Professional Documents
Culture Documents
BY
by Kirk Hamilton
Centre for Social and Economic Research on the Global Environment University College London and University of East Anglia
Acknowledgements The Centre for Social and Economic Research on the Global Environment (CSERGE) is a designated research centre of the U.K. Economic and Social Research Council (ESRC) The advice and comments of David Ulph, Malcolm Pemberton and David Pearce are gratefully acknowledged. The usual caveats apply. ISSN 0967-8875
1.
INTRODUCTION
Concern about damage to the environment and depletion of resources has made sustainable economic development a concept with both wide currency and wide interpretation, as Pezzey's (1989) exposition demonstrates. Although it is relatively easy to criticize various notions of sustainable development (see, for instance, Nordhaus (1992)), it is the goal of this note to explore a particularly simple definition, that per capita utility be non-declining, owing to Pezzey. The key question to be answered is whether, or under what conditions, sustainable development so defined is consistent with optimal growth and finite resources. Minimal sustainability is defined to be constant utility over time. If the utility function is a continuous and non-decreasing function of consumption, then minimal sustainability is equivalent to constant consumption over time. The initial problem to be examined is that of finding a development path with maximal consumption that is minimally sustainable. Stated this way, it is clear that this is equivalent to a maximin programme, which has been studied by Solow (1974), Hartwick (1977,1978), Dasgupta and Heal (1979), Dixit, Hammond and Hoel (hereafter DHH) (1980), and Dasgupta and Mitra (1983). The starting ethical position in this work is different, essentially a Rawlsian framework in which welfare across time is equal to that of the least well-off generation, but the end goal is the same: maximal constant consumption. A key result in this literature is Hartwick (1977), which showed that the "Hartwick rule", to invest resource rents, is a sufficient condition for a maximin programme. The first part of this paper fills in several of the gaps in the literature on the Hartwick rule: (a) a simple proof is given, for general production functions, that the generalized Hartwick rule combined with the Hotelling rule is necessary and sufficient for maximal constant consumption; (b) the behaviour of a Hartwick-Hotelling programme for CES production functions is thoroughly explored, including a correction to one of the results in Hartwick (1978); and (c) a weaker condition for maximal consumption than that in DHH (1980) is derived for CES production functions. Some conclusions are drawn concerning the debate on "strong" versus "weak" sustainability (Pearce, Markandya and Barbier, 1989). The second part of the paper examines the conditions under which utility can be increasing in a programme of optimal growth with finite resources.
2.
We assume that there is constant population (so that labour can be treated implicitly in the production function), and no disembodied technological growth. The initial endowment is a stock S01 of resources and K02 of capital. Output F3 is produced from capital K4 and
(1)
& F R = FK , FR
and
(2)
lim S t = 0.
t
(3)
The first of these is the familiar Hotelling rule; in the economy postulated holders of natural resource stocks must be indifferent between holding resources or the alternative asset, capital, which yields FK7. Any programme that left unexploited natural resources would clearly be inefficient, hence expression (3) which says that the programme must exhaust the initial resource stock.
The generalized Hartwick rule (to use the terminology of DHH) is given by:
(4)
where the return to resources, FR8, is the resource rental rate. What follows is a proof, for general production functions, that the combination of the Hotelling rule and the generalized Hartwick rule is necessary and sufficient for constant consumption. Applying expressions (2) and (4) we have:
Now define
so that
& & = FR Z . Z FR
This equation has solution Z = vFR11 for constant v12, and therefore,
& = F ( R + v). K R
It is clear from this derivation that the parameter v13 is simply a constant of integration - there is no obvious economic interpretation. Having established that the Hotelling rule and the
generalized Hartwick rule are together necessary and sufficient for constant consumption, the next questions to be examined are the behaviour of the system under different assumptions about v14, the attainment of maximal consumption and the existence of solutions under this programme.
The efficiency condition that S 0 15 implies that R 0 16. If v <017, eventually K <018, and therefore, assuming capital can be consumed and given the fixed initial endowment K019, both R20 and K21 will tend to 0; assuming that no output is produced purely by labour, constant consumption is impossible.
To take the argument further requires more structure for the production function. Since it is clearly the degree of substitutability between capital and resources that is of key importance in models with exhaustible resources, a convenient functional form is the class of constant elasticity of substitution (CES) production functions. Hartwick (1978) poses several questions about the behaviour of the system under the Hartwick rule for different values of the elasticity of susbstitution, but does not derive the answers.
Defining 22 to be the elasticity of substitution, we have, assuming constant labour force and normalizing per unit of labour,
1 1 1
F = (K
+ R
+1 )
For the case =123 this reduces to the familiar Cobb-Douglas form.
X = (K
and
+ R
+ 1 ),
(6)
< 1.
(7)
FR = X
F F = R = ( ) . X R
The examination of the behaviour of the system for v >024 will be divided into three parts, according to the assumptions about the elasticity of substitution. For the Cobb-Douglas function we have,
C = F FR R FR v
= F (1
v
R
).
If <128 then,
C = F FR R FR v
F = F F v R X
= F (1
1
R R + (1 ) R
).
Again, C29 becomes negative as R 030 (note that 1 31), contradicting C =032.
production is impossible when capital and resources are perfect substitutes in a CES production function. As derived in the Appendix, the growth rate of output for general 38 can be shown to be given by:
& =K & F
& R R ( 1) v ( ). R + v R ( ) v
(8a)
It will be convenient in what follows to use this expression for the case v =039:
(8b)
& K & F
& R v ( ). R + v R(1 ) v
If in the initial period R0(1- )- v >041, then for some time beyond this period,
R=
v,
at which point output is infinite (since the growth rate is positive and infinite - recall that R <042 because of efficiency condition (3)).
F = (K R
= ( (
+ R
+ 1 ) 1 R 1
1 R
1
K ) R
+ +
) 1
C = F FR ( R + v)
F = (1 ) F v( ) . R
Therefore if R0(1- )- v <046, consumption becomes negative as R 0 47.
To summarize the results concerning the value of v48: (i) if it is negative then output and consumption decline to 0; if it is positive then (ii) if the elasticity of substitution is less than or equal to 1, consumption eventually is negative; or (iii) for elasticities greater than 1, the Hotelling rule is violated for very large values of 49, or, in the case of small 50, either output is eventually infinite or consumption becomes negative.
3.
In what follows we employ the more widely known form of the Hartwick rule,
& =F R K R
i.e. that resource rents be re-invested.
(9)
An important question to be explored is the behaviour of output, consumption and investment under the standard Hartwick-Hotelling programme for varying values of the elasticity of substitution. We first consider the case <151.
FR = (K
+ R
1
+1 ) 1 R
1
K = ( ( ) R
+ +
(10)
FR 1 as
R 0.
K T = K 0 + FR Rdt and
0
Rdt = S
0
C = F FR R
=X
1
1
= X 1 ( X R = (K
1
)
1 1
+ R
+1 )
(K
(11)
+ 1 ).
Therefore, since K54 is bounded, consumption tends to 0 as R55 tends to 0 when <156.
This derivation can be compared with Dasgupta and Heal (1979, ch. 7), who show that if the elasticity of substitution is less than 1, then F / R57 is bounded, implying total output is bounded, and therefore that constant consumption is impossible.
The preceding derivation and the result from Dasgupta and Heal (1979) contradict Theorem 3 from Hartwick (1978). This is stated as,
<1 =1 <1
Recalling expression (8b), note that 1-58 as R 059, so that, while F60 may initially be increasing (i.e. for < <161), eventually it must decrease. In fact, since K62 is bounded, eventually K 063, so that F 0-64, contradicting the first part of the above theorem. Because total output is bounded, we know that F 065 in the long run.
Next we consider the case >166. Since resources are not essential in this instance, both Solow (1974) and Dasgupta and Heal dismiss this case. It is still worth exploring, however, the operation of the Hartwick-Hotelling programme under these conditions.
Recalling expression (8b), because >167 and <168, this expression implies that F <069. Because consumption is constant, this in turn implies that FR R 0 70 as R 0 71, since
F (K
+ 1 ) 1 as R 0 .
A further conclusion from the preceding expression is that K72 must be bounded since F <073. This is in spite of FR 74 being unbounded as R 0 75, as is obvious from expression (10).
The behaviour of the system for differing values of >176 is also worth exploring. As was alluded to earlier,
F K + R + 1
as .
This makes the large 77 case uninteresting because the Hotelling rule is violated.
What value of 78 maximizes consumption? This can be examined as follows, following from expression (11):
1 1
C (1) = X 1 ln( X ) (K ( 1) 2
1 1
ln( K )
=X
(K
ln( K )
ln( X ) ln( X ) ) (1 ) ). 2 ( 1) ( 1) 2
For small values of >179 we can approximate ln(X)ln()80. Using l'Hpital's rules we can see that,
ln( x) as 1+ . 2 ( 1)
Consumption is therefore a declining function of >181. Constant consumption under the Hartwick rule is consequently maximized as 1+82, i.e. the Cobb-Douglas production function yields maximal consumption.
Because resources are not essential for elasticities of substitution greater than 1, one strategy for achieving maximal consumption might be to consume all of the resource in the initial period. The derivation in the Appendix shows, however, that such a strategy will not yield constant consumption under the Hartwick rule.
The solution of the system for a Hartwick-Hotelling programme is remarkably simple in the Cobb-Douglas case:
F = K R , , > 0, + < 1.
K = FR R = F
implying,
C 0 = (1 ) F constant F constant
so that
K = K0 +
and therefore,
C 0 t, 1
1
C 0 C R = ( 0 ) (K 0 + t) . 1 1 9
The condition for the existence of a solution to the system is therefore > 83, i.e. the elasticity of output with respect to capital must be greater than that with respect to resources, since R84 must have a finite integral equalling S085. Performing the integration yields the value for maximal consumption,
C 0 = (1 )( )
S 0 1 K 0 1 .
(12)
To tie some of the literature together, it is worth describing the solutions of Solow (1974) and Dasgupta and Heal (1979) to the maximin problem, which did not use the Hartwick rule explicitly. Both choose the Cobb-Douglas production function after rejecting CES functions where total output is bounded ( <186) and where resources are not essential ( >187). For this production function maintaining constant consumption C088 implies that,
K = K 0 + mt
1
R = (C 0 + m) ( K o + mt )
(13)
where m = K89 is a constant. Both point out that efficiency requires that the integral of the above expression for R90 exist and be equal to S091, so that the condition > 92 is required. Performing this integration yields:
C0 = m (
S0 ) K0 m.
Dasgupta and Heal then maximize this expression with respect to m93 to yield the optimal constant consumption,
C0 = (
1 1
1 )( ) S 0 1 K 0 1 ,
By constructing phase
diagrams for the problem he arrives directly at the following expression for the parameter m94 in the system of equations (13),
& = C 0 m=K 1
10
(14)
for some fixed C095. He then integrates the equation for R96 in expression (13), and sets this equal to S097 to arrive at the maximal level of consumption, as given by expression (12).
Hartwick's key insight, that expression (14) embodies the rule "invest resource rents", is not derived explicitly in his 1977 paper; instead he proves the sufficiency of this rule for a programme of constant consumption.
It is also worth linking what has been presented so far to the much more general framework employed by Dixit, Hammond and Hoel (1980)1. If we assume the existence of a competitive output price pt98 and a positive constant exhaustion rent 99 (i.e. assuming efficient resource extraction), then expression (4) may be re-written as,
& = p F ( R + v) pt K t R = ( R + v)
& + v, = S
where v100 is constant. Rearranging terms then gives the analog of the "generalized Hartwick rule" of DHH:
& = v & + S pt K
The authors prove that this rule is necessary and sufficient for constant utility in their general framework. They go on to show that any path with v <0101 is infeasible.
A point of particular interest in the Dixit, Hammond and Hoel paper is their proof that, given any efficient path such that v =0102, any other path with v >0103 and a larger capital stock for all t >0 104 will yield a lower level of utility under the generalized Hartwick rule. Expressing their proposition in terms of the CES production function analysis presented so far, we can say that
11
is a necessary condition for C < C0105, where the "unprimed" variables represent their values 0 for v =0106 (this is a necessary condition because the capital stock can only be greater than its previous value for all t >0107 if its rate of change in the initial period is greater).
Changes in v108 will clearly affect R0109 since the capital stock level will be altered under the generalized Hartwick rule and the efficient path must both exhaust the resource and satisfy the Hotelling rule. For infinitesimal changes in v110 it is shown in the Appendix that the DHH condition for maximal utility is equivalent to
R0 < , v 1
(15)
while a more direct derivation of this condition, using the machinery of CES production functions presented so far, yields
R0 < . v 1
(16)
Recalling that the case >1111 was the most problematic in terms of analyzing its behaviour with respect to changes in v112, and recalling that <1113, we can conclude that the DHH condition is stronger than necessary for this case, since expression (15) is negative and expression (16) is positive. At least for infinitesimal changes in v114 the weaker condition (16) yields maximal consumption under the Hartwick-Hotelling programme when the elasticity of substitution is greater than 1.
12
4.
RISING UTILITY
Are there optimal programmes under which utility is increasing? This could be defined to be strict sustainability. To answer this question requires a different criterion of optimality than that employed to this point.
Given two programmes for which U >0115, it would be rational to prefer the programme for which the sum (or, in continuous time, the integral) of utility over time is the greater. Thus, at least for the purposes of exploring some simple models, we will assume that sustainability is consistent with a Utilitarian ethic as applied to current and future generations. The interesting question, it turns out, is whether there is a pure rate of time preference in the models. If there is, the Utilitarian ethic would suggest that we wish to maximize the present value of utility, where future utility is discounted at some rate r116.
U (C ) = C ( 1) , > 1.
(17)
This is one of the class of functions for which the elasticity of marginal utility, 117, is constant. In the analysis that follows the Cobb-Douglas production function will be used and the efficiency of any programme will again require that the resource be exhausted, as given in expression (3).
Finding the optimal programme can now be expressed as an optimal control problem:
max U (C )e rt dt
C ,R 0
& & + S H = Ue rt + 1 K 2
= Ue rt + 1 ( F C ) 2 R,
where 1118 and 2119 are the co-state variables. The optimum programme must satisfy the following first order conditions:
13
= U ' e rt 2 = 0 C = 1 FR 2 = 0 R = 1 FK = &1 K
= 0 = & 2 S
From expression (19) we see that
(18)
(19)
(20)
(21)
1 =
2
FR
FK d = ( 2 ). FR dt FR
2
or,
FK d = ( 2 ). FR dt FR
FK =
& F R . FR
This is just the Hotelling rule. It is derived directly as an efficiency condition for the optimal control problem.
U '' = . U' C
From expression (18) we see that
14
U ' = 1e rt .
Differentiating with respect to time gives
r +
& C = FK . C
(22)
This is the "Ramsey rule" (after Ramsey (1928)). Optimality requires that the social rate of return on investment equal the marginal product of capital. For the Cobb-Douglas production function the percentage rate of change in consumption is therefore given by,
& C R = 1 ( 1 r ) C K
(23)
Because resources are essential for production and in finite supply, resource use R121 will tend asymptotically to 0 in an optimal programme. The only way that the percentage rate of change in consumption, as given in expression (23), can be non-decreasing is if capital stock K122 eventually tends to 0 as well - i.e. eventually capital must be consumed. But if both capital and resources are declining, then so is output. Since the initial capital stock is fixed, consumption must eventually fall. The optimal path is not sustainable.
By equating the expression for the Hotelling rule and expression (22) it is possible to solve the differential equation to give,
e rt FR C =
for constant of integration 123, and by re-arranging terms,
C = (e
rt
K ) . R 1
1
15
If the discount rate for utility, r124, is 0 then it is clear from this expression that consumption will grow indefinitely as resource use R125 declines. As Dasgupta and Heal (1979, ch. 10) point out, however, the optimum path will only exist if the non-discounted integral of utility is finite, which requires that,
1 .
If the elasticity of marginal utility 126 approaches infinity, then consumption will approach the constant 127.
This latter point bears some elaboration. If any level of consumption yields the same level of utility (i.e. 128), then the optimal programme will be to maintain consumption constant. From the characterization of this programme given earlier, we know that investment will just equal resource rents and, moreover, that the constant 129 will be given precisely by expression (12), the maximum constant consumption level given the initial endowment of capital and resources.
16
5.
CONCLUSIONS
The definition of sustainability as non-declining utility entails strong inferences about optimal growth with finite resources. A central issue concerns the discounting of utility: if the pure rate of time preference is greater than 0, then the traditional Utilitarian maximand, the present value of utility, leads to an optimal programme that is not sustainable. It is worth recalling, therefore, Ramsey's view of discounting utility as "ethically indefensible". If, however, the discount rate for utility is 0 then, under fairly weak restrictions, the Utilitarian maximum yields continually increasing utility, and therefore strict sustainability.
If utility is constant for any level of consumption, or if the stated goal is minimal sustainability (perhaps for reasons of intergenerational equity) then the Hartwick rule, to invest resource rents, is the keystone. Solow (1986) refers to this as a "rule of thumb" for growth policy.
This study has drawn together several strands from a diverse literature on the Hartwick rule. The analysis has shown that, given virtually unrestricted production functions, the generalized Hartwick rule in combination with the Hotelling rule is both necessary and sufficient for consumption to be constant. The Cobb-Douglas production function (out of the class of CES production functions), in which the elasticity of substitution between capital and resources is exactly 1, yields consumption that is constant, positive and maximal when a standard Hartwick programme is followed in combination with two efficiency conditions, the Hotelling rule and complete resource exhaustion.
The magnitude of the elasticity of substitution for CES production functions has major implications for the operation of a Hartwick-Hotelling programme. summarized in Tables 1 and 2. These implications are
Table 1 17
<1131
=1132
135
v <0136
v >0146
<1156
=1157
160
The results for <1172 in Table 2 contradict Theorem 3 in Hartwick (1978). Two results when the elasticity of substitution is greater than 1 are of note. The programme that exhausts the resource in the initial period (which is feasible since resources are not essential in this model) yields non-constant consumption. And a weaker condition than that given by Dixit, Hammond and Hoel exists for the standard Hartwick rule to yield greater (constant) consumption than the generalized rule.
18
The derivationa in this paper emphasize the "knife edge" nature of the Cobb-Douglas production function. Although the generalized Hartwick rule promises constant consumption for very
general production functions, the requirement that a maximal constant consumption path exist places severe limits on the type of production function.
These results also shed light on the debate concerning "strong" versus "weak" sustainability. The weak sustainability position, as typified by Solow, and Dasgupta and Heal, is essentially that the elasticity of substitution between capital and resources is greater than or equal to 1. The results in Table 2 show that for weak sustainability, therefore, the Hartwick-Hotelling programme yields constant maximal consumption if the elasticity is exactly 1, constant submaximal consumption when it is greater than 1, and, in the limit as capital and resources become perfect substitutes, that efficient extraction is impossible (because the Hotelling rule is violated). The strong sustainability position of Pearce, Markandya and Barbier is that the
elasticity of substitution between capital and resources (especially certain critical resource stocks) is less than 1. The preceding results show that in this case the Hartwick-Hotelling programme leads to declining output and consumption. For the proponents of strong
sustainability, therefore, the rule "invest resource rents" is not sufficient to ensure sustainable development.
19
Appendix
Recalling the definitions of X173 (expression (6)) and 174 (expression (7)), a few basic results follow directly from the definition of the CES production function:
FR = X
1 1
F F = R = X R
(A1)
FRK =
1 1 1
1 R X
FK =
1 FK R
(A2)
FRR =
1 FR ( 1) < 0. R
(A3)
& 178. We begin with, The first item to be derived is the expression for F & =F R &+F K &. F R RR RK
Therefore,
20
& K & F
& R v as 1+ . R + v R(1 ) v
The next issue to be considered is the behaviour of the Hartwick-Hotelling system when >1 179 and, since resources are not essential for this value of the elasticity of substitution, when all
& = 0 180 if and only if, given F = F(K,R) of the resource is extracted in the base period. C
F ( K 0 + Fr ( K 0 , S 0 ),0) = F ( K 0 , S 0 ) FR ( K 0 , S 0 ) S 0 .
181,
(A4)
It will simplify the algebra considerably, without unduly affecting the generality of the argument, if we assume for this derivation that + =1182, so that
1 1
X = K
+ R
( ( K 0 + X 1 S 0
which implies that,
1
= X 1 ( X S 0
),
( K 0 + X
1 ) 0
=X
(K 0
),
1 K 0 = X 1 (K 0 1 S 0
).
For consumption to be constant this latter expression should be an identity. However, there are clearly choices of K0183 and S0184 for which the right hand side is less than or equal to 0, so the identity does not hold.
21
Finally, we consider how the level of consumption varies with v185 in the generalized Hartwick rule. The analysis will be based on infinitesimal positive changes dv186, to examine the
transition from the standard to the generalized Hartwick rule - negative values have already been ruled out because they lead to declining consumption. In order for the capital stock to be greater for all time (after the initial period) in the transition to the generalized rule, as Dixit, Hammond and Hoel hypothesize, a necessary condition is that K187 be greater under the generalized rule. This may be written as,
' C0 < C 0 if dv > 0 and FR0 ' ( R0 '+ dv) > FR0 Ro .
FRo v
dv,
R0 ' = R0 +
Therefore,
dv)( R0 + (
= FR0 R0 +
FR0
R0 dv + FR0 (
where terms in dv2189 have been dropped because they will go to zero in the limit. The question is therefore reduced to whether the sum of the second two terms in the preceding expression is greater than 0. Note that, since we are dealing with the initial period, K0190 is independent of v191, and therefore the relationship we wish to test for these terms may be written as,
FRR0
Recalling expression (A3), this may be written for CES production functions as,
R 1 R0 dv + ( 0 + 1)dv > 0, v v
22
so that, after dividing by dv192, the Dixit, Hammond and Hoel condition reduces to
R0 < . v 1
(A5)
Because K0193 is given and independent of v194, both C0195 and R0196 are functions of v197 under the generalized Hartwick rule. A more direct attack on this problem is therefore to evaluate
In order for C0198 to vary negatively with v199 we require the latter expression to be less than 0, or, again employing expression (A3) and dividing by FR 200, 0
1 1 R0 ( R0 + dv) < 1. R0 v
R0 < . v 1
(A6)
For the case >1202, expression (A6) is less restrictive than (A5) because its right-hand side is positive, while that of (A5) is negative.
23
References Dasgupta, P., and Heal, G., 1979, Economic Theory and Exhaustible Resources, Cambridge. Dasgupta, S., and Mitra, T., 1983, Intergenerational Equity and Efficient Allocation of Exhaustible Resources, International Economic Review, Vol. 24, No. 1, 133-53. Dixit, A., Hammond, P, and Hoel, M., 1980, On Hartwick's Rule for Regular Maximin Paths of Capital Accumulation and Resource Depletion, Review of Economic Studies, Vol. XLVII, 551-6. Hartwick, J.M, 1977, Intergenerational Equity and the Investing of Rents from Exhaustible Resources, American Economic Review, 67, No. 5, 972-4. Hartwick, J.M., 1978, Substitution Among Exhaustible Resources and Intergenerational Equity, Review of Economic Studies, Vol. XLV(2), No 140, 347-54. Nordhaus, W.D., 1992, Is Growth Sustainable? Reflections of the Concept of Sustainable Economic Growth, International Economic Association, (mimeo). Pearce, D.W., Markandya, A., and Barbier, E., 1989, Blueprint for a Green Economy, London: Earthscan. Pezzey, J., 1989, Economic Analysis of Sustainable Growth and Sustainable Development, Environment Dept. Working Paper No. 15, The World Bank. Ramsey, F., 1928, A Mathematical Survey of Saving, Economic Journal, 38, 543-59. Solow, R.M., 1974, Intergenerational Equity and Exhaustible Resources, Review of Economic Studies Symposium. Solow, R.M., 1986, On the Intergenerational Allocation of Natural Resources, Scandinavian Journal of Economics, 88(1), 141-49.
24