
All use subject to JSTOR Terms and Conditions
E. Bobenrieth H., J. Bobenrieth H., and B. Wright 611

The perfectly competitive market yields the same solution as the surplus
maximization problem. The Bellman equation for the surplus problem is


z

n(z) p max x{U(z ⫺ x) ⫺ f(x) ⫺ h(u)du ⫹ dE[n(z  )]},


x

subject to z  p x ⫹ q , x ≥ 0, z ⫺ x ≥ 0,

where E[.] denotes the expectation with respect to q .


The boundedness of U and G, and the convexity of f, imply that there
exists z ∗ ! ⫹⬁ such that, if x(z) is an argmax of the above Bellman equation,
then the next period’s available supply, x(z) ⫹ q , satisfies x(z) ⫹ q  ≤ z ∗.
Thus, without loss of generality, we set z ≤ z ∗. For (z, x) such that 0 ! x !
z ! z ∗, we impose the following conditions to ensure monotonicity and con-
cavity of the solution n and to ensure existence of a single valued optimal
policy function:

⫺U (z ⫺ x)h (z) ⫹ h (x)U (z ⫺ x) ⫺ h (x)h (z) 1 0,


f (z) ⫺ h(z) ≥ 0,
f (z) ⫺ h (z) ≤ 0,
f (z) 1 0.

Standard results apply, establishing that n is continuous and strictly in-


creasing and that the optimal policy function x(z) is continuous and strictly
increasing. Price is given by the function p(z) p f [z ⫺ x(z)], which is strictly
decreasing with z.
Our assumptions on the functions h and U imply that, if z 1 0, then
x(z) 1 0.
The policy function x satisfies the Euler condition

f [z ⫺ x(z)] ⫺h [x(z)] ⫹f  [x(z)] pdE {n  [x(z) ⫹ q ]} , for z 1 0,

and the envelope condition n (z) p f [z ⫺ x(z)] ⫺ h(z).


To address the convenience yield puzzle, we ask the following question:
what is the marginal value of an addition to available stocks at any time
t ≥ 0? The following result (based on theorem 1 of Benveniste and Scheink-
man [1979, p. 729]) provides the answer. In the absence of market distortions,
the existence of marketing costs implies that the shadow price of the marginal
unit of supply is the market price, net of its marketing cost. In this model
there is no extra return to holders of stocks. For z 1 0, we have:

dn
(z) p p(z) ⫺ h(z).
dz

This content downloaded from 130.240.43.43 on Tue, 09 Feb 2016 18:10:17 UTC
All use subject to JSTOR Terms and Conditions
612 Economic Development and Cultural Change

V. Backwardation and the Supply of Storage


Given current stocks x(z) 1 0, the Euler equation implies that the discounted
expected price change dE {p [x(z) ⫹ q  ]} ⫺ p(z) is equal to:
f  [x(z)] ⫺h [x(z)] ⫹dE {h [x(z) ⫹ q ]} .
Hence we can express the discounted expected price change as the spread
function S:
S(x) { f (x) ⫺ h(x) ⫹ dE[h(x ⫹ q  )].
If we assume that expected prices in our model with risk-neutral speculators
and fixed interest rate equal futures prices, we can now address one aspect
of the phenomenon, noted by Keynes (1930), namely, that the price spreads
implicit in forward or futures quotations do not usually cover the full carrying
charges for holding stocks of a commodity, including interest and storage
costs. It is now well known (see, e.g., Bresnahan and Spiller 1986; Williams
and Wright 1991) that the spread between far and near futures prices will not
cover carrying charges in the standard rational expectations storage model.
For our model, the fact that h is strictly decreasing implies, in addition, that
the spread between the present value of the price expected next period and
the spot (current) price will not cover full carrying charges, that is, S(x) !
f (x).
We derive strong restrictions on S , consistent with empirical observations
of the supply of storage initiated by Working (1934, 1948, 1949):
Theorem.
a) S(x) is strictly increasing.
b) Let x¯ p x ⫹ D, D 1 0. Define S¯ { S(x), ¯ S { S(x). Then:

S̄ ⫺ S
lim S p ⫺⬁, lim p ⫹⬁.
xr0 xr0 x̄ ⫺ x
Proof.
a) The result follows from the facts that h is strictly decreasing and
convex and that f  is increasing.
b) lim xr0 S p lim xr0 {f (x) ⫺ h(x) ⫹ dE [h(x ⫹ q  )]} p ⫺⬁, because
f (0) 苸 ⺢, h(0) p ⫹⬁ and E[h(q  )] ! ⫹⬁. It immediately follows that:


S̄ ⫺ S
lim p ⫹⬁.
xr0 x̄ ⫺ x
Q.E.D.
Part b of the above theorem refers to the slopes of arcs of the function
S(x). We can prove an analogous result about the derivative as x approaches
zero if we add the assumptions f (x) ≤ 0, h (x) ≤ 0.
Furthermore, there exists a unique invariant distribution of availability
z, which is a global attractor, by the proof in Scheinkman and Schechtman

This content downloaded from 130.240.43.43 on Tue, 09 Feb 2016 18:10:17 UTC
All use subject to JSTOR Terms and Conditions
E. Bobenrieth H., J. Bobenrieth H., and B. Wright 613

(1983, theorem 4). This in turn implies invariant distributions for price, stor-
age, and consumption.

VI. Conclusions
If the cost of marketing is a function of available supplies and carryout stocks,
some long-standing puzzles about commodity market behavior can be re-
solved. Consumption and stocks can be continuously positive, and the spread
between the expected price and current price can be at less than full carry.
Stocks are held at less than full carry, not because of convenience yield as
an extra implicit dividend in the form of a “negative storage cost” to be added
to market returns but because the stocks’ current marginal private (and social)
value is below market price and expected to rise. Restrictions implied by the
model are consistent with backwardation and an upward-sloped supply of
storage curve.
Thus our model replicates the qualitative relationship between price
changes and stocks that gave rise to the concept of convenience yield, but
our interpretation is quite different. Stocks are held in backwardation not
because their yield is higher than of the marginal unit consumed but because
their current shadow values are sufficiently low, due to marketing costs, to
make their expected rates of increase exceed the cost of capital.
These results are foreshadowed in some previous, less formal discussions
relating convenience yield to marketing costs and in the results of recent
spatial-temporal programming models of grain markets. The model furnishes
a microeconomic foundation for the inclusion of a “supply of storage” relation
in econometric models of markets with storage, which Miranda and Rui (1999)
have shown to be an effective means of meeting the challenge posed by Deaton
and Laroque (1992, 1996) to match the price autocorrelation exhibited in
commodity market data.
The admissible set of specifications of the function h includes cases that
produce very diverse types of price behavior. A sufficiently nonlinear spec-
ification of h, with f nearly linear, can produce a supply of storage relation
that is virtually flat for all values of carryover above an arbitrarily small
positive threshold. In this case, price behavior will be similar to that realized
in an analogous model with constant h and to occasional stockouts during
which the market can be in backwardation. Under a less nonlinear specification
of h, backwardations may be less severe but are a much more frequent feature
of the stochastic behavior of prices.
It might well be feasible to construct alternative approaches to the der-
ivation of supply of storage behavior. With respect to the approach pursued
in this article, an obvious project for the future is to explore further the
empirical relations between marketing costs (including transportation costs),
storage costs, available supply, and carryout stocks identified in Brennan et
al. (1997). Unfortunately, data of the necessary detail might be difficult to
find for other markets.

This content downloaded from 130.240.43.43 on Tue, 09 Feb 2016 18:10:17 UTC
All use subject to JSTOR Terms and Conditions
614 Economic Development and Cultural Change

References
Benirshka, M., and J. Binkley. 1995. “Optimal Storage and Marketing over Space and
Time.” American Journal of Agricultural Economics 77:512–24.
Benveniste, L. M., and J. A. Scheinkman. 1979. “On the Differentiability of the Value
Function in Dynamic Models of Economics.” Econometrica 47:727–32.
Bougheas, S. 1994. “Asset and Currency Prices in an Exchange Economy with Trans-
action Costs.” Journal of Macroeconomics 16:99–107.
Brennan, D., J. C. Williams, and B. D. Wright. 1997. “Convenience Yield without the
Convenience: A Spatial-Temporal Interpretation of Storage under Backwardation.”
Economic Journal 107:1009–22.
Brennan, M. J. 1958. “The Supply of Storage.” American Economic Review 47:50–72.
———. 1991. “The Price of Convenience and the Valuation of Commodity Contingent
Claims.” In Stochastic Models and Option Values, ed. D. Lund and B. Oksendal,
pp. 33–71. Amsterdam: North-Holland Elsevier.
Brennan, M. J., and E. S. Schwartz. 1985. “Evaluating Natural Resource Investments.”
Journal of Business 58:135–57.
Bresnahan, T. F., and P. T. Spiller. 1986. “Futures Market Backwardation under Risk
Neutrality.” Economic Inquiry 24:429–41.
Brock, W. A. 1974. “Money and Growth: The Case of Long Run Perfect Foresight.”
International Economic Review 15:750–77.
Deaton, A., and G. Laroque. 1992. “On the Behaviour of Commodity Prices.” Review
of Economic Studies 59:1–23.
———. 1996. “Competitive Storage and Commodity Price Dynamics.” Journal of
Political Economy 104:896–923.
Fama, E. F., and R. K. French. 1987. “Commodity Futures Prices: Some Evidence
on Forecast Power, Premiums, and the Theory of Storage.” Journal of Business 60:
55–73.
Frechette, D. L., and P. L. Fackler. 1999. “What Causes Commodity Price Backwar-
dation?” American Journal of Agricultural Economics 81:761–71.
Gibson, R., and E. S. Schwartz. 1990. “Stochastic Convenience Yield and the Pricing
of Oil Contingent Claims.” Journal of Finance 45:959–76.
Gray, R. W., and A. E. Peck. 1981. “The Chicago Wheat Futures Market: Recent
Problems in Historical Perspective.” Food Research Institute Studies 44:431–49.
Gustafson, R. L. 1958. Carryover Levels for Grains. USDA Technical Bulletin 1178.
Washington, DC: U.S. Department of Agriculture.
Johnson, D. G., and D. Sumner. 1976. “An Optimization Approach to Grain Reserves
for Developing Countries.” In Analysis of Grain Reserves, Economic Research
Service Report 634, ed. D. J. Eaton and W. S. Steele, pp. 56–75. U.S. Department
of Agriculture, Washington, DC.
Kaldor, N. 1939. “Speculation and Economic Stability.” Review of Economic Studies
7:1–27.
Keynes, J. M. 1930. A Treatise on Money, vol. 2, The Applied Theory of Money.
London: Macmillan.
Leuthold, R. M., J. C. Junkus, and J. E. Cordier. 1989. The Theory and Practice of
Futures Markets. Lexington, MA: Lexington Books.
Litzenberger, R. H., and N. Rabinowitz. 1993. “Backwardation in Oil Futures: Theory
and Empirical Evidence.” Unpublished manuscript, University of Pennsylvania,
Wharton School.
McDonald, R., and D. Siegel. 1986. “The Value of Waiting to Invest.” Quarterly
Journal of Economics 101:707–27.
Miranda, M. J., and X. Rui. 1999. “An Empirical Reassessment of the Commodity
Storage Model.” Working paper, Ohio State University, Department of Economics.
Morck, R., E. S. Schwartz, and D. Strangeland. 1989. “The Valuation of Forestry

This content downloaded from 130.240.43.43 on Tue, 09 Feb 2016 18:10:17 UTC
All use subject to JSTOR Terms and Conditions
E. Bobenrieth H., J. Bobenrieth H., and B. Wright 615

Resources under Stochastic Prices and Inventories.” Journal of Financial and Quan-
titative Analysis 24:473–87.
Pindyck, R. S. 1993. “The Present Value Model of Rational Commodity Pricing.”
Economic Journal 103:511–30.
———. 1994. “Inventories and the Short-Run Dynamics of Commodity Prices.” Rand
Journal of Economics 25:141–59.
Reed, W. J. 1993. “The Decision to Conserve or Harvest Old-Growth Forest.” Eco-
logical Economics 8:45–69.
Saving, T. R. 1971. “Transactions Costs and the Demand for Money.” American Eco-
nomic Review 61:407–20.
Scheinkman, J. A., and J. Schechtman. 1983. “A Simple Competitive Model with
Production and Storage.” Review of Economic Studies 50:427–41.
Telser, L. G. 1958. “Futures Trading and the Storage of Cotton and Wheat.” Journal
of Political Economy 66:233–55.
Thompson, S. 1986. “Returns to Storage in Coffee and Cocoa Futures Markets.”
Journal of Futures Markets 6:541–64.
Thurman, W. N. 1988. “Speculative Carryover: An Empirical Examination of the U.S.
Refined Copper Market.” Rand Journal of Economics 19:420–37.
Tilley, D. S., and S. K. Campbell. 1988. “Performance of the Weekly Gulf–Kansas
City Hard-Red Winter Wheat Basis.” American Journal of Agricultural Economics
70:929–35.
Weymar, F. H. 1974. “The Effects of Inventories on Price Forecasting.” In The Eco-
nomics of Cocoa Production and Marketing, ed. R. A. Kotey, C. Okali, and B. E.
Rouke, pp. 432–49. Legon: University of Ghana Press.
Williams, J. C. 1986. The Economic Function of Futures Markets. Cambridge: Cam-
bridge University Press.
Williams, J. C., and Wright, B. D. 1991. Storage and Commodity Markets. Cambridge:
Cambridge University Press.
Working, H. 1934. “Price Relations between May and New-Crop Wheat Futures at
Chicago since 1885.” Wheat Studies 10:183–228.
———. 1948. “Theory of the Inverse Carrying Charge in Futures Markets.” Journal
of Farm Economics 30:1–28.
———. 1949. “The Theory of Price of Storage.” American Economic Review 39:
1254–62.
Wright, B. D., and J. C. Williams. 1989. “A Theory of Negative Prices for Storage.”
Journal of Futures Markets 9:1–13.

Notes
* We gratefully acknowledge support from the U.S. Department of Agriculture
National Research Initiative project 200-35400-10599, from CONICYT/Fondo Na-
cionál de Desarrollo Cientifico y Tecnologico (FONDECYT) project 1010239, from
Dirección de Investigación, University of Concepción, and from Dirección de Inves-
tigación, University of Bı́o-Bı́o. We would like to acknowledge the invaluable assis-
tance of Carlo Cafiero of the University of Naples, Portici, in the preparation of this
article.
1. See Brennan and Schwartz (1985), McDonald and Siegel (1986), Morck,
Schwartz, and Strangeland (1989), Gibson and Schwartz (1990), Brennan (1991),
Litzenberger and Rabinowitz (1993), Pindyck (1993, 1994), and Reed (1993).
2. See, e.g., Working (1934, 1948, 1949), Brennan (1958), Telser (1958), Weymar
(1974), Gray and Peck (1981), Thompson (1986), Fama and French (1987), Thurman
(1988), Tilley and Campbell (1988), Pindyck (1993, 1994).
3. See Williams (1986) for a critique of much of this literature.
4. Benirshka and Binkley (1995) find indirect support for an explanation based

This content downloaded from 130.240.43.43 on Tue, 09 Feb 2016 18:10:17 UTC
All use subject to JSTOR Terms and Conditions
616 Economic Development and Cultural Change

upon spatial aggregation and transport costs in an empirical model of the U.S. grain
markets, while the results of Frechette and Fackler (1999) are more ambiguous. Bren-
nan et al. (1997) demonstrate the ability of spatial aggregation of stocks within a
marketing system to generate the appearance that stocks are held while the price spread
is in backwardation using a programming model based on actual cost and engineering
data from a Western Australian wheat market.
5. We assume a standardized competitively produced homogeneous commodity.
Marketing cost as defined here does not include advertising or other forms of demand
stimulation often associated with marketing.
6. The function G is not convex, since the determinant of the hessian of G is
⫺h (x)h (z) ! 0. Given the accounting identity z { c ⫹ x, the cost of marketing could
equivalently have been written as a function k(c, x). This function contrasts with a
transactions cost function familiar from monetary models which has similar arguments,
consumption and money stocks (Saving 1971; Brock 1974; Bougheas 1994) but is
assumed to be convex.
7. For grains, though storage is above ground, clean-out of the last bushels can
be an arduous, unpleasant, and even unsafe task. One of the authors thanks his former
summer employer, Pacific Seeds, Ltd., of Biloela, Queensland, Australia, for allowing
him to learn firsthand the difference between the effort involved in dumping the first
bushel of grain into a seed-drying bin and the effort involved in removing (with a
brush) the last bushel from the bin.

This content downloaded from 130.240.43.43 on Tue, 09 Feb 2016 18:10:17 UTC
All use subject to JSTOR Terms and Conditions

You might also like