Agricultural & Applied Economics Association

Computation Techniques for Intertemporal Allocation of Natural Resources Author(s): Duane Chapman Reviewed work(s): Source: American Journal of Agricultural Economics, Vol. 69, No. 1 (Feb., 1987), pp. 134-142 Published by: Oxford University Press on behalf of the Agricultural & Applied Economics Association Stable URL: http://www.jstor.org/stable/1241314 . Accessed: 06/02/2012 10:29
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.

Agricultural & Applied Economics Association and Oxford University Press are collaborating with JSTOR to digitize, preserve and extend access to American Journal of Agricultural Economics.

http://www.jstor.org

Techniques Computation for Intertemporal Allocation

of Natural
Resources
Duane Chapman
Application of optimal control theory to applied problems is limited by the difficulty of numerical solutions. Typically, terminal values for the production period, price, or production level have been assumed rather than optimized. The use of an objective functional with explicit discounting gives direct solution values for n, y(t), p(t), and rent (or consumer surplus) for continuous or discrete problems. The method is usable for numerical solutions to problems with linear demand, cost trend, or expropriation risk. It is illustrated with Fisher's widely used discrete problem and with application to parameters representing remaining world oil resources for competitive and monopolistic assumptions. Key words: control theory, natural resources, oil.

The theory of natural resource use over time has evolved considerably since Hotelling's pathbreaking work fifty years ago. Clark, Dasgupta, Gilbert, Fisher, Kamien and Schwartz, Heal, Smith and Solow have been among the leaders in this effort. However, applying this theory to numerical problems has been difficult. The purpose here is to describe in a precise and accessible way some problems and numerical solutions dealing with oil resources which can be used in teaching theory and analysis. The ambition, then, is small, intending to restate in simple form material which has been thoroughly discussed in the literature. Some interesting implications arise from the numerical solutions. Dramatically different time paths of oil use can have about the same objective values. Thus, with the present world economy, an oil cartel could for the present ignore depletion. Similarly, the socially optimal competitive path is unaffected by depletion for many years. Production paths for monopoly and competDuane Chapmanis a professorof resource economics, Cornell University. Assistance and criticismhave been offered by Simon Levin, Margaret Morgan,WilliamPodulka,RichardRand, WilliamTomek, WilliamLazarus, Gene Heinze-Fry, Nancy Birn, Joseph Baldwin,andPaulFassinger,all at CornellUniversity,anda Journal reviewer.

itive solutions are generally thought to cross, as occurs here. But cumulative production is always greater in the competitive solution, and the difference grows throughout the competitive period. The length of time of the production period is frequently assumed in optimization problems and is usually viewed as years to depletion. In part, this assumption reflects the difficulty of determining numerical answers to the problem in a continuous framework. But suppose that a monopoly has an n-year contract. Will it make radically different decisions than a monopoly that can choose n? Or suppose the expropriation question is defined as a probability problem for lost profit. Does this affect numerical solutions? Of course, both answers are affirmative.

Fisher's Problem and Optimal Control As a text for the discussion, Fisher's problem is useful because his Resource and Environmental Economics has attained wide usage. He gives this problem (pp. 14-18, 37-39): there are ten barrels of oil in the ground, the price demand function for oil is $10 - y, cost is a constant $2, and the discount rate is 10%. What is optimal use over two years? The objectives are to compare social opti-

Economics Association Copyright1987AmericanAgricultural

Chapman

NaturalResourcesAllocation 135

Table 1. Periods
Objective

Fisher's Oil Problem: 10 Barrels, 2

Monopoly Profit 4 4 $6 $6 8 $30.55 $15.27 Competitive Social Optimum 5.14 4.86 $4.86 $5.14 10 $28.57 $23.95

Period0 use, barrels Period 1 use, barrels Price, period0 Price, period 1 Total oil use Profitpresent value Consumers'surplus, presentvalue

Note: Cost is a constant $2 per barrel. Consumers' surplus before discounting is fpdy - 2y.

mality with competitive and monopoly solutions. If all of the oil must be used, the monopolist uses 4.95 barrelsin the first year and the remaining5.05 barrelsin the second year. But the competitive solution is socially optimal, and those values are 5.14 barrels in the first period and 4.86 barrelsin the second. If the monopolist need not produce all the amounts are four oil, the profit-maximizing barrels per year in each period, leaving two barrelsunused. The socially optimal,competitive solution would, if unconstrainedby a resource limit, use marginalcost pricing, and seek to use eight barrelsper period. Since this is not possible, the constrainedvalues for two years (i.e., 5.14 and 4.86) are correct. Table 1 shows these values. Fisher has specified the numberof time periods, the demandfunction as linear, and cost as constant. These assumptions rather than the specific values (i.e., 10 barrelsavailable,2 periods)are the key to easy solution. But if the numberof time periods, n, is made a control variable, the problem is more complex. In fact, for the monopoly the optimalnumberof periods is eight years. In what follows, a method of selecting n is developed that is applicable for finding numerical solutions for general continuous problems involving representative world oil data and alternativeobjectives. More formally, the problemis to maximize the present value of economic rent by finding the optimal time path of production and the optimal length of time for production.' The time path or trajectoryof productionis conThis definition of the problem and first-order canonical conditions is based upon Intriligator's discussion, chaps. 11, 14.

trolled by the decision maker to optimize the desiredvalue. Price is then determined proby duction, as is economic rent. Productionis always the rate of change in cumulativeproduction. Admissiblevalues of productionmust be such that price, production, and rent are always nonnegative. If all the resourcemust be used, then cumulative productionreaches the originalstock at the end of the optimal time. But if resource exhaustion is not required, then cumulative productionmust not exceed the originalstock, and optimization may leave part of the resource unused when productionceases. This is the basic problemfor any finiteresource, on any scale whethermicro(e.g., a smalldeposit) or macro (e.g., global oil). First, the continuousproblemis stated, and then Fisher's discrete problemis considered. In a later section, the monopoly profitobjective is replacedby social welfare and competitive objectives, and numericalresults follow. The continuous problem, initiallyfor a monopoly, is to maximize the present value of profits, V, with respect to n, the length of the production period, and y(t), the production path.
maximize V =
{y(t), n)

P(Y)Y - rydt;
e

where
(1) p(y)
= P2 -

PIY,

y(t)

dX(t)

dX( , dt S, -

y dt = X(n)
p,y py - ay ?-0,0.

Table 2 summarizesthe definitionsfor equation (1). For ease of solution,p(y) is linearand cost cxis constant for now. Cumulativeproductionat any time is X(t), a variable reflecting the state of remainingresources. Production, y(t), whatever form it takes, is the rate of change in cumulativeproduction. Originalresource stock is S; it cannot be exceeded. However, the not-more-than constraintcreates modest difficultiesfor solutions. Will the optimal y(t) use up all of the resourceby n, or will use cease with oil stocks still available?Finally, prices, quantities,and

136 February1987

Amer. J. Agr. Econ.

Table 2.
Variable a 3o

Definitions of Variables
Definition

Average cost per unit Optimalmonopolyproduction,unconstrained Parameters price demandfunction in 13, 32 P3 Competitivesocially optimalproduction, unconstrained CV Competitiveindustry,presentvalue of (3c) profit aX DV Discrete V; see V below at H Hamiltonian function It is still the case that y = dX/dt. Dynamiclagrangemultiplier,costate varih Finally, the second-ordercondition is able factor for regularinvestment SAccumulation d2H at interest 0. (4) 2 n ay2 Lengthof time period Price or price demandfunction p Before applyingthese relationships,the disr Interestor discountrate S crete problemis introducedin order to show Originalresourcestock SV Social value; presentvalue of consumers' that eight years is the optimal monopoly pesurplus riod and how y, is made explicit. The discrete Time t general analogue is summarizedwith slight V Presentvalue of monopolyprofit modificationof equations (1)-(4): X Cumulative production Productionor productiontime path y

The Hamiltonian provides the first-order conditions to solve (1): Either aH (3a) 0, or ay (3b) 0, and at

(5a)
n-1

profitmust always be nonnegative.The maximummonopolyprofitis foundby determining the functionalformof y(t), its actualnumerical values, and the length of time n. The best optimal control technique to use with this problem is the maximumprinciple. This principleasserts that study of a simpler function based upon the relationshipsin (1) can provideinformationuseful in findinga solution. By examining first-orderconditions, the natureof the solutionmay be determined.2 The Hamiltonian function in this context has two components. The first part is the integrandwhich is being maximizedover time. This is the discountedprofitfunction here:
(2) H p(y)y - iy = eYert --Xy.

maximize DV = {y,,n}
n-1

( + y (1 + r)t

(5b)
(5c)

t=0

y, = X(n - 1)
H - ay = p(Y)Y r)' (1 +

S,
Xy.

DV is the discrete version of net present value of monopoly profit. The other constraints, conditions, and definitionsin table 2 still hold. The logical interpretationof these relationshipscan be difficultbecause the true optimum may fall between the integer time units. The Fisher problem values are, as noted,
p(y)

The second part of the Hamiltonian is the productof the co-state variableh (similarto a lagrangianmultiplier)and y, the production rate which is the rate of change in the state variableX. In (2), h gives the change in discountedprofitassociated with a smallincrease in resource endowment. Approximately,X is dV/dS.
2 All of the functions are continuous and differentiable, and the

barrelsoriginalendowment,r = 10%per year discount rate, and a = $2 per barrelconstant

cost.

= $10 - y, P2 = 10, P1 = 1, S = 10

First, consider the solution when all the resource is used so X(n - 1) = S. Applying the

first-orderconditions, beginningwith (3a): (6a)

- 3p - ot aH OH p + y y _ - h = 0, y (1 + r)t

functionbeingmaximized (discounted profitin this section)is apconcave. propriately

Chapman

Natural Resources Allocation P2 -

137

(6b) (6c) y =

(1 2P1yr)'+ (1 + r)'

a o
(P2 -

, 2
Or)

Equation (12) shows optimal y when n is known. This can be put back into (5a) to find the optimal n: (13)
maximize DV maximize DV=
{n}

n--I t=o

(10 y,)yt
(1.1)'

2y,

Apparently, the co-state multiplier does not change over time, since applying (3c), aH (7) 0 ah

Since h is constant, y in (6c) can be used in the (5b) definition of cumulative production using all the resource.
n--I

where Yt is given in (12). This is easily programmed, and the maximum discounted present value is $51.57, the optimal time period is eight years (i.e., n = 7), and oil use declines from 2.08 barrels in the initial period to 0.25 barrels in the last period. Price rises from $7.92 per barrel in the initial

year to $9.75 in the last year. The value of X :

\(2(1+ r) + o
S=

(8)

where
n--1

2 -

2r1 (9) n30 (9 n + (1 + r)' = S. + 2

>311t=S

from (10) is + $3.85. The interpretation of X in this discrete format means that, given n = 7, dDV/dS = $3.85. (Remember that the initial period has t = 0, and n = 7 means an eightyear period.) Global, Discrete Oil Use: Monopoly Solution There are 1.189 trillion barrels of oil remaining, more or less, from the earth's original endowment of 2 trillion barrels. This includes proved reserves as well as oil remaining to be discovered.3 If the demand elasticity is -0.5 at current annual world oil use of 20 billion barrels, then for linear demand, p = $75 2.5y.4 Figuring average world production cost at $20 per barrel, parameters for a global oil problem are p(y) = $75 - 2.5y, P2 = 75, 13= 2.5; S = 1,189 billion barrels remaining oil resource; r = 10%per year discount rate; a = $20 per barrel constant cost; and = 11. 1o Using equation (11), oil production in year t is yr = 11 + (1189 - lln)(1.1)' = tix(n) gn Yt ll Searching for the optimal n is more time consuming with these values. It is helpful to find an initial value which may be close to the optimum. Suppose expropriation is not a problem and y declines to near zero as is often
3 From USGS, Masters et al., 1983 resource estimate, modified for 1983-84 actual use. Other accessible discussions of total oil resources are the Oil and Gas Journal; Exxon; Chapman; and Daly, Griffin, and Steele. 4 Long-run price elasticity values include Daly, Griffin, and Steele's -0.73; Pindyck's linear demand values which include -0.33 and -0.90; and -0.30 from Adams and Marquez. Kouris summarizes several studies; for retail gasoline in the U.S., values range from -0.36 to -1.02.

The summation of the interest terms is the familiar [(1 + r)" - 1]/r, the accumulation factor for n years of regular investment at interest. Define this as gi(n); n has to be determined. Consequently, solving for K in (9), (10) (11)

K = -211(S - nI3o) and p(n)

y = o + (S - nI3o)(1 + r) g(n)

Note that {o, n, and S determine whether y, increases or decreases over time. The 1o term defined in (8) is simply optimal y in the absence of a resource constraint. It is the usual result of maximizing py - ay. Equation (11), then, says that optimal production in the early years will be closest to the unconstrained level. As time passes, y, diverges from 3o. The sign and rate of change of this divergence (i.e., (yt - Bo)/at) depend upon the numerical values of resource endowment, the optimal production period, and the interest rate. Generally, if the resource constraint applies and n is properly chosen, y, is expected to decline over time. Using Fisher's values given above, the demand function is (12) Y, 4 + (10 - 4n)(1.1)t (12) y, = 4 + j(n)
-

(14) (14)

(t = 0,1,2,...,n).

138 February 1987

Amer. J. Agr. Econ.

assumed. Then, by setting y(t = n) = 0, an initialn can be found. Using (12), and then the (18) global values, (15) y(t = n) = o30 + r(+r)= + (1 (16)(16) n which gives (.1'.1 Following the simple iteration first used by Hotelling (p. 142), n = 118 is the closest integer value. Below, this method is shown to define the unique optimal n for the continuous problem. The present value of discretely discounted profit is $3.3274 trillion. Oil production and consumption would decline from 11 billion barrels in year "O" to 1.09 billion barrels in year 117. Price would rise from $47.50 per barrel in the initial period to $74.77 in the final period. Given oil use of about 20 billion barrels annually and a price of $27 per barrel in 1985, monopoly profit maximization is not attained, at least based on these assumptions.5 (17) n =
118.09r)n -

H 8y y

dHy p+ = - herr

dp y y-p ert

(P2 -

0,

(19)

2P1

2P1

a)

S +--1 1o r

(1 + r)-" r

As above, since (3c) leads to - aHIaX = 0, then ah/lat = 0, and h is constant for a given n. Consequently, if the resource is exhausted, then production over the period is
'-- kert

(20a)

(O

2pB2
-(er"-

+ Bo0dt = S, 21 2131
, Or

where 3o =

(20b)

1) + n[o = S.

r 2P1 r~ l So, with the accumulation factor, now g(n) = (er - 1)/r, - niPo) (21) - x = S2P1(S - n3) 23( I,(n)

(22) y(t) = (S - n3o)ert + t = 0, n. px(n) 3o, These expressions for y(t) and X are identical to the discrete forms for y, and h, [(10) and (11) above]. Clark (p. 146) outlined part of this approach ten years ago, noting, "Determination of T [the optimal exhaustion date], howContinuous Solutions: Competition, ever, is a non-trivial matter." He suggested Monopoly, and Social Optima iterating X. However, an explicit solution for the opAlthough continuous relationships are more timal present value of monopoly profit is deeasily interpreted, they have not attained fre- rived by using y(t) in (1),7 and quent use in numerical problems. In part, this is because evaluation of present value V in (1) _ 13(S - nIo)2 (23) V = p(n)P13o2 can be very difficult. However, new programs ern (n) such as MACSYMA now permit resolution of Also, note that maximizing V for n gives a this difficulty.6 Explicit answers to the problem in (1) for the global numerical assumptions unique solution to the problem of optimal n: follow. 1 S e-rn Beginning with equations (1)-(4) for the mo- (24) n = 1o + r er r nopoly case, the first-order (canonical) conditions are This of course provides the basis for the Hotelling iteration for equations (15)-(17). For this problem, (24) is the single unique opSThe $27 per barrel figure represents an average price for all of 1985. See Weekly Petroleum Status Report, recent issues (U.S. timum length of time for production. The
Department of Energy). 6 MACSYMA is a computer program for algebra and applied mathematics which, because of its calculus capabilities, is particularly useful for solving optimal control values. Rand offers a good program guide, and complete documentation is published by the MIT Mathlab Group and Symbolics, Inc. Microcomputers can use similar packages.
7 Equations (23) and (35) and the proof of their maximization by (24) and (33) are contributions by Margaret Morgan and William Podulka in a Cornell graduate seminar, "Economics of Resource Use."

Chapman

Natural Resources Allocation

139

value for n is found by simple iteration or by Newton's method and is 118.09.8 The present value of monopoly profit is $3.025 trillion. This is reassuringly close to the discrete solution. The second-order condition [from equations (18) and (4)] is ay2 -- e The same methods find numerical solutions to the socially optimal, competitive problems. Both have the same first-order conditions. The competitive problem is (25)
a2H = - 2P1<< 0.

(33) nc (34) ay

(S
c

1 +- r 3 e <0.

e-rn

As with the monopoly, optimal social value is solved explicitly by using equation (31) in equation (27), and n, maximizes this social value:
(35)

SV

2ern ~l~32p?(n)

(S-

2pi(n)

np3)2

In comparing monopoly and competition for nm} and {yc(t), nc} for these global (26) maximize CV values, the production and price paths in{y(t), n} tersect near nc. The socially optimal path begins at an annual 22 billion barrels and deP(Y)Y Xy dt, ap 0O, clines for 64.02 years until the resource is o -n ert exhausted. The monopoly path begins at 11 and the social optimum problem is, for social billion barrels, stays at this level for many value SV, years, and then declines toward zero. Equations (36) and (37) summarize these numerical results. (27) maximize SV
{Ym(t),
{y(t), n}

(36) =
yo(t)0 o(p(y)

Ym(t)=

11

- a) dy

dt. e-rt

.818 e and 10,000 and

.818 e.It

The competitive and socially optimal Hamand Hs, are iltonians, Hc p (28) H P(Y)Y - y (28) hy, 0; e Hr 'y ---

(37)

yc(t) = 22 - .0364 eit.

(29)

=
Hs OH
ey

The subscript c indicates competitive, socially optimal solutions, and m indicates monopoly solutions. The price paths do have the expected relationships to the interest rate and opportunity cost:
(38)

(P(Y er_

-)dy -

dpm
dpm

For each, the first-order condition is identical. (30) p(y) - ex - h = 0. p(y)ert

.5rKmer' > 0,

dtP (39) dt rcert > 0.

Again, since the first-order condition ah/at = - H/IaXis zero, h is constant for any time period n. Continuing, and using yc for the socially optimal, competitive path, (31) S(S =+
yc(t)
=

- nzP3) n) ert n3)e Ix(n) 3,r+

Here hm is .04 of 1? and hKis 9?. The depletion periods are considerably different. These results imply that an increment to world oil resources is more valuable to a competitive world economy than to a world cartel.

f33, =

where 13 (32)
- nr3)r1 X = - (S p (n)

The Numerical Paradox: Impatience and Myopia Are Almost Optimal Production paths that diverge sharply from the optimal can give essentially equivalent profit. This is equally applicable to solutions for both the social optimum and monopoly problems. Define a myopic planner as setting y at a

8 Conrad and Clark (chap. 1.6) show applications of Newton's solution method to optimal control resource problems.

140 February 1987

Amer. J. Agr. Econ. 50

-Ya: impatient paion yb: myopic paion Y: optimal plan

50

40 0..

T .5.74

.0S
.....-. ..........."..'.'.'.

30:::::: - ... ''- :' '_',,,i,,;_'::::: :

S~~~~~I:f~iiiiliiiiiitiiiiiiiiiiiiiil~iir~iiiiiiiilii

itii jijijiiiiiiiiitiijrit iiii~iiit~i~iiiiirii

: : : : : y :2:::::i: :::::::

.0

20
I
I0-

.0

(s 6.03

0 Figure 1.

16

24
Years"

32

40

48

56

64

1985 to 2049

Petroleum use profiles and net social value

constant 33 level until the resource is exhausted after S/33 years. (Recall 33 is the socially optimal level of production in the absence of a finite resource constraint.) The myopic planner's 22 billion barrels per year lasts 54 years, earninga discounted net social value of $6.023 trillion, rounded to $6.02 trillion for path yb in figure 1. The true optimum, Y* in figure 1, has productionwhich exhausts the resource after 64 years, attaining$6.030 trillion. In this instance, full optimizationimproves present value by a mere $7 billion. Define an impatient planner as wishing to initiate productionat the 33 level, gain a net social value comparableto the true optimum $6.03 trillion, and leave the oil business. This can be accomplished as follows. First, use a

36-year period. But the net social value for this dramaticallydifferent path is $5.74 trillion, 95%of the true optimum.Note in figure1 the contrast between Ya and Y*. Figure 1 raises a problem:optimalplanning may not always be necessary to gain optimal value. Stronglydivergentproductionpaths in some cases yield comparableobjective function values. This is not always the case, particularly where the resource is very limited. In the Fisher problem above, for example, this comparabilitydisappears. A sensitivity analysis indicatesthat the phenomenon may apply to a broad domainof parameter values applicable to world oil markets. In figure 1, the unshadedarea represents oil productioncommon to both divergentpath form ya(t) = 33 + 34ert. Second, define com- Ya and optimal path Y*. The shaded area is parable social value (or monopoly profit) as productionnot common to both paths. A di95%of the optimumvalue: vergence ratio can be the ratio of the sum of not-commonproductionto the total resource na} = .95 SV*{y*(t), n*}, or SVa{ya(t), produced on the optimal path. Graphically, this is the ratio of shadedarea to the appropriVa{ya(t), na} = .95 V*{y*(t), n*}. ate S for each case. Numerically,this resultis: Now find 34,which, meetingthe nonnegativity f - y*(t)l)/Sdt.9 constraints,gives the least years of production (lya(t) na. For figure 1, for example, with the basic parameters for the maximum social value 9 Since ya usually accelerates while Y* declines, and the proproblem,the impatientplannerpath is ya(t) = duction periodYl for is always shorter, a correlation measure of
22 + .577e"/. Petroleum use increases over a
divergence is less descriptive.

Chapman

NaturalResourcesAllocation 141

Table 3. Ratios

Sensitivity Analysis of Divergence

Competition or Socially Optimal Monopoly Plan .75
.53

Cases 1. Originalglobal oil parameters

2. But r = 5% 3. r = 10%, but S = 594.5

.53
.41

billionbl
4. r = 5% and S = 594.5

.53 .40 .39

.41 .39 .40

billionbl 5. Elasticityis -1.2, cost is

$10 bl, r = 5%, S =

594.5 billionbl

Table 3 shows a sensitivity analysis where divergence ratios are compared for several monopoly and social optimum-competition cases. In each instance, original resources, interest rate, cost, or elasticity are varied by sizable decrements for each market structure. Optimal paths and optimal lengths of production vary accordingly, as does the impatient path ya for each case. Figures analogous to figure 1 exist for each of the ten cases. Divergence ratios are noticeably large for all cases. A similar sensitivity analysis for the myopic planner (yb(t) = fo for monopoly, for social [3 optimum) is not shown, but in eight of the ten cases the value attained or exceeded 95% of the true optimum. The problem is not resolved. The observed comparability of solution values may be dependent upon the functional specification or parametric values used here.

Concluding Comments In mid-1986, average world contract and spot prices were fluctuating around $15 per barrel. This compares to a mid-1985 value of $27 per barrel. U.S. and world petroleum use was increasing as a consequence of lower prices. A general conclusion is that world oil markets moved from a mixed competition/cartel combination in 1985 to a competitive market in 1986. The presence or absence of an effective world monopoly is probably much more important in determining annual use, prices, and the years to exhaustion than are specific parameter values.

There is no intentionhere to determinethe empiricalparametersof world oil marketsand the natureor durationof the currenttransition to a primarily competitivemarket.Griffin'srecent article on the Organization Petroleum of Exporting Countries addresses these points. Solow observed twelve years ago thatbecause optimal n is longer for monopoly than for the competitive social optimum, "the monopolist is the conservationist's friend" (p. 8). However, the numbersshow an often-obscuredrelationship.Potentialmonopoly profitwith basic parametersis $3 trillion, one-half of the potentialnet social value of $6 trillion. The general point here is that numerical analysis with optimal control gives different economic interpretations thanwould be possible with theory alone, and that dramatically nonoptimalpaths can have almost optimalobjective function values. To recapitulate,findingnumericalvalues for the optimallength of a productionperiod has been difficult,and findingnumericalvalues for economic rent and social value functions has been nearly impossible. With computerassisted algebra and analytic solutions, the process can be shortened significantly. The steps are (a) define the first-ordercanonical conditions with discounting explicit; (b) express y, the equilibrium quantity,as a function of the parametersfor demand, cost, and the co-state multiplier;(c) solve for the co-state multiplier X as a function of the resource stock, the demand and cost parameters,and the interest accumulationfactor; (d) solve for y, now an explicitfunctionof n, t, the resource stock, and the demand and cost parameters; (e) find optimal n; (f) evaluate the objective function; and (g) evaluate the second-order conditions. The general technique should be followed for possible optima when all of the resources need not be used. In both cases only permissable values of quantities, prices, and profit should be examined. Compareas appropriate the constrained optimum path and length of time period in which all of the resource must be used with the unconstrainedoptimum. These steps lead quickly to numericalanswers for different decision goals, whether they be monopoly profit, consumers' surplus, or competitive equilibria. Both continuous and discrete formulationsare easily solved, and the method is easily extended to include the risk of expropriation and a technical change/environmentalcost continuous shift (Chapman 1986, pp. 21-23). With these im-

142 February 1987

Amer. J. Agr. Econ. Intriligator, Michael D. Mathematical Optimization and Economic Theory. Englewood Cliffs NJ: PrenticeHall, 1971. Kouris, George. "Energy Demand Elasticities in Industrialized Countries." Energy J. 4(1983):73-94. Mathlab Group. MACSYMA Reference Manual, 2 vols., rev. ed. Cambridge MA: MIT Laboratory for Computer Science, and Symbolics, Inc., 1983. Morgan, Margaret. "An Examination of the Effect of a Miscalculation of the Level of Available Oil Resources on the Optimal Rate and Length of Depletion." Paper presented in Economics of Resource Use seminar, Dep. Agr. Econ., Cornell University, 1985. Oil and Gas Journal, 31 Dec. 1984. Pindyck, Robert S. "Gains to Producers from the Cartelization of Exhaustible Resources." Rev. Econ. and Statist. 60(1978):238-51. Podulka, William. "Fisher's Problem: Some General Observations and an Example Using Variable Interest Rates." Paper presented in Economics of Resource Use seminar, Dep. Agr. Econ., Cornell University, 1985. Rand, R. H. Computer Algebra in Applied Mathematics: An Introduction to MACS YMA. Boston MA: Pitman, 1984. Solow, Robert M. "The Economics of Resources or the Resources of Economics." Amer. Econ. Rev. 64(1974):1-14. U.S. Department of Energy. Weekly Petroleum Status Report. Washington DC, various issues. U.S. Geological Survey, C. D. Masters et al. "Distribution and Quantitative Assessment of World Crude-Oil Reserves and Resources." Open-File Rep. 83-728, Reston VA, 1983. -. "World Petroleum Resources-A Perspective." Open-File Rep. 85-248, Reston VA, 1985.

proved solution techniques, classroom use and appliedresearch can grow substantially.
[Received July 1985; final revision received June 1986.]

References
Adams, F. Gerard, and Jaime Marquez. "Petroleum Price Elasticity, Income Effects, and OPEC's Pricing Policy." Energy J. 5(1984):155-228. Chapman, Duane. Energy Resources and Energy Corporations. Ithaca NY: Cornell University Press, 1983. . Working on Fisher's Problem with Computer Algebra. Cornell Agr. Econ. Staff Pap. No. 86-11, 1986. Clark, Colin W. Mathematical Bioeconomics. New York: John Wiley & Sons, 1976. Conrad, Jon M., and Colin W. Clark. Notes and Problems in Resource Economics. Cambridge: Cambridge University Press, forthcoming. Daly, George, James M. Griffin, and Henry B. Steele. "Recent Oil Price Escalations: Implications for OPEC Stability." OPEC Behavior and World Oil Prices, ed. J. M. Griffin and David J. Teece. Boston: Allen & Unwin, 1982. Exxon Corporation. "Improved Oil Recovery." New York, 1982. Fisher, Anthony C. Resource and Environmental Economics. Cambridge: Cambridge University Press, 1981. Griffin, James M. "OPEC Behavior: A Test of Alternative Hypotheses." Amer. Econ. Rev. 75(1985):954-63. Hotelling, Harold. "The Economics of Exhaustible Resources." J. Polit. Econ. 39(1931):137-75.