Long Taylors Rule With Errata
Long Taylors Rule With Errata
Long Taylors Rule With Errata
In 1977, Taylor proposed a constant elasticity model relating capacity choice in mines to
reserves. A test of this model using a very large (n = 1,195) dataset confirms its validity but
obtains significantly different estimated values for the model coefficients. Capacity is
somewhat inelastic with respect to reserves, with an elasticity of 0.65 estimated for open-pit
plus block-cave underground mines and 0.56 for all other underground mines. These new
estimates should be useful for capacity determinations as scoping studies and as a starting
point for feasibility studies. The results are robust over a wide range of deposit types, deposit
sizes, and time, consistent with physical constraints on mine capacity that are largely independent of technology.
KEY WORDS: Mine capacity, mine planning, feasibility studies, scoping studies, Taylors rule,
Hotellings rule.
INTRODUCTION
Taylor (1977, 1986) introduced an empirical
relationship between mine capacity and reserves,
informally known as Taylors Rule, as an aid to
determining installed capacity in mine planning.
Based on his experience in mine design, Taylor
suggested that installed mine capacity is roughly
proportional to the three-quarters power of estimated tonnage of reserves. He later confirmed
this with a sample of nearly 30 mining projects.
McSpadden and Schaap (1984) obtained the same
result with a sample of 45 open-pit copper mines.
Lasserre (1985), in an econometric study of some
North American metal mines, likewise found that
mine capacity increases proportionately less than
reserves. Lasserres (1985) estimated elasticities for
capacity with respect to reserves varied from 0.58 to
1
57
1520-7439/09/0300-0057/0 2009 International Association for Mathematical Geology
Long
58
DATA
Data on reserves, capacity, mining method, and
related information were obtained from a proprietary database, the Mining Operations Report,
developed by Geomineinfo of Tucson, Arizona. The
database contains comprehensive technical and historical information on more than 800 operating or
recently operating mines, principally in Americas
and Australia. A total of 539 mines were used in this
study, of which 342 are open-pit and 197 are
underground mines. The distribution of these mines
according to primary commodity is shown in
Table 1. Most of the mines (59%) are gold mines,
followed by copper (23%) and zinc-lead (12%).
Many of the mines in the database have
undertaken at least one significant expansion in
capacity during the life of the mine. On the
assumption that any expansion in capacity must
correspond to an expansion in reserves, individual
data points for this study were defined as a capacity
and its corresponding reserve for a newly opened
mine or an expansion thereof. The total number of
data points used in this study, over 1,200, thus
exceeds the number of mines from which data were
obtained. Generally, the reserve figure used was the
latest reported prior to completion of new capacity
or an expansion. In a very few cases, the appropriate
reserve estimate was reported well before or sometime after new capacity was installed.
A few open-pit gold and copper mines treated
higher and lower-grade ores, or oxide and sulfide
ores, by different methods in different facilities and
reported separate reserves for each ore type. These
mines, totaling 29, were treated as two mines in
one, thus increasing the number of open-pit
mines used from 342 to 375. Table 2 shows the
distribution of the total number of mines used in this
study by mining method and number of capacity
expansions.
These methods of handling the data may be
illustrated by the Sleeper gold mine in the Awakening Mining District, Humboldt County, Nevada,
USA. The mine was opened in 1986 as an open-pit,
heap-leach operation with an initial processing
capacity of 1,800 metric tons per day. The heapleach operation was subsequently expanded three
times to an ultimate capacity of 14,900 metric tons
per day. A vat-leach mill was installed in 1987 to
treat higher-grade oxide ores with an initial capacity
of 900 metric tons per day. The mill was subsequently expanded five times, reaching 2,500 metric
Open Pit
Underground
Total
94
226
5
1
0
1
7
0
8
342
31
93
3
3
1
2
9
1
54
197
125
319
8
4
1
3
16
1
62
539
Open Pit
Underground
Total
157
93
47
31
14
12
9
4
2
3
3
83
41
27
17
12
3
3
3
2
2
0
240
134
74
48
26
15
12
7
4
5
3
MODEL ESTIMATION
Taylors original model (Taylor, 1986, Eq. 3)
Tonnes per day 0:014 Expected tonnes)0:75
can be re-written as
C bT a
59
Table 3. Model Estimation Results
Model
Observations
R2
Coefcient
796
0.870
Underground
400
0.821
a
b
a
b
Estimate
0.649
2.093
0.563
1.214
Lower
Upper
0.631
2.392
0.538
1.603
0.666
1.794
0.589
0.824
Observations
F
Coefcient Standard
t
Ratio
Error
Ratio
796
5334
400
1824
so that ordinary least-squares methods for estimating the coefficients a and b could be used. The large
size of the database led to a prolonged effort in
outlier detection and exploratory analysis. Outliers
were investigated and found to comprise a handful
of data errors, some reporting problems, and a few
mines with unique circumstances. An example of the
latter was the Morenci copper mine in Arizona, an
open-pit mine-for-leach operation where about 93%
of the material mined in 2006 was treated by
leaching of run-of-mine or crushed ore. The large
capacity of this mine, wholly out of proportion with
its reserves, is thought to reflect the unique mining
and processing method used and a conservative
approach to reserve estimation. Most importantly, it
was found that underground mines, with the exception of block-cave mines, form a distinct group
which was estimated separately from open-pit plus
block-cave mines. Estimation results are shown in
Table 3.
All of the coefficients are highly significant
(Table 4) and residuals are normally distributed with
no apparent inhomogeneity of variance. Note that
the coefficient of greatest interest, a, in each model, is
significantly less than a = 0.75 value originally estimated by Taylor (1977). As expected, the value of a
for the underground model is significantly less than
that for the open-pit plus block-cave model. These
coefficients are <1, hence the capacityreserve relationship is relatively inelastic, or in other words,
there are decreasing returns to capacity with
increasing reserves. The good fit of the equation
suggests that the restrictive technological assumptions of the model, constant elasticity of scale, are not
a
b
a
b
0.009
0.152
0.013
0.198
73.03
13.75
42.71
6.12
MODEL EVALUATION
The models are subject to criticism in at least
two ways: (1) a possible lack of homogeneity due to
the mixture of many types of mineral deposits, and
(2) the assumption of constant elasticity. The two
objections are not unrelated. The data include mines
developed from 1908 to 2007, thus there might be
technological changes over time that result in technology-dependent values for the model coefficients.
Unique geologic characteristics of certain deposit
types may require unique technological solutions
that could also affect the value of the estimated
model coefficients. If these considerations are significant, subdividing the data by deposit type and
other criteria might result in better models.
The data for the open-pit plus block-cave and
underground models were first divided between
newly developed mines and expansions of mines.
The object was to test for significant differences in
the estimated parameter a between the open-pit plus
block-cave and underground mine models and
four new models derived by separating new and
expanded mines. No significant differences at the
5% level were found (Table 5). The full models are
applicable to new as well as expanded mines.
Long
60
Table 5. Test of Ho : a^ am Against the Alternative Ho : a^ 6 am ,
where ^
a is the Estimate of the Parameter a for a Subset of New or
Expanded Mines and am is the Estimate of the Parameter a for the
Full Models; am Equals 0.649 for the Open-Pit Plus Block-Cave
and 0.563 for Underground Models
Model
Open pit + block cavenew mines
Open pit + block cave expansions
Undergroundnew mines
Undergroundexpansions
Observations
t Ratio
338
459
157
243
0.624
0.641
0.556
0.551
1.71
0.71
0.36
0.72
Observations
t Ratio
291
331
83
0.668
0.636
0.656
1.13
0.65
0.23
87
0.564
0.03
132
0.506
1.97
31
0.601
0.89
t Ratio
0.547
0.793
0.551
0.677
0.484
0.476
0.304
0.571
2.19
1.68
1.06
0.70
1.30
0.70
1.94
0.14
Observations
t Ratio
51
63
198
241
59
51
45
87
81
28
0.583
0.640
0.658
0.612
0.637
0.584
0.565
0.530
0.536
0.518
1.91
0.35
0.48
2.02
0.33
0.80
0.06
1.00
0.86
1.00
61
Epithermal
AuX1, X2, X3
162
0.824
ALTERNATIVE MODELS
Epithermal
AuX1, X2
162
0.815
Epithermal
AuX1
Porphyry
CuX1, X2, X3
162
0.777
53
0.849
Porphyry
CuX1, X2
53
0.848
Porphyry
CuX1
Carlin
AuX1, X2, X3
53
0.795
79
0.726
Carlin AuX1, X2
79
0.723
Carlin AuX1
80
0.679
Cb
N
X
Xnan
n1
N
X
an ln Xn
n1
a1
a2
a3
b
a1
a2
b
a1
b
a1
a2
a3
b
a1
a2
b
a1
b
a1
a2
a3
b
a1
a2
b
a1
b
0.544
0.342
0.725
3.651
0.563
0.352
0.484
0.649
2.021
0.572
0.560
0.070
1.248
0.586
0.576
1.204
0.650
2.046
0.565
0.315
0.207
1.545
0.578
0.345
0.755
0.573
0.983
t
18.593
5.736
2.841
3.014
19.302
5.776
0.998
23.589
4.543
11.486
3.960
0.541
1.549
13.615
4.181
1.513
14.214
2.337
12.497
2.971
0.838
1.312
13.612
3.463
1.073
12.853
1.343
62
incorporate at least grade to simulate the cost distribution of known and undiscovered deposits.
DISCUSSION
There have been many attempts to derive
optimal mine capacity by various methods, including
modifications to the Hotelling (1931) model of
extraction from a finite resource (Cairns, 2001), net
present value criteria (Smith, 1997), and marginal
analysis (Sabour, 2002). None of the models obtains
results consistent with actual practice as empirically
observed in studies such as this one. All of the
proposed methods tend to specify optimal output
rates far in excess of observed capacity choices. This
is not surprising in that none of these methods
impose restrictions based on technology, which
suggest that there could be diseconomies to scale if
too large a capacity is selected.
The history of the mines examined for this study
show that, for sufficiently large deposits, reserves are
not fully delineated prior to commencement of
production, and that these mines often undergo
several phases of capacity expansion as reserves are
expanded over time. This phenomenon is consistent
with the hypothesis of technological and capital cost
constraints that limit actual capacity to levels much
lower than otherwise optimal. For a very large
deposit, fully exploring that deposit and developing
a mine to match, prior to realizing any revenue, is
probably too costly. Capacity expansions may also
be justified by an increase in metal prices or a
decline in costs such that lower-grade ores are added
to reserves. Cost-saving technical innovation is a
significant long-run factor in adding to reserves
(Long, in press). Capacity may also be expanded to
maintain a given metal output if grade is declining.
This study does not isolate the relative contributions
of these various factors, but does establish that
reserve additions and capacity expansion are common features of metal mining.
This study does show that Taylors empirical
relationship holds across time, deposit type, and
deposit size, with little or no significant variation in
the estimated coefficients. This remarkable fact
requires explanation. Taylor (1986) opined that
physical limitations on working a deposit are at
work. In the case of an open-pit mine, access can
only be from one direction, from the surface down,
and the amount of ore extracted from a pit is limited
by geometry of the ore body and other factors that
Long
affect pit design, such as rock strength. These factors
work together to require significant mining of rock
not of ore grade in order to access the ore. In the
case of an underground mine, with the significant
exception of block-cave mines, a large portion of the
ore reserve must be left in place during the main
phase of mining for roof support, and if ever
recovered, is removed only in the last stages of
mining. The depth and geometry of the ore put
limits on the number of working faces that can be
maintained at a time as well.
There may be a myriad of geotechnical factors
at work here, which bear further study. Comparison
with a better understood deposit typepetroleumis useful to understand this phenomenon.
Adelman (1993) and Cairns and Davis (2001) have
demonstrated that petroleum reservoir pressure
places practical limits on extraction rate or capacity.
Pressure declines as production advances, hence
individual well production declines over time,
regardless of any economic factors. Thus reservoir
pressure, a physical rather than an economic variable, is the limiting factor on production. A Taylors
rule for initial oil well capacity would have to
account for reservoir pressure as well as initial
reserves. We can infer from the results of this study
that physical factors governing access to ore limit
capacity for metal mines as well.
As Cairns and Davis (2001) show in the case of
petroleum, these physical limitations are the
scarce factor, not the size of the resource being
extracted. Hotelling (1931) proposed that the equilibrium price of an exhaustible resource, net of
marginal extraction costs, would rise at the rate of
interest. Such a phenomenon has not been empirically observed nor does it figure into the conscious
decision making of mine planners. Hotellings model
has a number of restrictive assumptions that, when
relaxed, often result in models that better fit with
observed reality. As Cairns and Davis (2001) show, a
physical constraint on production capacity substitutes a different kind of scarcity, reservoir pressure,
that completely changes the dynamics of pricing
mineral resources. Consistent with Adelman (1993),
Cairns and Davis (2001) show that the value of
petroleum reserves is only about half that predicted
by Hotellings model. This study shows that metal
mine capacity will increase proportionately less than
reserves, due to physical limitations on mine
capacity. Under Hotellings model, capacity would
ultimately be governed by the interest rate, there
being no physical limitations at play. Thus, we would
CONCLUSIONS
Taylors Rule relating capacity to estimated
reserves has been re-estimated using a large sample
of mines representing many countries with freemarket economies and several mineral commodities.
The empirical data were found to divide into two
groups, open-pit plus block-cave mines versus all
other underground mines, which were estimated
separately. Taylors three-quarter percent rule was
found to be too large in each instance, the proportional coefficient being 0.65 for open-pit plus blockcave mines and 0.56 for all other underground
mines. The good fit of the data to the model, and
several tests of alternative models, supports the
model assumption of constant elasticity of scale,
over the range of the data.
These new estimates, based on a large dataset
including underground mines, should be more reliable and have wider applicability than previous
estimates. For feasibility studies, the appropriate
model can be used to estimate an initial capacity
from which an iterative search for an optimal
capacity may begin. For scoping studies and other
rough calculations of probable mining costs, the
appropriate model can be used outright to estimate
capacity. The 95% confidence intervals estimated
give a useful range in capacity based on past and
current mining practice.
The constancy of the coefficient a over a range
of commodities, deposit types, deposit sizes, and
time, are consistent with Taylors intuition that
physical factors limit mine capacity. Put simply,
63
mine capacity can only grow about half to two-thirds
of the rate of increase in reserves, which is significantly less than what mine planners, using net
present value calculations, would likely choose. This
observed limitation on capacity renders Hotellings
rule irrelevant for mineral resources, just as it does
for petroleum. A better understanding of these
physical limitations to mine capacity will help
explain observed capacity choices, mineral pricing,
and the valuation of mineral deposits and mines.
REFERENCES
Adelman, M. A., 1993, The economics of petroleum supply;
Papers by M.A. Adelman, 19621993: MIT Press, Cambridge,
Massachusetts.
Cairns, R. D., 2001, Capacity choice and the theory of the mine:
Environ. Resour. Econ., v. 18, p. 129148. doi:10.1023/
A:1011114400536.
Cairns, R. D., and Davis, G. A., 2001, Adelmans Rule and the
petroleum firm: Energy J., v. 22, no. 3, p. 3154.
Camm, T. W., 1991, Simplified cost models for prefeasibility
mineral evaluations: U.S. Bureau of Mines Information
Circular 9298, 35 p.
Cox, D. P., and Singer, D. A., 1986, Mineral deposit models: U.S.
Geological Survey Bulletin 1693, 379 p.
Hotelling, H., 1931, The economics of exhaustible resources:
J. Petrol. Econ., v. 39, no. 2, p. 135179.
Lasserre, P., 1985, Capacity choice by mines: Can. J. Econ., v. 18,
p. 831842. doi:10.2307/135094.
Long, K. R., Economic life-cycle of porphyry copper mining, in
Ores and Orogenisis meeting, proceedings: Arizona Geological Society, Tucson, Arizona (in press).
Long, K. R., and Singer, D. A., 2001, A simplified economic filter
for open-pit mining and heap-leach recovery of copper in the
United States: U.S. Geological Survey open-file report 0100218, 21 p.
McSpadden, G. M., and Schaap, W. A., 1984, A test and comment
on Taylors rule of mine life: Bull. Proc. Australasian Institution Mines Metall., v. 289, no. 6, p. 217220.
Sabour, S. A. A., 2002, Mine size optimization using marginal
analysis: Resour. Policy, v. 28, p. 145151. doi:10.1016/
j.resourpol.2004.01.001.
Smith, L. D., 1997, A critical examination of the methods and
factors affecting the selection of an optimum production rate:
CIM Bull., v. 90, p. 4854.
Taylor, H. K., 1977, Mine valuation and feasibility studies, in
Hoskins, J. R., and Green, W. R., eds., Mineral industry
costs: 2nd ed, Northwest Mining Association, Spokane,
Washington, p. 117.
Taylor, H. K., 1986, Rates of working of minesa simple rule of
thumb: Trans. Institution Mining Metall., v. 95, sect. A,
p. A203204.
Errata for Long, Keith R., 2009, A test and re-estimation of Taylors empirical capacity-reserve
relationship: Natural Resources Research, v. 18, no. 1, p. 57-63.
R2
Observations
796
Coefficient
0.870
400
0.821
Estimate
95% Confidence
Interval
Lower
Upper
0.649
0.631
0.666
0.123
0.092
0.166
0.563
0.538
0.589
0.297
0.201
0.439
Observations
F
Ratio
Coefficient
Standard
error
t
ratio
796
5334
0.009
73.03
ln b
0.152
-13.75
0.013
42.71
ln b
0.198
-6.12
Underground
400
1824
Table 9. Extended model estimation results. All models are of open-pit mines in the United
States. The extended model includes the explanatory variables X 1 , reserve tonnage, X 2 , reserve
grade, and X 3 , a capital cost index (Bureau of Labor Statistics Index 112).
Model
Epithermal Au X1, X2, X3
Observations
R2
Coefficient
Estimate
162
0.824
a1
0.544
18.593
a2
-0.342
-5.736
a3
0.725
2.841
ln b
-3.651
-3.014
Epithermal Au X1, X2
Epithermal Au X1
Porphyry Cu X1, X2, X3
Porphyry Cu X1, X2
Porphyry Cu X1
Carlin Au X1, X2, X3
Carlin Au X1, X2
Carlin Au X1
162
0.815
162
0.777
53
0.849
53
0.848
53
0.795
79
0.726
79
0.723
80
0.679
C b X nan
n 1
a1
0.563
19.302
a2
-0.352
-5.776
ln b
-0.484
-0.998
a1
0.649
23.589
ln b
-2.021
-4.543
a1
0.572
11.486
a2
-0.560
-3.960
a3
0.070
0.541
ln b
-1.248
-1.549
a1
0.586
13.615
a2
-0.576
-4.181
ln b
-1.204
-1.513
a1
0.650
14.214
ln b
-2.046
-2.337
a1
0.565
12.497
a2
-0.315
-2.971
a3
0.207
0.838
ln b
-1.545
-1.312
a1
0.578
13.612
a2
-0.345
-3.463
ln b
-0.755
-1.073
a1
0.573
12.853
ln b
-0.983
-1.343