Sie sind auf Seite 1von 29

Resources and Energy 12 (1990) 143-171.

North-Holland

GEOLOGICAL INFLUENCES, METAL PRICES AND


RATIONALITY

Robert D. CAIRNS*
McGill University, Montreal, P.Q. H3A I W7, Canada
UniversitP de Montrhl, Montreal, P.Q. H3C 3J7, Canada

Received May 1989, final version received September 1989

Much empirical work has been performed to test the predictions of Hotelling-style models of
non-renewable resource production. Disparate, seemingly contradictory results have emerged,
some damaging to the predictions of rational models of present value maximization. This paper
presents a theoretical model of mining which incorporates the facts that varying qualities of
material must be extracted at any time from a deposit, so that a cut-off grade must be
determined, and that the extractive sector is embedded in an economy with more than one
sector. The model reconciles many of the disparate findings from a rationalist, Hotelling
perspective. Empirical results for Canadian gold mining are consistent with the theoretical
predictions.

1. Introduction
If there is a concept that is at the heart of the economic theory of
depletable resources, it is that the path of production levels is chosen to
maximize the dis&unted value of net proceeds. This concept may be traced
to the classical economists, but was given its first modern expression by
Hotelling (1931). Solow (1974a) popularized the insight that the path of
prices and~+utputs would tend to equilibriate both the product market for
the resource and asset markets, with in-situ resources viewed as assets.
In the simpler expressions of the model, the price of the resource is
expected to rise through time, and the central determining factor of the rate
of price rise is the rate of interest. That the simpler models were inadequate,
however, was obvious given the results of Barnett and Morse (1963) who
found, among other things, declining resource prices over long periods of
time. Realistic tests of the model required other approaches or the introduc-
tion of complicating influences.
*Earlier versions of this paper were presented at Universitt Lavai, IIASA, the Montreal
Econometrics Workshop and the AERE meetings. I thank Bijan Aiagheband for research
assistance and Steve Ambler, Scott Farrow, John Gaibraith, Nguyen Manh Hung, Tom
Kompas, Jeffrey Krautkraemer, John Livernois, Barry Ma, Pierre Mohnen, Margaret Siade and
Robert Soiow for helpful comments on earlier drafts. I am grateful to FCAR, Quebec Ministry
of Education, for financial support during the preparation of this paper.

0165-0572/90/‘$3.50 0 1990, Eisevier Science Publishers B.V. (North-Holland)


144 R.D. Cairns, Geological influences, metal prices and rationality

A unique method was developed by Miller and Upton (1985). They tested
the ‘Hotelling Valuation Principle’, by which the market value of resource
firms (in this case, a sample of firms in the U.S. petroleum industry) per unit
of reserves would be approximately the sum of two terms. One of the terms
was a factor, of slightly less than one, times the current profit per unit of
production; the other, they argued, was uncorrelated with the first and was
taken to be a constant in the estimations. The results tended to confirm the
theory.
A more traditional model was tested by Stollery (1983). After correcting
for the non-competitive structure of the Canadian nickel industry, Stollery
found a significant resource rent as predicted by theory. Cairns (1986a),
however, argued that Stollery’s results were also consistent with a myopic
model of mark-up pricing by nickel firms. His own estimations for the nickel
industry [Cairns (1981)] found that the rent was negligible. Even so, this
result was not inimical to the model; it simply meant that the resource was
so vast, and costs of extraction were rising so slowly, that no rent should be
perceptible. Still, it was clear that the verification of predictions of the model
was sensitive to the estimation method.’
One of the important factors explaining the declining resource prices found
by Barnett and Morse was exploration. Pindyck (1978) showed that explo-
ration, in conjunction with ultimate scarcity, could account for a U-shaped
path of prices over time; the declining portion could coincide with the
Barnett-Morse results. Slade (1982) proved that technological change also
could result in a U-shaped price path, and tested her results for several
metals.
Frank and Babunovic (1984) confirmed the importance of technological
change, and also pointed up a further influence on pricing, namely, time lags
in adjustment of production to changes in market conditions. Their empirical
results, however, showed wide variation across the depletable resources to
which they applied their model.
It is noteworthy that interest rates and lagged interest rates appear in
Frank and Babunovic’s results, in a way that suggests that both interest rates
themselves and changes in interest rates may be important determinants of
the price path. Another type of model used to explain resource pricing tries
to capture arbitrage in asset markets. Heal and Barrow (1980) argue that
asset markets do not exhibit the full-information equilibrium assumed by
other models. Rather, they find that changes in interest rates are important
in determining metal price movements. There are second-order effects which
arise through uncertainty of future price trends of resources and alternative

‘Stollery’s estimations were carried out over the period 1952-1973, while Cairns’ were done
looking forward from 1974. Some further light may be cast by Halvorsen and Smith’s (1984)
results: they estimated that in-situ values of Canadian mineral resources fell over 80 percent
between 1956 and 1974.
R.D. Cairns, Geological influences, metal prices and rationality 145

assets. Also, interest rates themselves do not appear in the model, and the
empirical results confirm the model. Despite this confirmation, however, Heal
and Barrow (1980, p. 175) are not satisfied with the implication that prices
will remain constant if the interest rate remains constant - a fact totally at
variance with rationality in the asset market for a resource. They propose
alternative model specifications by which the interest rate itself may appear.
Their conclusion is that resource prices may be expected to exhibit move-
ments that are substantially more complicated than economic theory has
hitherto predicted.
One problem with the Heal-Barrow approach is the naive supply side of
their model, which does not adequately take into account the depletable
nature of the resource. Smith (1981) points up some related and other factors
- extraction costs, new discoveries, changes in market structure, and changes
in the institutional environment - and stages a tournament for several
alternative arbitrage models. He finds the Heal-Barrow model to be among
the ‘best’ models, but that its predictive power is not strong compared to
simple time series models. Like Frank and Babunovic, Smith found the
forecasting performance of the models tested to vary widely across minerals.
A later model by Heal and Barrow (1981) addresses some of these issues
by including variables held to be proxies for depletion and the change in
costs through time. Although they cannot include directly the effects of
technological progress and new discoveries, which drove the results of
Barnett and Morse (1963), Pindyck (1978) and Slade (1982), Heal and
Barrow claim an improvement in the specification of the model. The simple
equilibrium version of Hotelling model is nested within their model, and is
rejected by their tests.
The strongest challenge to the Hotelling analysis is that of Farrow (1985).
Using data provided by a mining company on a confidential basis, he tests
the predictions of a Hotelling model for a single mine. The empirical results
are not consistent with the predictions of the model. Farrow (1985, p. 477)
concludes that ‘it is clear that the Hotelling model is insufficiently robust to
be confidently applied.. .‘. The results are, however, informally consistent
with a rule of thumb employed by mining firms to determine the lowest
quality of ore that is extracted: in times of rising prices the firms lower the
cut-off grade. ‘The argument that is used to justify this rule of thumb is that
the current higher price pays the higher extraction and processing costs
implied by lower-quality ore’ [Farrow (1985, pp. 46&467)]. He adds that it
is not certain whether the rule of thumb is consistent with profit maximiza-
tion, but that preliminary results of now-published work by Krautkraemer
(1988) suggest that it is not.
Farrow’s findings are doubly troubling. First, they suggest that the
Hotelling approach as a positive description of practice may be incorrect.
And, this is not merely a result of the fact that the resource may not be
sufficiently scarce to justify an observable rent, as found by Cairns (1981).
146 R.D. Cairns, Geoiogical influences, metal prices and rationality

Second, it points to a possibility of non-rational behaviour, without an


explanation for the departure from rationality nor for its extent, whereas
economists would be inclined to view rules of thumb to develop as an
approximation to completely rational behaviour in a situation where perfect
information cannot be obtained.
Slade (1988, p. 194), too, notes that ‘anyone familiar with the mineral
industries . . . knows that when nominal price increases (falls) lower (higher)
grade ores are mined’. In order to resolve the seeming contradiction between
theory and practice, she models (in agreement with her empirical evidence)
metal prices as martingales. Thus, the apparently myopic responses by mine
managers to price changes can be explained as optimal responses to
uncertainty. However, her model assumes that quality (grade and size) varies
across deposits, not within deposits. Explicitly modelling the cut-off grade
decision, Krautkraemer (1988) does find that in some cases a martingale
specification of price can explain the anomaly. Yet he is unconvinced,
concluding that ‘the simple rule of choosing the cut-off grade on the basis of
current profitability is not a general rule for maximizing the present value of
returns to the mine’ (p. 159).
This paper is an attempt to reconcile the disparate findings, some
supportive of, some challenging to, the Hotelling perspective. Here, we
present a stylized Hotelling-type model that treats explicitly the fact that a
mining company is constrained by geology to mine ores of different grades
simultaneously, a relaxation of assumptions that Farrow (1985, p. 478)
considers desirable. Using a very similar approach to that of Krautkraemer
(1988) and of Farrow and Krautkraemer (1987), we find the rule of thumb
observed by Farrow to be consistent with rational behaviour as in a
Hotelling approach. Furthermore, the model predicts that the ‘rule of thumb’
will hold even for anticipated price changes, and for situations in which
capacity is not constrained. Farrow and Krautkraemer’s (1987) theoretical
analysis suggests that predictions are strongest when capacity is fixed, but
their empirical results suggest that ‘extractive capacity is not constrained’ (p.
23). Their theoretical model also holds that only unanticipated price changes
tend to a negative relation between price and cut-off grade. The nature of
our data are such that it is plausible to view price changes as being
anticipated. We do not deal with the empirical question of whether price
movements for certain metals can be represented by martingales.
The model is a two-sector neoclassical equilibrium model which, combined
with a simple assumption on savings behaviour, also obtains a prediction of
second-order effects on prices and interest rates. Consistent with Krautkraemer’s
and Farrow and Krautkraemer’s results, the actual equation obtained is very
complicated. The second-order effects depend on the parameters of the
production technologies in the mineral and goods sectors. In reality, there
are more than two sectors, and the variation in technology across sectors
R.D. Cairns, Geological influences, metal prices and rationality 147

may explain some of the variation other researchers have found in estima-
tions for different depletable resources. The interest rate itself appears. It is
argued below that, contrary to Heal and Barrow’s theoretical analysis, this
finding is not inconsistent with their empirical results after all; at the same
time it is more plausible, as Heal and Barrow themselves noted. The result is
also consistent with the results of Frank and Babunovic.
The price-interest rate equation is not estimated, but derived only to show
the possibility of second-order effects arising from the Hotelling model with
varying grades. Rather, two other conditions, which together imply the
second-order effects, are derived from (1) the dynamic condition arising from
the maximum principle and (2) the first-order conditions for an individual
mine, in conjunction with equality of factor rewards in the two sectors. These
cannot be estimated immediately, but are manipulated using a well-known
expression due to Lasky (1950a, b) for the ‘tonnagegrade’ relation. This
relation enables substituting average grades into the formula where cut-off
grades appear. Even so, as has been found by other researchers, the
necessary data are not readily available. An appropriate series of data does
exist, however, for Canadian gold mines which qualified for government
assistance under the Emergency Gold Mining Assistance Act, in force from
1948 to 1972. Among other things, the data allow for circumvention of the
important problem of estimating the in-situ value of ore. Also, it cannot be
held that the price of gold behaved as a martingale in this period. The single-
mine conditions can be tested for several of these mines, and the predictions
of the model tend to be accepted.
Therefore, while it may be argued that the peculiarities of individual mines
may give mine managers additional degrees of freedom to pursue ‘manager-
ial’ objectives, the theoretical and empirical findings of this paper give no
support to calls to abandon the postulate of rationality implicit in Hotelling’s
approach.

2. A stylized model of minera extraction


We assume there are a large number of identical mineral deposits, each
held by a different price-taking firm, j, and taking the form of a cylinder of
known length. Ore quality, expressed in terms of grade, g, is a declining
function of the radial distance of a sample of ore from the axis of the
cylinder: g=g(r) and g’(r)<O. Ore is exploited by each firm, j, through its
cylinder. A cut-off grade (equivalently, radius) is determined at each instant.
Thus, at time t, the distance along the axis of the cylinder at which mining is
taking place will be H(t), and ore out to radius r(t) will be mined. In
addition, extraction costs are assumed to rise exponentially with H(t); for
example, if the cylinder is viewed as being vertical, H(t) would be ‘depth’ and
costs would increase with depth.
148 RD. Cairns, Geological inJuences, metal prices and rationality

We further assume that, once ore has been exploited at any distance along
the cylinder, any ore left behind of lower grades than the cut-off grade will
never be exploited thereafter. This is a strong assumption. Occasionally,
mining firms can return to re-work already worked stopes; technological
change usually occurs, however, to enable ths change. [Compare Boldt (1967,
137ff.).] We offer only the following explanations for this strong simplifi-
cation. First, it is assumed in this paper that there is no technological
progress.2 Second, when royalty rates in British Columbia were raised
substantially in the mid-1970s base metal mining firms responded with
complaints that huge quantities of close-to-marginal ore would be ‘lost
forever’, because it later would be uneconomic to return to extract ore left
behind as a result of the policy. This argument suggests that, most plausibly,
the assumption may be viewed as the limiting case of there being a very high
cost to returning to exploit an already worked part of a mine. This is clearly
true when mined-out portions are tilled in, as is common in underground
mining.
The economy is divided into two competitive production sectors: one for
producing mineral and one for producing a composite capital-consumption
good. Following Solow (1974b) and Dasgupta and Heal (1979), we assume a
Cobb-Douglas technology. This assumption is made for both sectors. In the
mineral sector, capital and labour combine to extract mineral from a certain
volume of ore; thus, the lower the average grade, the more capital and labour
are required to produce a given amount of metal. In the goods sector, metal,
labour and capital combine to produce composite output. Zero population
growth is assumed, so that total labour supply I., is fixed. The capital stock,
K, is also assumed malleable between sectors. The behavioural assumption is
that of (discounted) profit maximization on the part of each firm.
The problem can then be set up if we make a few more definitions. Let
h(t) =dEI(t)/dr be the length of cylinder being exploited at time E. Then a
total volume of ore, V, is produced by a Cobb-Douglas technology:
rcr’h= V=L’ KlaCemkH, where O<c< 1 and eekH reflects increasing factor
use (cost) tl eitract ‘deeper’ ores .3 The subscript ‘rn’ denotes the mineral
sector. Total output of metal (which is input to the second sector), however,

ZThis factor was found to be important by Frank and Babunovic (1984). Recognizing its
significance, we abstract from it in order to display the effect of geological factors more clearly.
We assume below that there is no technological progress in our empirical tests, and provide
justification for the assumption.
3The production function has the advantages of famiiia~ty and ease of manipulation, but the
disadvantage of restrictiveness. We argue, however, that the restrictiveness is not so great as to
overcome the advantages, Farrow (1985, p. 469, n. 12) tested for separability of depth and
output and did not reject that hypothesis. Halvorsen and Smith (1984, p. 960) found
homotheticity of production in reproducible inputs. They did, however, find biased technical
change, which cannot be represented here; but, as our empirical estimations assume no
technological progress, we neglect this matter. It is interesting to note that for petroleum,
Livernois (1987) found a depletion effect (our e -kH) but that otherwise marginal extraction costs
were constant.
R.D. Cairns, Geological influences, metal prices and rationality 149

is M =nmh, where m=& 2xg(x) dx. Thus, M =(m/r2) I/: The factor multiply-
ing r/; m/r2, will then be recognised to be the average grade, which we shall
call G. Finally, M is an input in the production of Y: Y=L;KK,bM’-“-b,
where a>O, b>O, and a+ b-c 1. The final good, I: is treated as the
numeraire.
The mining firm’s problem is then to maximize

$(pnmh-wL,-pK,)c”dt

such that 7 hdtgHH,, and V=nr2h=L’,K,!-ce~kH.


0

The problem of the lit-m thus becomes a straightforward control problem.


The Hamilton-Lagrangean is, in current value form,

The control variables are L,, K,, h and r and the state variable is H.4 We
assume that H is concave in the control variables. [As Krautkraemer (1988)
observes, this imposes restrictions on the grade function, through m]. First-
order conditions and the dynamic condition then yield.5

k/s = kh+ G/P- d(Wr2d - 1). (1)

In addition, the first-order conditions and the condition of equal factor


rewards in the two sectors yield

This is comparable to Farrow’s model, although much simpler in terms of


the dual cost function. It may be simplified to

g = pp$, 5 -c/a ek”,


(2)

where /I>0 is a constant involving a, b and c

4The model assumes that the cut-off grade responds relatively quickly to changes in p, p, etc.
In practice, this is true; the mine manager can order a change in g on a daily basis.
‘The manipulations, all straightforward but lengthy, are done in Appendix A.
150 R.D. Cairns, Geological influences, metal prices and rationality

5=c((l-C)/C-b/a)20

and

(+c/a= 1 +(~/a)(1 -a-b)> 1.

In (2), we express g in terms of p and p, which are the commonly used


right-hand side variables. Also, (2) involves similar variables to (1).
Finally, under the proportional savings assumption (R,,, + R,,= sY where s
is constant), a rather formidable-looking equation is obtained:

sp/((l-a-b)(l-c))=--
“[P
i-(
a
+b) i-e
(P Jl
(d/P) - P b
’ WI+p-(c/4(1 --a--b)Wp)+(1
-a-Ml -4 1

_ r((ii~-d~)/~~)-P-((cla)(l-a-b)(ti~-ri~)lp~)_ ""~,;;~;hj
5(PI~)-(c/a)(l-a-b)hlp-p 1

x (1-W(41 -cMW-P)
aPIP)-ct +(c/m 1.

In order to deal qualitatively with this equation, which is not amenable to


intuitive interpretation, we re-express it as

f(p, rj,ii,P, P,ii; a,b,4 = 0. (3)

3. Interpretations

These equations have several implications. We begin with (3). The first
thing to notice is that we obtain second-order effects on the price and
interest rate, a result which approaches that of Heal and Barrow [1980, p.
163, eq. (5)]. There are several differences, however.
First is the degree of complexity of (3) and its non-linearity. If the
technologies of production for different minerals are different, so that our
stylized model of mineral production may be viewed as a simplification, then
the complexity is consistent with the wide variation in performance across
minerals of Heal and Barrow’s model and of its forecasting properties as
reported by Smith (1981), or indeed of other models which do not take
quality variations into account. Certainly, without the background of the
R.D. Cairns, Geological influences, metal prices and rationality 151

theoretical model, one would be hard pressed to find a potential for


consistency of price movements across resources if anything like expression
(3) is close to the truth.
A second difference is that the third time-derivative of price does not
appear in (3), but does in Heal and Barrow’s model. This turns out to be
consistent with their results, as the non-appearance of 2;’ would imply the
non-appearance of a term in @ lagged two periods, or r,(t-2) in their
relation. In the empirical results [Heal and Barrow (1980, pp. 168-171)], the
coeflicient of r,(t-2) is not often significant. Also, it drops out theoretically if
it is assumed that economic agents use the same form to generate expec-
tations of price changes as of interest rates6
The main difference is that the interest rate itself appears. Much is made of
its non-appearance in their model by Heal and Barrow (1980), and Farrow
(1985, p. 459) remarks on how ‘uneasy’ they are with the conclusion. In
Appendix B we follow up on the logic of the last paragraph of Heal and
Barrow’s (1981, p. 102) second paper, by successively differentiating the naive
Hotelling prediction @/~=p. From this perspective, we can argue that it is
the presence of a term in Ij/p in Heal and Barrow’s model, not the absence of
a term in p, that ‘needs’ to be explained. Their empirical results can be
reinterpreted to test whether or not p/p appears. One can say that their
results reject the twice-differentiated version of the naive Hotelling rule. But
that result is implicit in all the empirical work discussed in the introduction
to the present paper. We should be very surprised if such predictions were
not rejected.
In the present paper, the second-order effects (on prices and interest rates)
appear because of the interplay of two factors. First, there is the dynamic
condition (1) arising from maximization of the present value of its resource
by each mining firm. Second, as the derivation following eq. (A.1 lb) in
Appendix A points out, the mining sector must compete in the capital
market for the available savings of society. The growth in the mining sector’s
capital stock depends on the growth of rents it can generate, and hence on
depletion (depth) and the grade distribution. The feedbacks operate through
the cut-off grade distribution. The introduction of an explicit role for capital
in the model highlights the importance of the other sector of the economy to
equilibrium in the resource sector. This is a fact also noted by Farzin (1984)
who presents a model in which there is a resource sector and a backstop
sector, i.e., a second resource sector. Here we do not lose sight of the
importance of resources as assets competing with other assets, but emphasize
the role of all decision-makers in the economy.
In addition, the two-sector approach sheds important light on the role of
prices and interest rates in determining the grade margin. In eq. (2) we see

6This would entail a2=a3 in their equations (3) and (3’) [Heal and Barrow (1980, p. 163)].
Compare also the motivation for the second paper [Heal and Barrow (1981, pp. 95-96)].
152 R.D. Cairns, Geological influences, metal prices and rationality

that the sign of the effect of the interest rate depends on the relative capital
intensities in the two sectors. Again, our observations are similar to those of
Farzin (1984). An increase in the rate of interest increases (capital) costs in
both sectors, but more so in the more capital-intensive sector. If the goods
sector is more capital intensive, capital will move to the mineral sector and,
abstracting for the moment from effects on h, the increased capital will allow
for greater metal output through a reduction in g, which will substitute for
capital in the goods sector. If the resource sector is more capital-intensive,
capital will flow out. Richer ore will be substituted for capital, so that the
cut-off grade will be raised.
The effect of a change in the interest rate on the rate of depletion is more
complicated, and again not in line with simple models with homogeneous
resources and no capital, From Appendix A7 we obtain

N-4 l-c&e-k”p-“-” ’ *+Ga- 1


A= -.._.
[ c 1 i-C [/I1

g c(1 -a-b)
_.
11 r2
(4)

There is an effect of p on h because of the cost of using capital in production,


through both the direct effect of p in (4) and the indirect effects of the
expressions involving rz and G/g. A conservation effect arises through using
available capital and labour in the mining sector to change r. As &/dp has
the opposite sign to ag/ap, even the conservation effect may be seen to
depend on relative capital intensities. It may well occur that a reduction in
the discount rate hastens depletion. It is noteworthy that the sign of the
effect of p on h depends on the grade distribution (through G/g) as well as
capital intensities. Finally, one recalls that the influence of p on p is given by
(31.
From (2) we also conclude that agjap ~0. Thus, Farrow’s (1984) ‘rule of
thumb’ is consistent with profit maximization, at least in this model. We
stress that this result follows directly from the first-order conditions (see
Appendix A) for maximization of profit at the mine level, and the conditions
for equilibrium in factor markets. Indeed, it is established in Appendix A that

pg=(p/(l -~))‘-~(w/lc)ee~~.

The dual cost function in the mineral sector is8

C,(p, w, H, I’) = (p/( 1 -c))’ -‘( w/cf’ ekHK where ‘I/= nr2h.

‘In particular, eqs. (A.8b) and (A.13) in conjunction with the mineral-sector production
function.
8Note that this is the cost function for ore. In the cost function for metal, V is replaced by M/
G. Average cost is decreasing in G, increasing in r.
R.D. Cairns, Geological influences, metal prices and rationality 153

Thus, pg=X,/aV That is to say, one can express the mine manager’s rule
of thumb in selecting a cut-off grade such that revenues from mining the
marginal unit just cover marginal costs. There is no rent at the extensive
margin, and all resource rents appear as differential rents. This result was
found by Cairns (1986b) in a one-sector model of mining.
This result raises the question of the similarity of our model to that of
Krautkraemer (1988). In fact, his results show that the rate of change of the
cut-off grade depends on the rate of change of the price, less the rate of
discount. He assumes that there is no depletion effect, which in our model
means that k=O. Under this assumption, our dynamic condition (1) implies
that

qg/gya 5 --p = G/g- 1 >o.


(* )

Thus, our results are in agreement with those of Krautkraemer.


Krautkraemer’s results, however, do not relate to the ‘rule of thumb’, but
pertain to the rate of change of cut-off grade rather than to the cut-off grade
itself. Similarly, he finds a relationship between price and the rate of change
of h, where h is the rate of depletion. In models where multiple grades are
taken into account, then, extrapolations of the predictions of simple Hotell-
ing models (with no depletion effect) tend to hold for the rate of change of
the rate of depletion, not for the rate of depletion.
When there is a depletion effect, as in (l), the prediction has to be
modified further, to take into account any changes in h occasioned by a
change in (@/p--p). [See (2) and (4).19
The question remains as to how closely eqs. (l), (2) and (3) conform to
reality. Our development indicates that (3) is derived from naive assumptions
about savings behaviour and the grade distribution across mines. Interpre-
tation and estimation of this equation, then, would seem pointless. This
suggests that devising appropriate tests of the consistency of resource pricing
with the Hotelling analysis is a formidable undertaking with unclear benefits.
This is especially true when we recognize the long list of other factors -
technological progress, new discoveries, etc. - that we have neglected.
Fortunately, we do not have to estimate (3) to test the Hotelling predictions.
Eq. (3) is implied by eqs. (1) and (2), and it is these equations which reflect
the optimizing behaviour of resource firms in response to their environment,

9A further point is that if p is perturbed discretely, as Krautkraemer hypothesizes, then in


general g, as well as g/g, will be perturbed discretely in order to satisfy the necessary conditions.
The shift of g will be picked up by (2) but not by (1) which is analogous to Krautkraemer’s
equations.
154 R.D. Cairns, Geological influences, metal prices and rationality

as predicted by the Hotelling model. Unfortunately, estimating (1) and (2)


would require data - on cut-off grades and depth - which are not readily
available. The next section outlines how these are transformed into variables
for which data are available.

4. The grade-tonnage relation

The exponential distribution was used by Lasky (1950a,b) to model the


quantity of ore at various grades. Musgrove (1971,1976) considered the
mathematical implications of the distribution and applied it to several
minerals, obtaining ‘excellent fits’ in almost all cases. Unfortunately, his data
were drawn from large regions (e.g., the United States, the world) over fairly
long periods, where there could be several contaminating influences. [Cairns
(1986a, p. 94) discusses one of them.] His results cannot be considered to be
valid tests of the model, and it is probably best to view this particular law of
deposition as applying to individual deposits and small geographical regions.
In this paper, we apply the distribution at the deposit level. First, we note
that, in physical terms, average differential rent per unit ore is G-g, and
average metal output per unit ore is G, so average rent per unit metal output
is

R= 1-g/G.

We use the grade-tonnage relation to obtain an expression for &‘g in


Appendix C. Then, recalling that rn/r’=G, we get, by substitution for g/g in
(l),‘O

We still need to estimate h. The variable H, or depth, was used in the


stylized model solely as a physical motivator to represent the increase in cost
as mining proceeded, i.e., the depletion effect. In the literature, a common
way to represent the depletion effect is to assume that costs rise with
cumulative extraction. If we let H be cumulative extraction, then h=fi is
current output, 4.’ 1 This yields the equation

‘*Musgrove gives interpretations for the parameters of the model. The average grade in the
deposit, for example, is K.
“Strictly speaking, this would imply theoretically that h=q=nmh, so that grade is constant.
But there are few other proxies for depletion. (Remaining reserves, for example, would also call
for use of q where h appears.) One candidate is time itself, In this case, q would be replaced by a
constant in (5). Empirical results using time as the depletion effect are very similar to those
reported below.
R.D. Cairns, Geological influences, metal prices and rationality 155

- ; & G(P-f)+&(P-~).
~=~(RGq)-k(Rq)

This equation corresponds to (1). By logarithmic differentiation of (2), we can


express g/g in terms of d/p, p/p and h=q to obtain another equation:

+(C+;)(ZZf)+$RCq-kI7q. (6)

These are the equations to be tested for equality of various coeflicients, for
the signs of various coefficients, and for the value of < +c/a, which should
exceed unity.

5. The data and estimations

What we wish to test is the response of mine managers to changes in the


external environment, in terms of rates of depletion and especially the choice
of cut-off grades. The derived equations, (5) and (6) involve average grades,
under the assumption of a particular grade distribution; data on average
grades are not difficult to obtain. But they also involve the Ricardian rents
on a mine-by-mine basis. These are usually very difficult to find.
The data set used here relates to Canadian gold mines qualifying for
assistance under the Emergency Gold Mining Assistance (EGMA) Act, which
was in effect in the post-war period until the price of gold rose very rapidly
with the loosening of constraints on it after 1971. The political economy of
the Act need not detain us here; it has been ably reviewed by Lasserre (1983).
S&ice it to say that gold mining was becoming unprofitable as inflation ate
away at the fixed figure of U.S. $35/fine ounce through World War II. The
Act provided subsidies to gold mining which varied according to mining
costs. Annual payments to individual mines, from which costs and rents
(including subsidies) may be directly deduced using the subsidy formula, are
tabulated by Energie, mines et ressources Canada (1976).
In one way, the choice of a data set for this particular metal may appear
156 R.D. Cairns, Geological injuences, metal prices and rationality

strange because gold’s being a monetary asset with its price set by monetary
authorities would appear to be inconsistent with the two-sector general
equilibrium by which prices, etc. are established in the theoretical model. But
endogeneity of price comes in only at the stage of eq. (3), and hence after
(A.1 lb) in the appendix. Eqs. (1) and (2) can be viewed as responses to
parametric parameters p, p, etc. Thus, we can divorce the single-mine results
entirely from endogenous determination of p and p and still obtain (1) and
(2) as expressions of producer behaviour. As this behaviour ‘is what we wish
to model, there is no inconsistency.12
Indeed, the data set has several useful features. The stylized model was
derived on the assumption that various influences on pricing other than the
depletion effect and the cut-off grade - such as unanticipated price develop-
ments, by-products and co-products, technological changes, exploration,
institutional change, etc. - were controlled. The fact that gold was a
monetary asset provided conditions as near to iaboratory conditions as can
be dreamt of in economics. The mines subsidized by the EGMA Act had
several broad characteristics, which we summarize here.
First, the price of gold was easy to predict, at least in nominal terms: it
was fixed at U.S.$35/ounce. The mine manager was clearly a price taker.
Furthermore, all he needed to do to predict price was to predict the rate of
inflation and the exchange rate. There were no unforeseen shocks, except in
these variables, and the impact of such shocks was minimized by the
assistance formula (which was adjusted from time to time). Yet, these two
variables did result in changes in the real price of gold in Canadian dollars.
In order to qualify for aid, the mines had to derive at least seventy percent
of their revenues from gold. Thus, co-products were a minimal influence on
the choice of cut-off grade, and we have effectively controlled for the effects
analysed by Pindyck (1982) and found to be troublesome by Farrow and
Krautkraemer (1987, p. 18). In addition, there is an independent estimate of
Ricardian rent in each year for each mine, which is consistent across mines
and across time.
In the individual mines, technological change over the period of obser-
vations (which was short} was not likely an important factor: major

t2The fact that the object of the present paper’s empirical investigation is gold mining has
some incidental interest in the history of thought. An earlier literature, including an important
article by Paish (1939) sought to determine the causes of fluctuations in gold supply with
changes in economic conditions. Many of the considerations which motivated this paper were
discussed at length by Paish. For example, it was observed that in times of high pro~tabiIity,
short-run profit maximization would suggest high grading of the deposit. In reality, however,
changes in conditions would lead to changes in the marginal grade of ore mined, with increased
prices leading to the mining of lower grades. Regulations and taxation to discourage high
grading were discussed. Maximizing mine life or total yield was also a value inculcated in
engineers. Engineering practice, then, set itself against rapacious exploitation in the short run or
‘picking the eyes of the mines’ [Paish (1938, 395 ff.)]. Such practice was seen as consonant with
the self-interest of engineers, but not of shareholders. In the present paper, we are investigating
to what extent engineering practice would be agreeable with shareholders’ interests.
R.D. Cairns, Geological influences, metal prices and rationality 157

technological changes would tend to be diffused at new mines, rather then


producing ones. But these changes, and the change induced by new mine
openings (the result of exploration) were unimportant to an individual mine
manager, as they did not affect the conditions of the market with its fixed
price. In addition, the low grades of gold ore mined .made cut-off grades an
important decision. Grade did indeed vary over time at each mine.
For estimation purposes, we may rewrite (5) and (6) as follows:

-~3(&+- ;))+Q, (5’)

-& RG; +~,R;+E~, (6’)


( >

Rates of change of variables are approximated by first differences in the


usual way, and all variables are expressed in real terms. The theoretical
section suggests testing the following hypotheses:

PI = 8283; and pO=l.

The tests are described more fully in Appendix D. SutXce it to say that the
first and last of these are tested using a t-test, while the others utilize a cross-
equation technique based on that of Mishkin (1982, pp. 22-31). There is one
major complication, however: G (the average grade of ore mined) and R (the
Ricardian rent accruing to intramarginal ore) are determined simultaneously.
Because Mishkin’s tests are maximum likelihood tests, we have utilized
instrumental variables estimation in order to correct for simultaneous
equation bias when using the tests.13
Of all the gold-mining firms in Canada which qualified for assistance, there
were only eight for which there were enough observations (at least eleven) to
perform the tests on both equations. These results, reported in Appendix E,

lsWe used instrumental variables as we did not have data on all predetermined variables.
158 R.D. Cairns, Geological influences, metal prices and rationality

are consistent with the model developed above.14 The values for R2
generally indicate that the equations have explanatory power. Standard
errors on individual coefficients are high, however; thus, our t-tests are not
powerful. Multicollinearity may be a problem, given the values for R2 and
high standard errors of coefficients. As our theory does not allow for deleting
variables, ridge regressions were performed and individuai coefficients were
much closer to the values predicted by theory. Reported standard errors are
biased, however, and cannot be used for testing.
In interpreting the empirical results, one should put more weight on tests
involving eq. (5’) [and hence (I)] alone rather than the cross-equation tests,
for three reasons. First, eq. (2) was derived under the assumption of a Cobb-
Douglas technology. But it is straightforward to show that (1) may be
derived if the production function in mining is the more general eSkNF(K, L).
We have given theoretical reasons above why the Cobb-Douglas may be a
reasonable approximation in this industry. Indeed, eq. (2) arises as a
simplification of (1) in the model. Because of the simplification and introduc-
tion of additional information, one should not expect perfect, but close to
perfect, correlation between the error terms if the Cobb-Douglas assumption
is correct. This is not a rigorous test, and only suggestive, but a comparison
of the residuals from (5’) and (6’) for the different mines shows that they
have correlations ranging up from 0.9, and averaging about 0.97. Still, other
empirical work cited above suggests that the Cobb-Douglas may not be an
accurate description, whereas that empirical work is consistent with the more
general form for which (I) is valid.
The second and third reasons for putting more weight on (5’) relate to the
fact that there are fewer coefficients to be estimated. This gives more degrees
of freedom to the estimations for the eight original mines; yet another is
added if (5’) is estimated under the constraint PO= 1. Furthermore, there are
four mines for which we have enough observations to estimate (5’) but not
(6’). For these mines, too, the hypothesis that ,fIi=f12fi3 is not rejected.
Finally, it proved possible to check whether the laws of deposition were
such as to lead to the same grade-tonnage relation (same value of K) at all
mines in a given geological region. The hypothesis that &( = I/K) is the same
across mines was examined using the data for the seven gold mines in
Ontario (namely, mines numbers 1, and 3-8). Due to limited number of
observations, a test based on Mishkin’s (1982) test was used. The data did
not support the hypothesis. The calculated test statistic for the hypothesis
was 65.4 as compared with the theoretical value of chi-square with 30
degrees of freedom at 0.05 probability level of 43.8. This could be due to
geological differences among the mines examined as well as technical change

14The model was also tested using OLS, the results were little changed. In addition, the cross-
equation technique allowed the taking into account of non-zero covariances of the errors across
equations, as a further generalization of Mishkin’s (1982) method.
R.D. Cairns, Geological influences, metal prices and rationality 159

as the sample period was not the same for all the mines in question. This
result suggests that aggregation across mines for purposes of estimating
depletion and price trends may be a dangerous practice. [Compare Livernois
and Uhler (1987) for the oil industry.] In terms of our theoretical model, one
of the major simplifying assumptions used in deriving the already inelegant
(3) is not valid. Thus, the influence of interest rates on prices is undoubtedly
even more complicated than suggested by (3).

6. Conclusion
The purpose of this paper has been to present a theoretical model of a
non-renewable resource industry which reconciles the apparently contra-
dictory empirical facts found by other researchers. These findings have
beclouded the Hotelling-style vision of resource managers as rational profit
maximizers.
We have found that the seeming contradictions can be reconciled in a
simple perfect-foresight model which takes into account the facts that (a)
geological conditions are such that varying qualities of ore must be mined at
any given time and (b) the mining industry is embedded in the industrial
economy. This is not to say, however, that reality does not throw up other
important influences - including adjustment cost, imperfect information, etc.
- which have been used by others to explain their findings. These are clearly
important. But the present paper has tested a model which controls for
several of these additional influences. The industry, Canadian gold mining,
has provided an almost ideal laboratory in this sense. The theoretical
development has suggested one reason why various researchers may have
had mixed success in explaining the influence of interest rates on metals
prices, and why that task may be a very daunting one indeed. In the
empirical work, we have run up against another problem of those who have
tried to disaggregate non-renewable resource industries. We have been
unable to obtain a large number of detailed observations, and have thus
been limited to only a small proportion of the producing mines. A number of
assumptions have been required in order to make use of what data were
available. Therefore, the empirical results are only suggestive.
But our results do shed some light on the behaviour of mine operators.
Contrary to the assumptions of Hotelling-style models, it could be, for
example, that the operators use a proportion of their rents to pursue
managerial objectives such as maximizing the quantity of ore mined or the
life of the mine. The uniqueness of each deposit could leave ample room for
expression of such values. It is less likely, however, that such wealth-
destroying behaviour would be systematically expressed in profession-wide
rules of thumb respecting the response of cut-off grades to price changes. The
present paper suggests rather that such rules may also lead to present value
maximization. The empirical results, although not robust, do not reject the
I60 R.D. Cairns, Geological influences, metal prices and rationality

theoretical expectations. While these mines did need assistance - and hence
profitability was a constraint as much as an objective (cf. perfect compe-
tition) - the combined weight of theory and evidence argues for renewed
caution in the face of calls to abandon the postulate of rationality.

Appendix A: Equilibrium conditions


The Hamilton-Lagrangean for the problem of the firm is, in current value
form:

The control variables are h, r, L, and K,, and the state variable is H. We
obtain from the maximum principle the following first-order conditions:

(A4

64.2)

(A-3)

(A.4)

=-kvdh if v>O

= -nkpgr’h by (A.2). (A.9


From (A.l) and (A.2),

u = np(m - r2g), (A.6a)

ti = nj(m - r’g) + xp(2rgf - 2rgi - r28), (A.6b)

Substituting (A.6a) and (A.6b) into (AS), and simplifying,

g/g = kh-t G/P- t4W’r2g


- 1). (A-7)

From (A.2), (A.3) and (A.4) for each mine in production,

(AXa)
R.D. Cairns, Geological injluences, metal prices and rationality 161

= w( 1 - c)/(pc). (A.8b)

From profit maximization in the final good sector, we have

aYJL,=w; bY/K,=p; (l-a-b)Y/M=p. 64.9)

These results imply

w = a(( 1 - a - b)/p)” a - W”(b/p)W”,


(A. 10)

From (A.8a), (A.8b), (A.1 l),

l-cp-lekH(l__)-(l-c)c-c,
g=w’p (A.1 la)

g = /jp$, F -c/a ekH,


(A.1 lb)

where

r= 1 -c-(bc/a)=c((l -c)/c-b/a)pO,

t+c/a=(c/a)(l -a-b)+ 1 > 1,

c(l-o-b)/a(l
/I=(l-a-b) _C)~(l-c)bbc/aacc-c,

To obtain an equation for p and p, let us assume that g,/G, is the same for
all mines j in production. This places strong conditions on the distributions
of producing mines (e.g., the grade distribution is the same at each). We are
reminded of Livernois’ (1987) finding of aggregation biases. The relationships
we obtain even assuming no aggregation biases are still quite complicated,
and the following analysis is intended to be indicative only. Under the
assumption,

=(G/g)w’p’-‘p-‘(1 -c)-(l-c)c-cLC,jK$c, by (A.lla),


162 R.D. Cairns, Geological influences, metal prices and rationality

But by (A.9), M = (1 - a - b)pK,/(bp). Thus,

K,=[b/(l-a-b)(l-c)](G/g)K,. (A.12)
Similarly,
L,=[a/(l-u-b)c](G/g)L,. (A.13)

Now, under the constant proportional savings assumption,

k,+k,=sk: where from (A.12) we have

K,+K,=K,[l +bG/((l -a-@(1 -c)g)],

sY=k,[l+bG/((l-a-b)(l-c)g)]

+K,,W(l -u-Ml -Wk-SGYg21, (A.14)

L = (LILJL = Cpg( 1 -c) e ~kH/p]“cL,/[l +uG/((l -u-b)&],

by (A.8a), (A.13),

%,,/K, = ( l/c) WP) + (k/d -(b/P) - khl + (@I,


where
O=[c(l-a-b)g]/[c(l-a-b)g+uG],

&‘J = [(g/g) - (~lGMW)( 1-a -W/G) + 11.


Now,

Y/K,,,= L;K,Gb 1
- a - b K,(l-c)(l-a~b)~c(l~a-b)e~kH(l-a-b)/~
m IfI.

This simplifies to

Y/K, = Gg 1- @/d(p/p)(a/d+ 1-0 -b eWdk~fibld/(( 1 _ c)( 1 _ u _ b)).

Using this result and dividing (A.14) by K,, we obtain

o/c+(l-amb)ekHa/c e/c
sg ““(P/P) P /(Cl -a-b)(l-c))

= CC l/c) [(d/p) +(2/s) - (P/P) - khllIN + M(l - a - b)(1- c))l

+ C(dg) - @lG)l I31- W4 1- cMG/g + c( I- a - b)/u)]. (A.15)


R.D. Cairns, Geological influences, metal prices and rationality 163

The three equations from which we now operate are (A.7) (A.1 1b) and
(A.15), in order to eliminate g, h and H.
We can logarithmically differentiate (A.1 lb). The result, and (A.7), give two
expressions for (g/g - kh). Equating them gives an expression for G/g in terms
of p, p, p and j. Substituting into (A.15) we obtain

sp/((l-a-@(l-c))=-1
“L
p-(
a
+b)
(P
d-P
Jl
1
(P/P)-P b
WP)-p-(c/4(1 -a-bWpj+(l -a-b)(l-4

_ S((iip-p2)lp2)-P-_((cla)(l---bb)(iip-d2)lp2@P-?WP~)~_P~
WP)-(cl4~~ -a-b)Plp ~ - PIP-P 1
x (1-W(a(l -~MWP)-P)
WP)-(5+(+4)p 1’
(A.16)

Appendix B: Re-examination of Heal and Barrow’s equation (5)

The simplest Hotelling model predicts j/p = p. Differentiating once,

iilP-wP)2 =P.
Differentiating again,

k/P - 3WPWP) + w/P)3 = ii.


Hence,

‘ii/P- 3GvPWP) + w/P)3 + W/P) - wP/P)2 = li + BP,

for any constant B. Solving Heal and Barrow’s eq. (5) where rc-fi/p, gives

F/P - ~WP)(@/P) + W/P)~ + &/PI - W/P)~ + W/P) = aii + bfi.

The two equations are very similar. Note that p itself does not appear in the
differentiated Hotelling model. Nor does the linear term in j/p. (In the Heal-
Barrow equation, all non-linear terms in p/p are a part of Vcand i,.) A test of
the Hotelling prediction would be to check if C=O; a = 1; and b=B. Informal
164 R.D. Cairns, Geological i@uences, metal prices and rationality

checks of the implications of these for Heal and Barrow’s estimated


coefficients suggest that these would be rejected in the sample. But recall that
the equations were derived by purely mechanical manipulations of the
prediction of the very simplest Hotelling-style model.
Consider in this light the simpler model [Heal and Barrow (1980, pp. 173-
174; 1981, p. 96)]. [There is a printing error in eq. (2) in the 1981 paper.] It
implies that

P/P+ wP)2 = c(PlP)(dlP).

The first derivative of the prediction j/p=p gives

ii/P - (dlP)2 = P =(dP)(d/P).

[The implied conditions cannot be tested using Heal and Barrow’s results
(1981, p. 98) because their eq. (3), the basis of their estimations, is not derived
from (2) of the 1981 paper. Rather, it arises in the more complicated model
of the 1980 paper, pp. 161-165.1

Appendix C: Derivations of equations to be estimated

Suppose the quantity of ore by grade follows the probability distribution


L(g) throughout the ore body (and hence at each ‘wafer’ of thickness h),
where 0 <g gg(O) =g,,, 5 1, with g, being the maximal grade. Let f(g) =L’(g)
and F(g) = 1 -L(g). Then L(g,) = 1 means F(g,) = 0 and ore content at grade
at least equal to g is distributed as F(g). Also, ore content in an annulus of
width dr at radius r is

2m dr = - Af(g) dg, (C.1)

where A is a constant relating ore content to volume. If we normalize A to


equal 1, then

d(nr2) = dF(g).

F(g) = m2, since 0 = F(g,) = F(g(0)). Average grade is given by

G=(l/lrr’)~2axg(x)dx=(l/r2)~F~‘(~y)dy, (C.2)
0 0

c= -2ri(G/r2) +(l/r2)2riFp ‘(nr2) = -2(i/r)(G-g),

-2(i/r)=(l -R)(d/G),
R.D. Cairns, Geological i@uences, metal prices and rationality 165

where we have observed that rent per unit ore output is G-g, so that rent
per unit metal output is 1 -g/G=R.
From (C.l),

For estimation purposes, we assume that grade is distributed according to


the formula given by Lasky and Musgrove:

fk)=(l/Wexp(-g/K),

F(g)= exp( -g/K) - exp( -g,lK).


Then

gf(g)lQg) = (g/K) exp ( -g/WlCev ( -g/K) - exp( -g,lWl.

We can also integrate (C.2) to find G:

G=KCU +g/K)ew(-g/K)-(1 +g,lK)exp(-g,lK)l/Cexp(-glK)

-exp ( -gmlW1,

G/K = 1+ CCWK)exp( -g/W -k,lK) exp( -gmlKMw ( -_glW

-ev ( -s,/K)ll

= 1+ CWWev ( -dWlCexp ( -g/K) - exp( -g,lK)ll~

for large g,/K. (We note that Musgrove interprets K as the average grade,
down to grade zero, in the deposit. It would be common to lind g, close to 1
and K very small.) Thus,

S/g= CKIR(G- K)l@;IGb

Appendix D: Description of the tests

Test 1

H,: b, > 1.

H,:b,Sl.
166 R.D. Cairns, Geological influences, metal prices and rationality

Test statistic: (6, - l)/SE(&)

as compared with t-statistic, where SE(b^,) is the estimated standard error of


b, in the second equation.

Test 2

H,:b;=b,,b;=b,.

H,:b; #b, and/or b; # 6,.

Test statistic: T(ln s”- In S)

to be compared with chi-square with two degrees of freedom, where

T is the total number of observations in both equations,


S is the unconstrained residual sum of squares for both equations when
normalized for the variances,
S is the constrained residual sum of squares for both equations when
normalized for the variances.

Test 3

H,: b;/b; = b,/b, = 1 6=b3

HI: one or more of the hypotheses noted above are not true.

Test statistic: T(ln s”- In S)

to be compared with chi-square with three degrees of freedom, where

T and S are the same as in Test 2,


s” is the constrained residual sum of squares for both equations when
normalized for variances.

Test 4

H,: b, = b,b,.

H,: b, # b,b,.

Test statistic: T,(ln s”, -In S,)

to be compared with chi-square with one degree of freedom, where


R.D. Cairns, Geological influences, metal prices and rationality 167

TI is the number of observations in the first equation,


S,1 is the unconstrained residual sum of squares in the first equation,
S, is the constrained residual sum of squares in the first equation.

Test 5

H,:b,=l.

H,:b,#l.

Test statistic: (6, - 1)/%(&J

to be compared with t-statistic, where SE(6,) is the estimated standard error


of b, in the first equation.

Appendix E: Test results

Table E.l
Gold mines included in this study.
~~_______.
Name Sample period”
Series A
Aunor Gold Mines Ltd. 196&71
Cariboo Gold Quartz Mining Co. Ltd. 1952-64
McIntyre Porcupine Mines Ltd. 1960-71
Upper Canada Mines Ltd. 196&71
Renabie Mines Ltd. 1958-69
Preston East Dome Mines Ltd. 195667
Paymaster Consolidated Mines Ltd. 1952-63
Madsen Red Lake Gold Mines Ltd. 1962-7 1

Series B (small-sample mines)


1’ Lamaque Mining Co. Ltd. 1954-56,
1962-63,
1969970
2’ Hollinger Consolidated Gold Mines Ltd. 1952-55,
1958,
1962-63,
1967-68
3’ Pamour Porcupine Mines Ltd. 1954,
1962-64,
1967-7 1
4’ Sigma Mines (Quebec) Ltd. 1952-56,
1963,
1969-70
“A lagged value for all these data points was available for use in
computing first differences, etc.
168 R.D. Cairns, Geological influences, metal prices and rationality

Table E.2a
Least-squares regression results.

Eq. (1)
___-.- Eq. (2)
~._
Mine RZ D.W SEE R2 D.K! SEE
__-
-1 0.315 1.46 0.074 0.465 1.76 0.018
2 0.108 2.29 0.404 0.150 1.99 0.455
3 0.714 1.58 0.090 0.610 1.79 0.125
4 0.722 1.73 0.083 0.769 2.09 0.089
5 0.671 2.92 0.122 0.753 2.63 0.125
6 0.550 2.01 0.159 0.789 1.93 0.129
7 0.604 2.38 0.119 0.681 2.40 0.126
8 0.442 2.23 0.253 0.521 2.52 0.302
1’” 0.819 2.18 0.096 _
2’ 0.908 2.04 0.069 _
3’ 0.669 0.95 0.618 _
4’ 0.938 2.01 0.019 - - _
_-..-
“Corrected for AR( 1).

Table E.2b
Estimated parameters?

Mine Int. b, b, b, bo
__._-.
Method: instrumented variables”
1 -0.10 0.11 -0.30 124.00 - 38.1
-1.09 0.96 -0.87 1.75 - 1.69
2 -0.27 0.20 -0.32 81.0 -33.6
-0.41 0.81 -0.85 0.65 - 0.78
3 0.47 0.29 - 1.33 774.00 -222.00
0.40 0.54 -0.50 0.74 -0.76
4 -0.00 -0.16 0.55 36.5 - 14.6
-0.03 -0.65 0.74 0.17 - 0.22
5 0.23 -0.13 0.39 - 39.3 10.3
1.30 -1.50 0.91 -0.55 0.69
6 -0.58 -0.22 1.76 21.5 - 4.99
-3.08 -- 2.85 3.34 0.26 -0.31
7 0.14 -0.11 0.44 202.00 - 34.2
0.41 -0.99 0.80 0.58 -0.52
8 -0.46 -0.20 0.95 -45.5 19.3
-1.11 - 1.06 1.67 -0.39 0.49

Method: Instumental variables/ridge


1 -0.02 -0.00 -0.00 1.88 0.45
2 - 0.06 0.00 -0.00 - 2.58 - 1.04
3 - 0.07 0.01 0.04 - 2.05 - 0.74
4 *x0.01 0.00 0.01 -0.84 -0.29
5 0.15 -0.04 0.04 3.60 1.28
6 -0.15 0.00 0.15 0.68 0.09
7 0.02 -0.03 0.14 10.2 1.29
8 -0.46 -0.10 0.63 4.71 1.34

?-ratios are given beneath the estimates.


R.D. Cairns, Geological influences, metal prices and rationality 169

Table E.2c
Estimated parameters; concluded.

Mine Int. b; h; b, b, b, b,
Method: Instrumental variables”
1 - 0.09 - 0.03 0.16 3.14 - 0.97 - 127.0 38.9
-0.27 -0.18 0.20 0.13 -0.14 -0.44 0.42
2 -0.74 0.33 -0.42 2.84 - 0.75 - 84.2 33.8
-0.83 1.03 -0.88 0.2 1 -0.16 -0.66 0.79
3 0.49 0.04 -0.49 134.0 - 39.9 - 1075.0 314.0
0.88 0.10 -0.34 1.25 - 1.25 -1.88 1.91
4 - 0.04 - 0.47 1.47 129.0 - 39.7 -918.0 289.0
-0.21 - 1.48 1.55 1.14 - 1.15 -1.16 1.17
5 0.25 -0.11 0.28 - 12.1 2.58 6.24 -2.58
0.83 -0.64 0.32 -0.19 0.20 0.04 -0.07
6 0.13 0.12 -0.88 - 30.7 8.60 28.7 -1.19
0.14 0.26 -0.25 -0.81 -0.81 0.24 - 0.06
7 -0.05 -0.25 1.41 - 7.05 1.43 - 322.0 55.2
-0.03 -0.13 0.11 -0.05 0.06 - 0.08 0.08
8 -0.44 -0.15 0.82 - 2.76 0.01 52.3 -18.6
-0.87 -0.71 1.17 - 0.07 0.00 0.19 - 0.22

Method: Instrumental variables/ridge


1 0.01 0.01 -0.44 1.53 -0.16 - 9.20 1.08
2 0.03 0.01 -0.01 0.06 0.02 1.97 0.95
3 -0.16 0.01 0.06 0.14 0.04 4.32 1.48
4 - 0.06 0.00 0.02 0.18 0.06 3.22 1.01
5 0.12 -0.03 0.01 0.36 0.09 -2.87 -1.01
6 -0.49 -0.19 1.46 - 3.70 1.04 - 14.1 3.98
7 0.01 -0.00 0.01 0.05 0.01 -0.73 -0.15
8 -0.18 -0.00 0.14 -0.12 -0.15 0.67 0.07
V-ratios are given beneath the estimates.

Table E.3
Test results, using OLS.

Mine Test-l Test-2 Test-3 Test-4 Test-5


1 0.41 1.34 6.02 4.55 -1.73
2 0.77 0.30 1.70 0.70 -0.81
3 1.91 0.31 7.35 0.92 -0.77
4 1.17 1.16 3.41 0.06 -0.23
5 -0.10 0.04 0.57 0.43 0.63
6 -0.10 1.21 1.77 0.22 -0.37
7 0.08 0.05 1.00 0.59 -0.54
8 -0.23 0.11 1.32 0.24 0.46
170 R.D. Cairns, Geological injluences, metal prices and rationality

Table E.4
Test results using instrumental variables.

Mine Test-l Test-2 Test-3 Test-4 Test-5


-0.38 0.20 1.64 0.73 -0.59
0.55 0.15 0.40 0.16 -0.44
0.24 0.66 0.65 0.25 0.10
- 1.21 2.02 5.82 1.27 0.79
0.52 0.12 2.51 0.60 0.79
0.12 3.41 8.43 0.03 0.15
-0.83 1.37 2.16 0.02 - 0.02
-0.45 0.52 1.97 0.59 0.64

Table E.5
Test results using instrumental variables;
cross-equation technique.

Mine Test-2 Test-3


6.63 2.46
0.12 2.40
3.46 7.60
0.27 1.62
0.81 1.06
0.19 0.53
0.12 1.52
0.31 0.69

Table E.6
Test results for small-sample mines.

Mine Test- 1 Test-4


1’ 0.86 3.03
2’ - 1.68 6.41
3’ - 1.05 1.90
4 -0.15 0.11

References
Barnett, Harold J. and Chandler Morse, 1963, Scarcity and growth (Johns Hopkins Press,
Baltimore, MD).
Boldt, Joseph R., 1967, The winning of nickel (Longmans, Toronto, Ont.).
Cairns, Robert D., 1981, An application of depletion theory to a base metal: Canadian nickel,
Canadian Journal of Economics 14, Nov., 635-648.
Cairns, Robert D., 1986a, More on depletion in the nickel industry, Journal of Environmental
Economics and Management 13, no. 2, June, 93-98.
Cairns, Robert D., 1986b, A model of exhaustible resource exploitation with Ricardian rent,
Journal of Environmental Economics and Management 13, Dec., 313-324.
Dasgupta, P.S. and G.M. Heal, 1979, Economic theory and exhaustible resources (Cambridge
University Press, Cambridge).
Energie, mines et ressources Canada, 1976, Rapport final concernant I’application de la loi
d’urgence sur I’aide a I’exploitation des mines d’or, Ottawa.
Farrow, Scott, 1985, Testing the efficiency of extraction from a stock resource, Journal of
Political Economy 93, no. 3, 452-487.
R.D. Cairns, Geological influences, metal prices and rationality 171

Farrow, Scott and Jeffrey A. Krautkraemer, 1987, Metal supply and grade selection: Responses
to anticipated and unanticipated price changes, Mimeo. (Carnegie-Mellon University,
Pittsburgh, PA).
Farzin, Y. Hossein, 1984, The effect of the discount rate on depletion of exhaustible resources,
Journal of Political Economy 92, no. 5, 841-851.
Frank, Jeff and Mark Babunovic, 1984, An investment model of natural resource markets,
Economica 51, 83-9.5.
Halvorsen, Robert and Tim R. Smith, 1984, On measuring natural resource scarcity, Journal of
Political Economy 92, no. 5, 954-964.
Heal, GeolTrey and Michael Barrow, 1980, The relationship between interest rates and metal
price movements, Review of Economic Studies XLVII, 161..181.
Heal, Geoffrey and Michael Barrow, 1981, Empirical investigation of the long-term movement of
resource prices: A preliminary report, Economics Letters 7,95-103.
Hotelling, Harold, 1931, The economics of exhaustible resources, Journal of Political Economy
39, 137-175.
Krautkraemer, Jeffrey A., 1988, The cut-off grade and the theory of extraction, Canadian Journal
of Economics XXI, no. I, 146-160.
Lasky, S.G., 1950a, Mineral resource appraisal by the U.S. geological survey, Colorado School
of Mines Quarterly 4.5, Jan., l-29.
Lasky, S.G., 1950b, How tonnage and grade relations help predict ore reserves, Engineering and
Mining Journal 157, April, 81-85.
Lasserre, Pierre, 1983, L’aide aux mines d’or ou les silences du fant6me de Bretton Woods,
Canadian Public Policy/Analyse de politiques IX, no. 4, Dec., 446457.
Livernois, John R., 1987, Empirical evidence on the characteristics of extractive technologies:
The case of oil, Journal of Environmental Economics and Management 14, no. 1, March,
72-86.
Livernois, John R. and Russell S. Uhler, i987, Extraction costs and the economics of
nonrenewable resources, Journal of Political Economy 95, no. 1, 195203.
Miller, Merton H. and Charles W. Upton, 1985, A test of the Hotelling valuation principle,
Journal of Political Economy 93, no. I, l-25.
Mishkin, Frederick S., 1982, Does anticipated monetary policy matter? An econometric
investigation, Journal of Political Economy 90, no. 1, 22-51.
Musgrove, Philip A., 1971, The distribution of metal resources (tests and implications of the
exponential grade-size relation) Proceedings (American Institute of Mining, Metallurgical
and Petroleum Engineers, Council of Economics, New York) 349-419.
Musgrove, Philip A., 1976, Mathematical aspects of the grade-tonnage distribution of metals, in:
Wiiliam A. Vogeley, ed., Economics of the mineral industries, 3rd ed., (American Institute of
Mining, Metallurgical and Petroleum Engineers, New York) 192-202.
Paish, F.W., 1938, Causes of change in gold supply, Economica 5, 379-409.
Pindyck, Robert S., 1978, The optimal exploration and production of nonrenewable resources,
Journal of Political Economy 86, no. 5, 841-861.
Pindyck, Robert S., 1982, Jointly produced exhaustible resources, Journal of Environmental
Economics and Management 9, Dec., 291-303.
Slade, Margaret E., 1982, Trends in natural-resource commodity prices. An analysis of the time.
domain, Journal of Environmental Economics and Management 9, June, 122-137.
Slade, Margaret E., 1988, Grade selection under uncertainty: Least cost last and other
anomalies, Journal of Environmental Economics and Management IS, June, 189-205.
Smith, V. Kerry, 1981, The empirical relevance of Hotelling’s model for natural resources,
Resources and Energy 3, 105-l 17.
Solow, Robert M., 1974a, The economics of resources or the resources of economics, American
Economic Review LXIV, Papers and proceedings, May, 1-14.
Solow, Robert M., 1974b, Intergenerational equity and exhaustible resources, Review of
Economic Studies. Symposium issue.
Stollery, Kenneth R., 1983, Mineral depletion with cost as the extraction limit: A model applied
to the behaviour of prices in the nickel industry, Journal of Environmental Economics and
Management 10, June, 151-165.

Das könnte Ihnen auch gefallen