You are on page 1of 22

An Industrial Economy’s Response To Declining Energy Resources

David Comerford∗† 5th April, 2012 - preliminary, comments appreciated

Abstract This paper explores how a decline in the availability of easy to exploit energy resources feeds into economy-wide prices that affect the profitability of energy production. A model economy is constructed in which energy is produced with capital services, which are themselves produced ultimately with energy inputs. In such an economy, removing the easy resources causes energy prices to rise, but it also causes the price of capital services to rise. After an energy quality shock, more marginal energy resources will only be exploited as a substitute for the lost easy resources if the energy price to capital services price ratio rises. It is found that there are conditions related to the returns to scale in the capital services sector which cause the energy price to capital services price ratio to fall as easy energy resources become unavailable. As the price ratio falls, the profitability of exploiting marginal resources falls, these resources are abandoned, and the economy can collapse. JEL Classification: Q30, Q43; Keywords: Non-Renewable Resources, Energy, General Equilibrium, Returns to Scale;

∗ Corresponding author: David Comerford, School of Economics, University of Edinburgh, 31 Buccleuch Place, Edinburgh, EH8 9LN, Scotland, UK; Email: † I would like to thank the Scottish Institute for Research in Economics (SIRE) for financial support via a conference participation grant to present this work.




The purpose of this paper is to explore the economy’s response to being faced with only lower quality energy resources, with a view to characterising the situations under which this is problematic. “Problematic” in this context is interpreted as an inability to use all available resoures, or to a price signal that makes the use of the resources less efficient, and does not incentivise the use of alternatives, as the resources becomes scarcer. Capital assets are used to supply energy to the economy, but the manufacture of capital assets can be an energy intensive process. If higher quality energy resources are no longer available, and the use of lower quality resources is to be expanded by applying more capital inputs to their exploitation, then the relative energy (output) to capital (input) price movement will have to be consistent with this expansion. A three good economy is therefore of the minimum complexity required to investigate this issue, and given the feedbacks between available energy resources and economy wide prices, a general equilibrium approach is appropriate1 . The theoretical natural resource economics literature usually has fairly sanguine conclusions with regards to the exhaustion of non-renewable resources. Since Hotelling’s (1931) [10] article, the first principle is that resource prices rise at a compound rate related to the rate that can be earned by extraction and investing the financial proceeds. This rising price ensures that resources that are initially unprofitable to exploit eventually become profitable, and that there are incentives both to economise on the use of the resources and to develop alternatives. This foundation to the literature has subsequently been built upon, e.g. Holland (2008) [9] describes models of resource extraction that generate a peak in the extraction rate during the extraction period using a partial equilibrium approach (since interest rates and backstop prices do not depend upon energy used in the aggregate economy); Dasgupta & Heal (1974) [8] extend The Hotelling’s framework to a general equilibrium setting without changing the conclusion that non-renewable resource prices rise without bound as they become more scarce; and Aghion and Howitt (1998) [2] describe a two sector general equilibrium model in which growth can be sustained despite declining availability of non-renewable natural resources, that are essential for production, through investment in intellectual capital. In all these cases, and in general throughout the literature, natural resource scarcity is accompanied by a rise in their price2 . Holland claims that price movements will be smoothly increasing because “oil is virtually costless to store in its natural reservoir ... even completely myopic firms without secure property rights would wait to produce from these [higher cost] deposits until the price were high enough to cover the extraction costs”. This statement reveals, I believe, a possible shortcoming in this approach: yes, the resources can be left in the ground at zero cost, but there is no guarantee that the capital assets that are used to extract these resources will be reasonably priced in future. It may be the case that as resources become scarce, capital prices rise faster than energy prices, and so lower quality energy resources can never be profitably exploited. The analysis in this paper reveals a crucial role for economies of scale in determining how the economy responds to declining availability of energy resources: Constant returns, or a low level of increasing returns, in the capital services sector are consistent with energy scarcity causing energy prices to rise faster than the prices of capital services, and so for the economy to profitably expand into lower quality energy resources; However, if the degree of returns to scale in the capital services sector is strong enough then the reduction in this sector’s productivity,
1 As Ayres (2001) [1] shows however, prices in a three sector economy do not necessary reflect marginal contribution and can respond in complicated ways to changing factor abundances - so it need not be the commodity that is becoming scarce whose price responds most strongly to its own rising scarcity. 2 A possible exception to this is Gaitan, Tol & Yetkiner (2004) [19] which finds that the resource to capital price ratio tends to a constant, but this result may be driven by their use of a constant savings ratio rather than being the result of some optimal plan.


rather it would not have been possible at all with the technology available. As the scope for adding to production from new geographical areas declines. even allowing for technological advances. The paper introduces the three sectors of the economy in sections 2 .caused by the restriction in its factor inputs. rather than predominantly through technological advances as applied to existing sources. and this data suggests that. and some dispute about whether the evidence of the 20th century is consistent with this belief. However. given that there is some expectation (by at least some people) that technology will win out over depletion. The intellectual experiment that is then explored is how these economies compare with each other. The contribution that this paper makes is to draw attention to the possibility that price movements in response to scarcity may not be favourable to bringing on substitutes or for using capital intensive but energy efficient alternatives. This rising price is at odds with the observed price history of non-renewable resources. The explanation for this is usually technological advances. This increase in the cost of supplying the same amount of energy comes despite improvements in technology over the period. Section 6 presents some illustrative results from this basic model and section 7 does likewise with a generalised model. and section 9 concludes. Barnett & Morse (1963) [12]. However. boosts the price of capital services by more than the price of energy. It is in this sense that EROI can be said to be a technologically adjusted index of the cost of obtaining these resources. and there is some dispute about which of these effects is “winning”. the suggestion is that the era of rising production could soon end.4 and makes some basic propositions about the economy in section 5. and Simon (1996) [15]. Extracting deep water oil in 1930 would not have cost an extra 6% over the oil that was being extracted at that time. So for example oil and gas from 1930 had an EROI of (greater than) 100 : 1 and so obtaining 100boe (barrels of oil equivalent) required spending energy (including the energy embodied in the capital used to extract the energy) that contained ∼ 1boe so that gross energy production would have had to be ∼ 101boe to supply the final economy with this 100boe. Hamilton (2011) [7] details the history of global crude oil production over the last century and a half and finds that the production increases have been achieved mainly through the exploitation of new geographic areas. By 2005 oil and gas EROI was ∼ 15 : 1 and supplying the final economy with 100boe would have required gross energy production of ∼ 107boe.g. Full proofs of section 5’s propositions are given in the appendix. A set of data that is broadly consistent with Hamilton’s interpretation is the energy return on energy invested (EROI) for fossil fuels over the past century (see figure 1). a possibility which is not considered in the existing literature. 2 The Energy Sector The Hotelling model takes the incentives of an energy market participant as given. EROI can be considered as a technologically adjusted index of the cost of obtaining energy resources. in the model presented in this paper I look at economies in which technological progress and depletion are exactly offset. and eventually the economy can collapse. the resources that we are extracting are becoming more costly. Section 8 discusses the interaction between conclusions from these models and the incentives to innovate. given that they only differ in the availability of high quality (low cost) energy resources. have concluded from this price history that it is evidence of declining rather than increasing scarcity of energy resources. and generates a rising price as the resources are depleted. and some economists e. 3 . There are two effects going on: depletion and technological progress. The supply of energy therefore contracts rather than expands at the margin.

The agents therefore extract all the surplus and the license fee for resource j. Energy resources with higher costs of production and/or low EROI tend to be more capital intensive. Fj . The production function is: Ej = Qβ − j . to a continuum (0. ∞) of potential energy firms. for a given quantity of inputs.e. by optimising over the quantity of inputs used. This gives: Qj = βpE P 4 1 1−β (2) . will be equal to the one period profits that can be made from exploiting this resource. As an illustrative example we can consider the wooden derricks used for Pennsylvanian oil production in the 19th century against the deep water drilling rigs used today in places like the Gulf of Mexico. Figure 2 also shows that we are a long way from any limits in the availability of fossil fuel resources. Ej using capital services Qj in a decreasing returns to scale production function that also exhibits costs indexed by j i. Fj . taking the prices as given. πj = pE Ej − P Qj . Qj . “high j” firms are exploiting poorer quality energy resources than “low j” firms and so. or we can compare the pick and shovels used for easily accessible coal seams. The resulting energy market is competitive (i. at the start of each period. price taking) with a continuum of differentiated firms. Energy resources are owned by the agents in the economy who auction the right to exploit these resources for one period. where pE is the unit energy price and P is the unit capital services price. j β ∈ (0. 1) (1) Once they have paid the license fee. each producing homogenous output. notwithstanding any efforts on our part to leave some resources unused because of climate change concerns.e. The model presented here effectively takes this as its definition of energy quality. they produce a lower quantity.Figure 1: Figure from Murphy and Hall (2010) [5] The availability of low quality (high cost) energy resources is taken to be infinite. to the machinery required for mountain top removal in the Appalachian Mountains. but is increasingly in lower quality deposits. firms maximise profits. j. This is broadly consistent with the evidence in that the remaining fossil fuel resource is massive. Ej of outputs.

e. there is free entry in the energy sector and firms continue to enter. r] where e ≥ 0 is the parameter representing the highest quality energy resources available. r QE E Π = e r Qj dj = Ej dj = e r βpE P βpE P 1 1−β (r − e) 1 (r − e) − (r2 − e2 ) 2 r (4) (5) (6) = = e β 1−β πj dj = pE E − P QE = e Fj dj = F The energy sector uses capital inputs. and its output responds endogenously to the relationship between output energy prices and input capital prices. 5 .Figure 2: Adapted from Brandt & Farrell (2007) [3] by Murphy (2011) [13]. This is independent of j i. If it is optimal to exploit a particular resource. High quality resources are those which require low capital inputs per unit of energy produced whereas low quality resources require higher capital inputs per unit of energy produced. however these unlimited resources will be of increasingly poor quality.e.e. We define j on [e. whilst r is an endogenous variable that is defined by πr = 0 i. This gives: r = Qβ (1 − β) = j βpE P β 1−β (1 − β) (3) The total inputs and outputs from the energy sector are calculated by summing over the firms from e to r i. then it is optimal to exploit every resource of higher quality. until the marginal firm makes zero profits and the agents can no longer extract any surplus. Shows resources with their production cost. all energy firms use the same quantity of inputs. and pay a positive license fee. Proven reserves are dark bands on left. There is no limit imposed upon energy availability. The value of a firm is zero however since Fj = πj . Therefore profits and energy output are both decreasing in j. uncertain resources are lighter bands on right.

firm level. but agree that if we model the aggregate economy as a representative firm then increasing returns to scale are appropriate.g. Hall (1989) [6] explains the correlation of factor productivity with exogenous demand shocks using increasing returns and finds that increasing returns are particularly evident in the aggregate economy and in manufacturing sectors. ψ≥1 (7) 3 i. ψ ∈ (1. the contraction of the automobile and aerospace sectors could affect the manufacturing supply chain in such a way as to drastically increase costs/decrease productivity in the sector that manufactures equipment for deep oil drilling. and so it would not exist in cheap oil’s absence. Perhaps if there was no cheap oil available.25. This in turn may mean that. The comparative static analysed is the steady state of an economy with a particular level of highest quality available energy resource. In this section we present a general returns to scale capital services sector which converts the economy’s capital stock into capital services using the production function: Q = θK ψ . Caballero & Lyons (1989) [11] split the returns to scale evident in the aggregate economy into internal. The interaction between possible increasing returns to scale and resource constraints has not been considered. and increasing returns to scale are ubiquitous in new-trade models of intra-industry trade4 . We can imagine that this is the same economy after an energy quality shock and consequent resetting of beliefs (i. in the notation of equation 7. e. business cycle models may use increasing returns as a partial explanations for size of fluctuations. The agents assume that energy quality. Basu & Fernald (1997) [4] explain similar data as [11] as a reallocation effect towards more efficient firms rather than any real increasing returns. some endogenous growth models rely on increasing returns to scale.e. it is easy to imagine that they may be important: for example perhaps the ability and profitability of deep oil drilling is only possible because there is a full manufacturing supply chain that is predicated on the existence of automobile and aerospace industries. 4 In the empirical work of Mohler & Seitz [14]. without cheap oil sustaining the automobile and aerospace sectors. which corresponds to a returns to scale parameter. 1. which is also subject to diseconomies of scale caused by declining energy resource quality).5. 3 Returns To Scale In The Capital Services Sector There is some evidence to suggest that manufacturing industries behave as if they are subject to increasing returns to scale.and so the available high quality resources are always exploited. economic geography models with agglomeration effects. that technological progress and depletion are equal and offsetting. However.3 (ignoring the fact that capital services are only a subset of the whole economy. is a constant. Their best estimate of the degree of scale economies in the US is that a sector which increases its inputs by 10% will see an increase in output of 8%. If this energy quality shock leads to a rise in the ratio of energy prices to capital prices then the usage of lower quality resources will expand to (partially) offset the loss of the high quality resources. e. and positive external returns to scale. at the new lower highest quality available energy resource level). Models with increasing returns are widespread e.5) 6 . This is equivalent to the agents believing that the forces of depletion and technological progress are equal and opposite.e. this translates as ψ ∼ 1. the deep water drilling sector is unprofitable. This can be compared to the equivalent steady state of an economy with a lower highest quality available energy resource. Exploitation of lower quality resources is an increasing function of the energy price to capital price ratio. but if the whole economy increases its inputs by 10% then output will rise by 13%3 . the elasticity of substitution in the Dixit-Stiglitz production of European economies is found to lie in the range 3 . constant or decreasing returns to scale.

and external economies due to the benefits of product variety. θ is a parameter that will be used to normalise the output from this sector such that at a particular capital stock and energy resource availability. L. (with effectiveness/technology parameter A). while the price that this industry supplies capital services for is P . Kt+1 = Kt (1 − δ) + sYt (15) To complete the model economy we now only have to consider the accounting identities linking the sectors. ψ. w pE P such that Y = Y E α QF γ = (1 − α − γ) AL AL L E α−1 QF γ Y = α =α AL AL E E α QF γ−1 Y = γ =γ AL AL QF = wL + pE E + P QF (1 − α − γ)A (11) (12) (13) (14) Capital accumulation dynamics are introduced in a very simple way: a constant savings rate. 4 The final goods market and the whole economy The final goods market in this economy is a perfectly competitive. and so if a rental rate for capital.6. ρ. Capital services are fully utilised and the agents receive labour income. capital rents and energy sector license fees. Dixit-Stiglitz production is a conflation of internal economies due to fixed costs. ρ = P θK ψ−1 (8) (9) Section 7 is used to present a specific microfounded form for the returns to scale in the capital services sector using Dixit-Stiglitz aggregation of a monopolistically competitive sector [17].where K is the capital stock. The assumption of a constant savings rate will be generalised in Section 7. E. and used in sections 4 .e. then the following relationship holds: P Q = ρK i. energy. The more simple model presented here. is paid to the agents in this economy. and capital services QF to produce output: Y = E α Qγ (AL)1−α−γ F (10) Being perfectly competitive and exhibiting constant returns to scale means that. and depreciation rate. abstracts from the source of scale economies. δ. and ψ determines the degree of returns to scale. Q = and Y = QE + QF wL + pE E + P QF = wL + ρK + F (16) (17) 7 . (K ∗ . s.e. e∗ ). Cobb-Douglas economy which uses labour. as in the Solow model [16]. Zero profits are made in this sector. in equilibrium. factors must be paid their marginal product i. Q are the capital services supplied. which buys the output of the economy i. constant returns to scale. the capital services supplied to the economy will be independent of degree of returns to scale.e.

∞) therefore represents all possible price ratios that are associated with positive output from the energy sector. Since no stable productive economy (i. Proposition 2. ∃ψ ∗∗ = such that ∀ψ > ψ ∗∗ . h = 1+ α αβ 2β e 1 γ γ (1 − β)(1 + β + 2 )z z + 2β α 1 − β 2 αβ γ 1 + β 1 + 1+β γ z+ e 1−β 1 β −1 1 1 1 1 1 c0 ψ ψ z (z + g(e)) ψ (z + h(e)) ψ ( β −1) θ ≥ 0. There are therefore no problems of interpretation with a multiplicity of equilibria. the market equilibrium exists and is unique. K ≥ 0.e. ∀e ≥ 0 (18) The variable z is a positive transformation of the energy to capital services price ratio: z= βpE P β 1−β − e 1−β z ∈ (0. The phase diagram for this extreme case of Super Strong Increasing Returns To Scale is shown in figure 3. Clearly K(z) is a bijection on z > 0. ∆K(zt ) where c1 . Y ∗ > 0) exists even when all resources are available under these parameters. For any given capital stock. The equilibrium price ratio as a function of capital stock is given by: θK ψ = Q = so K where c0 . f ≤ g ≤ h. G(zt ) is monotonically decreasing and so there is only one (unstable) steady state K ∗ > 0. g ≤ h. and we are interested in how economies that can exist cope with resource restrictions. For ψ > ψ ∗∗ and e = 0. g. Phase space is characterised by the following expression: ∆K(zt ) sY (zt ) − δK(zt ) 1 α+γ γ( β −1) 1 α+γ γ γ 2β e e α+γ = s(AL)1−α−γ (1 − β 2 )α+γ zt zt + zt + 2 αβ 1+β 1−β 1−β 1 1 1 1 1 c0 ψ ψ −δ z (z + g(e)) ψ (z + h(e)) ψ ( β −1) θ 1 1 1 1 1 1 c0 ψ ψ α+γ z (z + g(e)) ψ (z + h(e)) ψ ( β −1) = s(AL)1−α−γ c1 zt (zt + f (e))α+γ (zt + h(e))γ( β −1) − δ θ = = α+γ s(AL)1−α−γ c1 zt (zt + f (e))α+γ (zt + h(e))γ( β −1) 1 − 1 1 ψ i. g . f and H(0) = = δ c0 (AL)α+γ−1 sc1 θ H(zt )[1 − G(zt )] f > 0. 1 2β 1+β α+γ 1 ztψ −α−γ zt + g(e) zt + f (e) α+γ (zt + g(e)) ψ −α−γ (zt + h(e))( ψ −γ)( β −1) (19) 1 1 1 ≥ 0. limz→0 ∆K(z) is negative and limz→∞ ∆K(z) is positive.e. we do not consider this case further. As we shall see though. Proposition 1. h > 0. ∀e ≥ 0 0 .5 Some propositions on this model economy Full proofs of all the propositions in this section are given in Appendix 1. there are multiple steady states. K = 0 is a steady state. 8 .H > 0 ⇒ H > 0 ∀zt > 0 Proposition 3.

is shown in figure 4. For 1 ≤ ψ < ψ ∗∗ and e = 0. z ∗ > 0 with e = 0. Super Strong IRS have two steady states: a stable state at K = 0 and a higher unstable state. G(zt ) is monotonically increasing and so there is a unique stable steady state K ∗ > 0. Proposition 4. limz→∞ ∆K(z) negative. We can find an analytic expression for the relative price at the stable steady state. limz→0 ∆K(z) is positive (and so K = 0 is unstable). z∗ = sc1 θ (AL)1−α−γ δ c0 9 1 ψ 1 ( 1 −γ)( 1 +1)−2α ψ β (20) . For 1 ≤ ψ < ψ ∗∗ . economies with 1 ≤ ψ < ψ ∗∗ have two steady states: an unstable state at K = 0 and a stable state at K ∗ > 0. economies with ψ > ψ ∗∗ . Figure 4: At e = 0.Figure 3: At e = 0. The phase diagram for e = 0 and Returns To Scale that are not Super Strong. i.e. The steady state capital stock is then K(z ∗ ).

then with e = 0 we can evaluate the ∆K(z) function and see that the result is akin to figure 4. However. or Weak Increasing. economies with ψ ∗ < ψ < ψ ∗∗ . for 1 ≤ ψ < ψ ∗ . i. With 1 ≤ ψ < ψ ∗ the phase diagram in figure 4 holds for e > 0 (though we cannot strictly prove monotonicity for G(z). This is unlike Constant. always exists. e > 0) is negative. Proposition 6. the steady state changes discontinuously from K ∗ > 0 to K ∗ = 0. z ∗ > 0 ⇐⇒ K(z ∗ ) > 0. For 1 ≤ ψ < ψ ∗ .e. there is a point in the e parameter space at which the economy collapses as the K ∗ > 0 steady state ceases to exist. e > 0) is negative. Strong IRS. Significantly. the phase diagram changes to that shown in figure 5. Then as we raise e. ∃ψ ∗ = α+γ < ψ ∗∗ such that ∀ψ ∈ (ψ ∗ . i. Once we raise e past a critical value. we can experimentally construct a particular economy with Strong Increasing Returns to Scale. with a globally stable K = 0 steady state. i. All we know here is that for e > 0. ∞). we cannot prove the existence of a non-zero steady state for Strong Increasing Returns to Scale. a stable steady state K ∗ > 0. limz→0 ∆K(z) is positive and limz→∞ ∆K(z) negative. may have three steady states: stable states at K = 0 and K ∗ > 0 with an unstable state between. We can describe the transition from figure 5 to figure 6 as a collapse because there is a discontinuity in the steady state that the economy can reach. ψ ∗∗ ) and with e > 0. Returns 10 .e. Proof.e. Figure 5: At e > 0. limz→0 ∆K(z) is negative (and so K = 0 is stable). For 1 ≤ ψ < ψ ∗ and e > 0. phase space looks like figure 6. limzt →0 ∆K(zt . Indeed. ∆K(z ∗ + ) < 0 for very small . and general e ≥ 0. Previous analysis has shown that.1 Proposition 5. for ψ ∗ < ψ < ψ ∗∗ . Further ∆K(z) is a continuous function of z with no singularities over z ∈ (0. the signs at very low and very high z have been established). Therefore there must be some z ∗ ∈ (0. And as we raise e further. ∞) for which ∆K(z ∗ ) = 0 and ∆K(z ∗ − ) > 0. Strong Increasing Returns To Scale. Further. and that limzt →∞ ∆K(zt . limz→0 ∆K(z) is positive (and so K = 0 is unstable). Therefore it is possible that the phase diagram for Strong IRS looks like that shown in figure 6. Therefore there is a stable steady state for the economy with 1 ≤ ψ < ψ ∗ and general e ≥ 0. we observe that the experimentally constructed phase diagram looks like figure 5.

e. economies with ψ ∗ < ψ < ψ ∗∗ . These results are somewhat unsatisfactory in that there is such a clear distinction between economies that will collapse under resource restrictions and economies that will not: from the outset at low e. These comparative statics show how the steady state of the economy varies with ψ and e . r. Figure 6: At e > 0. then a rising r(e) curve indicates that the use of lower quality resources substitutes for the high quality resources that are no longer available to the Scale in which the economy exists at some positive level of production. varies with the index of the highest quality available energy firm. Figure 7 shows how the index of the marginal. Let ψ = 1 ⇒ θ = 1 and define θψ = (K ∗ )1−ψ where K ∗ is given by equation (20) for ψ = 1. ψ. and the results from the following section. The parameters used to generate these results. We want the economies considered to differ only in their returns to scale. may only have one stable steady state at K = 0.they do not show timepaths. i. Therefore we normalise the economies so that they coincide for e = 0. 11 . θ = 1. and e = 0. that are imposed. but the results generated do however illustrate the salient features of the model: it is possible that the use of low quality resources does not expand as high quality resources become unavailable. e. and this occurs because of price incentives. but for their results to be comparable. If we imagine that the economies along a specific ψ line are connected by a series of depletion shocks (as discussed in Section 2). zero profit energy firm. 6 Basic Model Results In this section we present illustrative results from the basic model developed in the preceding sections. A falling r(e) curve indicates that marginal resources are abandoned as high quality resources cease to be available. Strong IRS. irrespective of the severity of the resource restrictions. e. the strong IRS economy starts abandoning marginal resources as high quality resources become unavailable. are listed in Appendix 2. This is not suggestive of the real world.

If the true state of the world is that depletion will outpace technological progress. and Strong IRS with ψ > ψ ∗ . For Strong IRS economies.Figure 7: r vs e for three economies: CRS with ψ = 1. so why would any (risk neutral) agent seek to save these resources against future scarcity? In other words. Weak IRS with 1 < ψ < ψ ∗ . In generalising the model we lose the ability to interrogate it analytically and since we’re just producing a numerical result we may as well build in microfoundations for the capital services sector. this is a poor model in the sense that agents myopia is sub-optimal precisely in the cases that it doesn’t tell us anything interesting. where economies can cease to function because of lack of resource availability. these profits are falling with e.they cannot gain by hoarding because the price of extracting the resources is going to rise by more than the price that they can sell them for. then is this belief sub-optimal? In the above results. at least for e greater than some e∗ . 12 . then for this to hold there must be some range in which firms/licensing agents with resources j > j ∗ have no incentive to hoard because their profits are falling in e. the agents myopia is optimal . but we have not proved that this is the case in general. the profits at the energy firm exploiting the resources at j = 0 (and hence the license fees that the agents can extract) are rising with e only for the CRS or Weak IRS economies. In the interesting cases. r (e) < 0. Figure 8 illustrates this general case schematically by showing an economy where π0 is rising with e but which r (e) < 0. We can experimentally verify that the propositions of Section 5 appear to hold. We now examine the assumption that the agents believe depletion and technological advances in reducing costs in the energy sector exactly offset. The generalisations that we are adding are to endogenise savings and to have two input factors for the capital services sector: capital stock and energy. Since r is defined as a zero profit condition. This is a general result in that the defining feature of the economies that will collapse is that. 7 A generalised model and the impact of policy In this section we generalise the model to produce more interesting results and then show that the collapse of a Strong IRS is due to prices rather than to any fundamental limits.

and have the agents chose the amount of energy used today to balance present consumption against the value of future consumption. Et ). The aggregation sector has production function and profits as follows: n Q = 0 qi PQ − σ−1 σ di σ σ−1 . the rental rate of capital. Weitzman (2009) [18] shows that infinite marginal utilities at zero consumption and the possibility of catastrophe do not mix. Savings are endogenised by giving the risk neutral agents some rate of time preference and assuming full depreciation.Figure 8: Schematic showing that if r (e) < 0 then there must be some resources for which there is no incentive to hoard. now we have the price of capital constant and the steady state savings rate. will be a constant equal to the inverse of the agents’ discount factor. At steady state.and we should not want to exclude this possibility without good reason. We could postulate an equation of motion et+1 = f (et . fEt > 0. It may be a more controversial assumption in a model in which agents rationally anticipate depletion outpacing technological progress. disaggregated. ρ. 13 . All this paper seeks to achieve is to signpost that extreme conditions may lie in this direction. σ>1 (21) (22) n πA = pi qi di = 0 0 5 The assumption of risk-neutrality is uncontroversial here since the agents expect no change in the value of the parameter e. varying (increasing) with Y the parameter e5 . However in order to generate catastrophic/collapse outcomes. The capital services sector is generalised by splitting it into two. Rather than the price of capital varying and a constant proportion of output being saved as in the model of the previous sections. s = K . The analysis of Section 6 on the incentives to hoard suggests that even a forward looking model with risk neutral agents and Strong IRS may still collapse. We do not have good models of how people behave under such extreme conditions and it is not the purpose of this paper to propose such a model. A perfectly competitive aggregation sector buys the output of an imperfectly competitive. capital services sector. it would be necessary that the agents not make these choices based on CRRA utility optimisation .

The demand schedule that each monopolist faces. This can be seen more clearly by zooming into figure 9: 14 . Weak IRS.The disaggregated capital services sector consists of measure n (endogenous) firms each producing a differentiated good with some monopoly pricing power. gives: EQ P = = η ρ K 1 − η pE ησ 1 1 η ρ 1−σ σ ρ (σf ) σ−1 K 1−σ σ − 1 φ(1 − η) 1 − η pE ησ σ σ η ρ σ−1 (σ − 1)φf (σf ) 1−σ K σ−1 1 − η pE (26) (27) (28) Q = The only other change from the model presented previously is that the total output of the energy sector now has to be split across the final goods and capital services sectors. We now initially have the (infinitely) abundant low quality resources providing a substitute for unavailable high quality resources under strong IRS. Solving this model. Defining f = f (σ) allows us to normalise the economies with different σ’s so that they all coincide for e = 0. assuming free entry drives the profits of each monopolist to zero. and ultimately the economy collapses. Simulating this model (with the parameters detailed in Appendix 2) gives: Figure 9: r vs e for three economies: CRS. and Strong IRS. and their production and profit functions are: qi qi πi = Qp−σ P σ i = η 1−η φ(Ei Ki (23) − f) (24) (25) = pi qi − pE Ei − ρKi Where f is fixed costs. Eventually as e continues to rise. the r(e) curve has a turning point and as e rises further the economy starts to abandon the marginal resources despite their abundance. EF + EQ .

where e∗ is the point at which this IRS economy collapses. in that our experience of the real world does not lead us to automatically allocate either of the CRS/Weak IRS or Strong IRS lines to an ‘implausible’ category. then the pareto-optimal tax rate is zero for CRS and is increasing in the degree of scale economies. the more at risk of collapse the economy is from the interaction of increasing returns and scarce factors.3 . This is done by allocating factors to industries rather than by using the price mechanism for allocation. the more amenable this situation is to policy intervention. 15 . All technological conditions are satisfied and it is a steady state in the sense that consumption is the same as that achieved by the CRS economy and savings are sufficient to replace the depreciating capital stock. If we were to impose lump sum taxes on the agents and use the proceeds to subsidise energy production. Is the collapse of the economy due to fundamental limits in how the technologies of the sectors can exploit the resources available? No. and the market allocation cannot be improved upon .Figure 10: As Figure 9 but zoomed in. and as an example of this we reproduce figure 9 with the same Strong IRS economy operating at a point e > e∗. the degree of returns to scale needed for Strong IRS and collapse is much lower in the generalised model of this section than in the simple model of the previous sections6 . 6 Indeed the Strong IRS line in figures 9 and 10 was generated with a returns to scale of 1.although no calibration has been performed. The economy does not collapse under CRS.this follows from The First Welfare Theorem (all sectors are price taking under CRS). This illustration of how allocating the factors under a command economy system can outperform allocation using prices is an example a more general phenomenon: this collapse is not due to fundamental resource limits but is rather due to a failure of the price mechanism to allocate these resources efficiently when there are returns to scale in the economy and factors feed-back upon themselves by being both inputs and outputs. As described in Appendix 2. This is a much more plausible result than that presented in Section 6.which is the number estimated by Caballero & Lyons [11] . However.

Rising r occurs because of a rising energy to capital services price. as energy became scarce and expensive there was an increasing incentive to develop this technology. The switch price will eventually be reached for the CRS and Weak IRS economies and they will ultimately use the energy efficient technology. further declines in high quality energy availability could see the switch price being reached again on the way down i. rising e always causes rising r in the CRS and Weak IRS economies. Considering only the model of Section 7. For expositional 16 . 8 Innovation When we imagine technological solutions to energy scarcity. There would be some price ratio at which these technologies used energy and capital services in the same ratio to produce the same output level. A real world example of this could be modern cars with computer optimised engines.e. The switch price may or may not be reached under Strong IRS. it may then abandon it. and this price ratio would be the price ratio that the economy switched from one technology to the other. In Aghion & Howitt’s model. the Strong IRS economy may choose never to take up the energy efficient technology. then advanced energy efficient products may not be available.we can well imagine that productivity in advanced sectors depends on sufficient scale. We could suppose that some nondepletable backstop was available at some energy to capital services price ratio. and even if the economy does adopt it.Figure 11: As Figure 9 with a command economy. we could formulate two alternative production technologies for final goods: a technology that used energy very efficiently by applying lots of capital services. and a technology that used capital services efficiently by applying lots of energy. Specifically. This is the situation that Aghion & Howitt were abstracting from. in the model presented in this paper I suggest that. if we live in a world of strong IRS. and causes rising r in the Strong IRS economy initially. This is intuitive . as opposed to simpler vehicles that use petrol inputs much less efficiently. but even supposing that it is. then energy scarcity may not motivate us to use this advanced energy efficient technology. However. and if scale is hit hard enough by a shortage of energy. because the price of the high-tech computer optimised capital goods could rise by more than the price of energy. we often think of high technology goods that use energy more efficiently. The same issue arises for a putative backstop technology.

Further.and the interaction does not mitigate the problem. as estimated as occurring in. Future research must focus on testing these conclusions in a model in which the energy shocks are rationally anticipated by the agents. the backstop technology will eventually be deployed. Again the productivity of the sectors that can produce the solar technology depends upon the scale at which it operates. renewables feed-in-tariffs. Subsidies can support the scale of industry so that innovations or technologies are within the reach of a Strong IRS economy. It appears that the solar backstop is feasible but the energy price is too low to justify its deployment. but the costs are positively related to the capital services price. This is therefore (theoretical) support (though not necessarily support in any specific case. The pricing is such that despite the rise in energy prices. which may create an industry of sufficient scale to be profitable. And we have shown that increasing returns to scale. we are no closer to profitably deploying the backstop technology. However capital sectors across the economy contract and productivity falls. the more it is amenable to policy intervention . under CRS or Weak IRS. whereas they may be out of reach without policy interventions. At low levels of scarcity there is large demand for semiconductor technology so this sector is large and productive. e. and at no point was it profitable to deploy the backstop technology. Energy scarcity rises and energy prices rise. 17 . then it will eventually be optimal to undertake the innovation effort under CRS or Weak IRS. If the benefits to innovating are positively related to the energy price. The economy eventually collapses for lack of energy. the more that this phenomena is a real problem. 9 Conclusion We have shown that it is possible that the price movements caused by declining energy resources may not be conducive to the exploitation of more marginal resources.which does allow society to mitigate the problem through activist policy. or for our real world economy) for subsidies. and on calibrating such models to uncover whether they suggest we live in a world of Strong Increasing Returns to Scale or not. However. As in Section 7 this problem is amenable to policy intervention. and often assumed for.g. we have discussed how this phenomena may impinge upon innovative or technological solutions to future energy shortages . is a sufficient condition for this phenomena to be manifest.purposes let us suppose it is large scale deployment of solar panels in deserts. The description here only applies to the Strong IRS economy. In general. in contrast with the basic Hotellings model and almost all of the non-renewable natural resources and energy literature. It may be the case that it is never be optimal to undertake the innovation effort under Strong IRS. industrial economies. this story applies to any innovation effort that may allow an economy to grow or continue at the same level under resource restrictions.

Oil prices. [9] S Holland. 1996.ub. Journal of Political Economy. 2004. University of Hamburg Working Paper. [6] R Hall. 1998. and economic growth. 1989. 2010. Monopolistic competition and optimum product diversity. Returns to scale in u. NBER Working Paper. 2011. [12] H Barnett C Morse. Energy. Modeling peak oil. [4] S Basu J Fernald. The hotellings rule revisited in a dynamic general equilibrium model. 2009. [5] D Murphy C Hall. The minimum complexity of endogenous growth models: the role of physical resource flows. 1931. 1956. [3] A Brandt A Farrell. 1989. [19] B Gaitan R Tol I Yetkiner. The Journal of Political Economy. Annals of The New York Academy of Sciences. Howitt P. 2007. A contribution to the theory of economic growth. 2011. Review of Economic Studies. Scarcity and Growth. exhaustible resources. manufacturing. [7] J Hamilton. 1974. The role of external economies in 1997. physics. [2] Aghion P.s. The economics of exhaustible resources. On modeling and interpreting the economics of catastrophic climate change. The Energy Journal. Climatic Change. The gains from variety in the european union. Scraping the bottom of the barrel: greenhouse gas emission consequences of a transition to low-quality and synthetic petroleum resources. NBER Working Paper. [16] R Solow. 2010.s. Prepared for Handbook of Energy and Climate Change. [17] A Dixit J Stiglitz. Munich Discussion Paper http://epub.eroi or energy return on (energy) invested. Garcaia-Peanalosa C. [11] R Caballero R Lyons.ucsd. [15] J L Simon. The Ultimate Resource Princeton University Press. [14] L Mohler M Seitz. American Economic Review.References [1] R Ayres. Year in review . 1963. [10] H Hotelling. Brant-Collett M.uni-muenchen. 18 . [18] M Weitzman. The Review of Economics and Statistics. 1977. Invariance properties of solow’s productivity residual. [8] P Dasgupta G Heal. Endogenous Growth Theory. 2001. [13] T Murphy. Johns Hopkins Univ Pr. Peak oil perspective. production: Estimates and implications. 2008. The Quarterly Journal of Economics. The optimal depletion of exhaustible resources.

Proof. = = = = = Q − QF Y γ QF Y α E β αβ QF 1−β e − γ E 1−β 1−β γ e Q− E(z + ) β αβ 1−β z→∞ Q 0 QD − QS E S −Q z→∞ 0 ∀z > 0 Clearly QS → −∞ as E = < XSD = = with XSD(0) XSD → ∞ as dXSD and > dz 19 . Given a relative price level. K ≥ 0. the total capital services available to the economy. For any given capital stock. (1 − β)z(z + dQD E > 0. we have the energy output as a function of z: E Clearly E(0) = 0 = . is also given. 2β e 1 (1 − β 2 )z z + 2 1+β 1−β dE > 0. z. z so QS E and QS (0) E dQS E with dz So. ∀z > 0 dz z→∞ and E → ∞ as The supply of capital services to the energy industry can be determined given the total capital services available and the demand from the final goods sector: QS E P pE i. Q = θK ψ . dz z→∞ 1 e )β 1−β ∀z > 0 and QD → ∞ as E and so we have demand monotone in relative prices.e. Given K. the market equilibrium exists and is unique.Appendix 1: Proofs of propositions Proposition 1. It is then useful to perform a positive transformation on the energy price to capital services price ratio: β βpE 1−β e z= − P 1−β Now consider the demand for capital services from the energy industry as a function of z: QD E Clearly QD (0) = 0 E = . and assuming that the energy industry’s demand for capital is satisfied.

Proof. Since K(z) is a bijection. K = 0 is stable ∀e ∈ [0. z = 0 ⇒ K = 0. Proposition 4. e) = 0. We have already seen that ψ > ψ ∗∗ ⇒ K = 0 stable. Proof. At K = 0 steady state. Further: ∆K(zt . ∆K(z) depend upon the sign of the term in brackets. Here we consider only 1 < ψ < ψ ∗∗ and e = 0. Proposition 3. there is a root for XSD and it is unique. For 1 ≤ ψ < ψ ∗∗ and e = 0. Clearly K(0) = 0. ∞) and. For ψ > ψ ∗∗ and e = 0. there is a unique K ∗ > 0 steady state that is unstable. so the sign of.e. This is a steady state since Kt+1 = (1−δ)Kt +sYt = (1 − δ) × 0 + s × 0 = 0. Further. limz→∞ G(z) = 0. ∀e ≥ 0 since limz→0 G(z) = 0 and limz→∞ G(z) = ∞. It remains to show that zeq (K = 0. e = 0) = s(AL)1−α−γ c1 zt 1 2α+γ( β +1) 1− δ c0 (AL)α+γ−1 sc1 θ 1 ψ zt ψ 1 ( 1 −γ)( β +1)−2α = H(z. e = 0)] 20 . by continuity and monotonicity. Proposition 2. ∞) for all parameter values when the parameter e = 0. We have: ∆K(zt ) α+γ = s(AL)1−α−γ c1 zt (zt + f (e))α+γ (zt + h(e))γ( β −1) 1 1 1 1 1 δ c0 ψ ψ −α−γ zt + g(e) α+γ 1− (AL)α+γ−1 zt (zt + g(e)) ψ −α−γ (zt + h(e))( ψ −γ)( β −1) sc1 θ zt + f (e) = H(zt )[1 − G(zt )] 1 Taking limits with 1 ≤ ψ < ψ ∗∗ we get limz→0 ∆K(z) is positive and limz→∞ ∆K(z) is negative. Therefore. e = 0 ⇒ limz→0 G(z) = ∞. and roots of. Kt+1 > Kt ⇐⇒ zt+1 > zt i. we also have limz→0 G(z) = ∞ and limz→∞ G(z) = 0 even for e > 0 (where we cannot definitively claim monotonicity). sign(∆z) = sign(∆K). H (z) > 0. where: ψ ∗∗ = 1 2β 1+β α +γ (29) Given this parameter restriction. ∞). Have ∆K = Kt+1 − Kt = sYt − δKt = H(zt )[1 − G(zt )] and since K(z) is a bijection with K > 0. e = 0)[1 − G(z.e. at least for e = 0. We define ψ ∗∗ as the degree of increasing returns required such that G(z) is monotonically decreasing on z ∈ [0. ∃ψ ∗∗ = 1 2β 1+β α+γ such that ∀ψ > ψ ∗∗ . G(zt ) is monotonically increasing and so there is a unique stable steady state K ∗ > 0. G(zt ) is monotonically decreasing and so there is only one (unstable) steady state K ∗ > 0. H(0) = 0 and H(z) > 0. K = 0 is a steady state. limz→0 ∆K(z) is negative and limz→∞ ∆K(z) is positive. limz→∞ ∆K(z) negative. we have Q = 0. Therefore there exists a market equilibrium and it is unique. and G(z ∗ ) = 1 occurs only once for unique z ∗ because of monotonicity. limz→0 ∆K(z) is positive (and so K = 0 is unstable). If we have z = 0 then we will have QD = 0 and E γ e E = 0 so QS = Q− αβ E z + 1−β E 1−β β = 0. ∀z ∈ (0. i. we also have K = 0 ⇒ z = 0. For 1 ≤ ψ < ψ ∗∗ . Therefore we do indeed have z = 0 if K = 0 and so K = 0 is a steady state. For G (z) < 0 we need ψ > ψ ∗∗ . the proposition holds.

∃ψ ∗ = α+γ < ψ ∗∗ such that ∀ψ ∈ (ψ ∗ . Further ∆K(z) is a continuous function of z with no singularities over z ∈ (0. 21 . ∞). ψ ∗∗ ) and with e > 0. K ∗ > 0 corresponding to K(z ∗ ). e = 0) = 0. Previous analysis has shown that. For ψ > ψ ∗ . ∞) for which ∆K(z ∗ ) = 0 and ∆K(z ∗ − ) > 0. limz→0 ∆K(z) is negative (and so K = 0 is stable). stable. Therefore. K = 0 is stable. limz→0 ∆K(z) is positive (and so K = 0 is unstable). e = 0) → ∞. limz→0 ∆K(z) is positive and limz→∞ ∆K(z) negative. We have: ∆K(zt ) = α+γ s(AL)1−α−γ c1 zt (zt + f (e))α+γ (zt + h(e))γ( β −1) 1 1 1 1 1 δ c0 ψ ψ −α−γ zt + g(e) α+γ 1− (AL)α+γ−1 zt (zt + g(e)) ψ −α−γ (zt + h(e))( ψ −γ)( β −1) sc1 θ zt + f (e) = H(z)[1 − G(z)] 1 Therefore. Proof. Therefore there is a stable steady state for the economy with 1 ≤ ψ < ψ ∗ and general e ≥ 0. to limz→∞ G(z. For 1 ≤ ψ < ψ ∗ .Given the parameter restriction 1 ≤ ψ < ψ ∗∗ holds. z ∗ > 0 ⇐⇒ K(z ∗ ) > 0. Since β ∈ (0. the function G(z. 1 Proposition 5. a stable steady state K ∗ > 0. for e = 0 there is a unique. so G(z ∗ ) = 1 occurs only once for unique z ∗ . ∆K(z ∗ + ) < 0 for very small . for 1 ≤ ψ < ψ ∗ . For 1 ≤ ψ < ψ ∗ and e > 0. clearly 1 1 ψ∗ = < 2β = ψ ∗∗ α+γ α+γ 1+β Proposition 6. Proof. always exists. 1). e > 0) = α+γ s(AL)1−α−γ c1 zt f (e)α+γ h(e)γ( β −1) 1 1− δ c0 (AL)α+γ−1 sc1 θ 1 ψ 1 ztψ −α−γ g(e) f (e) α+γ g(e) ψ −α−γ h(e)( ψ −γ)( β −1) 1 1 1 1 1 The quantity in the square bracket is positive if ψ − α − γ > 0 but negative if ψ − α − γ < 0. we can take the limit as z → 0 with e > 0: zt →0 lim ∆K(zt . Therefore there must be some z ∗ ∈ (0. e = 0) is monotonically increasing in z from G(0. and general e ≥ 0. 1 Therefore ψ ∗ = α+γ and K = 0 is unstable for 1 ≤ ψ < ψ ∗ . Further.

Appendix 2: Parameters of simulated economies 22 .