Resources Policy 27 (2001) 107–117

www.elsevier.com/locate/resourpol

Labor productivity, costs, and mine survival during a recession夽

1,*
John E. Tilton
Colorado School of Mines, Golden, CO 80401, USA

Received 3 April 2001; received in revised form 9 June 2001; accepted 13 June 2001

Abstract

The ability of a mine to survive cyclical downturns depends, according to economic theory, largely on its variable production
costs. Since labor accounts for a sizeable share of the variable costs of mining, a mine that enters a recession with relatively high
labor productivity and that manages during the recession to raise its labor productivity should be more likely than other mines to
avoid cutbacks and closure.
The US copper industry over the 1975–90 period provides empirical support for this expectation. But surprisingly, it also suggests
that mine survival depends (a) more on labor productivity than variable costs, and (b) more on the ability of a mine to increase
its labor productivity once in a recession than on a high level of labor productivity at the start of a recession. An important factor
affecting the extent to which mines increase labor productivity once in a recession is the life expectancy of their reserves.  2001
Elsevier Science Ltd. All rights reserved.

Keywords: Copper mining industry; Labor productivity; Competitiveness

In 1975 the world copper mining industry fell into a
deep recession that persisted for over a decade. As
shown in Fig. 1, the average annual copper price on the
LME (London Metal Exchange) in 1997 dollars dropped
to 1.49 dollars a pound in 1975 from 2.72 dollars the
previous year. It then stayed at this relatively depressed
level for the next several years. After a modest increase
at the end of the 1970s, it plunged to new lows. By 1986,
the bottom of the decline, the real price of copper was
nearly 70% below the 1974 high. During the latter half
of the 1980s, the price of copper recovered, and by 1990
was at roughly the same level in real terms as in 1975.
The United States with many of the world’s marginal
copper mines was particularly hard hit by this severe
downturn in the global copper market. Profits fell
sharply, and by the early 1980s nearly all US mines were
losing money. Many were not even recovering their vari-
夽
An earlier version of this study was presented at the Eighth
Annual Meeting of the Mineral Economics and Management Society
held April 15–17, 1999, in Ottawa
* Tel.: +1-303-273-3485; fax: +1-303-273-3416.
E-mail address: jtilton@mines.edu (J.E. Tilton).
1
William J. Coulter Professor of Mineral Economics, and at the
time this study was conducted, Visiting Scholar at the Centro de
Mineria of the Pontificia Universidad Catolica in Chile.
able costs of production. The US industry petitioned the
government for protection against imports in 1978 and
then again in 1984. On both occasions its request was
denied.
Nevertheless, the US industry did survive. As
described elsewhere (Tilton and Landsberg, 1999), the
industry greatly reduced its costs and increased its labor
productivity. As a result, when the price recovered in
the late 1980s, it once again was profitable. Chile, it is
true, had become the world’s largest copper mining
country, a distinction the United States had held since
the beginning of the century. Still, the US industry
accounted for 22% of Western world copper production
in 1990, and its output was above its 1970 level and
considerably above its 1975 level.
Aggregate figures, however, paint an incomplete and
somewhat misleading picture. Not all the US copper
industry survived to enjoy the market recovery and the
price increases during the latter half of the 1980s. As
Table 1 shows, some 24 US mines produced 10,000 t or
more of copper in concentrate in 1975. By 1990, five
of these mines had ceased production completely, and
another six were producing very modest tonnages (under
4000 t). The other 13 mines remained important pro-
ducers—three had cut back their output (from 2 to 29%),
while ten had expanded their production (from 4 to
590%).

0301-4207/01/$ - see front matter  2001 Elsevier Science Ltd. All rights reserved.
PII: S 0 3 0 1 - 4 2 0 7 ( 0 1 ) 0 0 0 1 2 - 5
108 J.E. Tilton / Resources Policy 27 (2001) 107–117
Fig. 1. Average annual LME copper price in real (1997) US cents per pound, 1970–95. Source: US Geological Survey.
Why did 13 mines manage to survive—a number even
to grow—over the 1975–90 period, while others failed?
Here we focus on the contribution of labor productivity
in answering this question. The next section considers
why we expect labor productivity to affect a mine’s
ability to survive a recession. The third section then
examines empirically the relationship between labor pro-
ductivity and changes in mine output over the 1975–
1990 period. The fourth section turns to production
costs, presumably the critical link between productivity
and output. The fifth section focuses on the role of
reserves and mine size in explaining why some mines
greatly improved their productivity (and in turn their
prospects for survival) while others did not. Finally, the
sixth section highlights the major findings and examines
a few of their implications.
Labor productivity and survival: theory
According to conventional economic theory, a profit
maximizing or net-present-value maximizing firm with-
out any market power should remain in operation as long
as the market price remains at or above its average vari-
able costs of production.2 Variable costs cover all those
expenses that rise and fall in the short run with output,
in contrast to fixed costs associated with buildings,
equipment, and other long-run investments.
This means that a copper mine should remain in oper-
ation as long as the price of copper equals or exceeds
its variable costs. In the real world, of course, a mine
may deviate from this expected behavior as a result of
2
Where mining firms exploit a fixed stock of mineral resources,
variable costs should include user costs, that is the net present value
of the profits lost in the future as a result of producing one more unit
of output this period, rather than saving the nonrenewable resource
required for that unit of output for use in the future (Hotelling, 1931).
In the copper industry, however, there is little to suggest user costs
are positive, or that firms consider them in determining their copper
output. As a result, they are not considered further here.
government pressures or incentives, shut down and start
up costs, management’s expectations about future price
movements, and other discrepancies from the simplify-
ing assumptions of economic theory. Still, as the price
of copper falls during a recession, we normally expect
those mines with relatively high variable costs to shut
down before mines with low costs.
Conventional economic theory is the basis of com-
parative cost analysis, a technique widely used by min-
ing companies, consulting firms, and government agenc-
ies to construct short-run supply curves for metal and
other mineral commodities.3 The analysis entails esti-
mating average variable costs of production and
capacities for individual mines, and then ordering or
ranking mines according to their variable costs. It nor-
mally assumes a mine will close whenever price falls
below its variable costs, and then reopen when price
rises back to or above its variable costs, though in some
cases closure costs and other considerations that may
cause mines to deviate somewhat from the predicted
behavior are taken into account. While most comparative
cost studies are proprietary, some are available to the
public (see, for example, US Bureau of Mines, 1987;
Torries 1988, 1995).
Comparative cost analysis is used to assess the com-
petitiveness of undeveloped mineral deposits and the
survivability of existing operating mines. Like conven-
tional economic theory, it assumes that the costs of the
marginal producer, whose output is needed to satisfy
prevailing demand, determines the market price.
Experienced analysts of the metal industries, however,
have for some time noted that the cause-and-effect
relationship between costs and price is not just one-way
3
Comparative cost analysis can also estimate intermediate to long
run supply curves. This, however, requires estimates of average total
costs for both existing mines and known but undeveloped mineral
deposits. Moreover, the resulting supply curve is not truly long run,
since it cannot take into account the future discovery of new deposits
or the effects on costs of new technological developments.
 J.E. Tilton / Resources Policy 27 (2001) 107–117 109

Table 1
Output and labor productivity for 24 US copper mines, 1975 and 1990a

Outputb Productivityc
Mines 1975 1990 Growth 1975 1990f Growthf

Expanding minesd
Bagdad 20 136 590 20 102 414
Chino 53 145 172 60 91 51
Morenci 125 324 158 53 95 78
Ray 49 112 129 44 68 55
Tyrone 75 155 106 51 95 87
Bingham Canyon 247 371 50 31 153 394
Pinto Valley 66 88 34 59 77 31
San Manuel 109 142 30 22 36 63
Cyprus Miami 45 57 28 42 52 24
Sierrita 132 137 4 35 57 61
Contracting minesd
Butte 91 90 ⫺2 43 123 184
Missione 106 79 ⫺26 28 62 122
White Pine 71 51 ⫺29 13 24 82
Non-surviving minesd,f
Silver Bell 19 4 ⫺80 45 44 ⫺1
Mineral Park 27 2 ⫺93 34 14 ⫺59
Superior 44 3 ⫺94 18 15 ⫺18
Yerlinton 31 2 ⫺94 35 30 ⫺14
Bisbee 13 1 ⫺96 44 35 ⫺20
Esperanza 24 0 ⫺99 38 49 30
Continental 16 0 ⫺100 32 22 ⫺32
Ajo 33 0 ⫺100 42 29 ⫺31
Battle Mountain 20 0 ⫺100 30 21 ⫺31
Ruth McGill 31 0 ⫺100 16 14 ⫺12
Sacaton 20 0 ⫺100 39 45 16
All other minesh 75 98 30 g g g

Total industryh 1542 1995 29

Sources: Brook Hunt and Associates; Rio Tinto Mine Information System; US Mine Safety and Health Administration.
a
All US copper mines whose 1975 output equaled or exceeded 10,000 t or more of contained copper equivalent in concentrates are included
in this table with the exception of Twin Buttes. Although Twin Buttes’ 1975 output was 13,800 t of contained copper, it was excluded because
its 1975 production was abnormally low, causing its productivity for that year to be unusually low as well.
b
Output is measured in thousands of tons of copper equivalent contained in concentrate production. Output growth is the percent change in
output between 1975 and 1990.
c
Productivity is measured in tons of copper equivalent contained in concentrate produced per 1000 manhours of labor input. Productivity growth
is the percent change in productivity between 1975 and 1990.
d
Expanding mines survived the recession in the copper market during the 1975–90 period and even managed to increase their output. Contracting
mines survived the recession but suffered a decline in output. Non-surviving mines ceased to be significant producers in the sense that their output
fell below 4000 t of copper equivalent.
e
The Mission mine also includes the Pima mine.
f
Labor productivity reported for non-surviving mines for 1990 is actually for their last normal year of operation: 1975 for Ruth McGill and
Bisbee, 1976 for Battle Mountain, 1977 for Yerington, 1980 for Mineral Park, 1981 for Silver Bell, Superior, Esperanza, and Continental, and
1983 for Ajo and Sacaton.
g
Productivity data for All other mines are not available.
h
Output for All other mines is the contained copper in concentrate production, and does not include the copper equivalent of byproduct output.
Productivity for ‘All other mines’ is measured in tons of copper contained in concentrate per 1000 manhours of labor input, and does not take
into account the copper equivalence of byproduct output. ‘Total industry’ includes the copper equivalency of byproducts for all mines except those
included under ‘All other mines’.

(Crowson, 1984, 1998, pp. 139–141). While costs influ-
ence price, price can in turn influence costs. In particular,
a sharp fall in price that threatens a mine’s survival may
force managers to take difficult personnel decisions.
Such conditions may similarly make workers and labor
unions more willing to accept painful changes. Thus
mines in the bottom cost quartile at the start of a
recession may find their relatively favorable position
deteriorating as other mines successfully reduce their
costs. This suggests that survival depends both on a
mine’s relative cost position at the start of a recession
and on changes in that position as the recession pro-
ceeds.
Conventional economic theory stresses the importance
of the former and discounts the latter, as it presumes
firms are fully aware of their technological options, and
110 J.E. Tilton / Resources Policy 27 (2001) 107–117
are always maximizing profits. This implies firms are
operating as efficiently as possible, and so the scope for
reducing costs during a recession is limited. When cost-
reducing opportunities do arise, they are associated with
new technology rather than a fall in the market price.
New developments in the theory of the firm over the
past several decades, however, question the assumptions
of the conventional theory that firms maximize profits
and operate with known and given technology. These
developments include bounded rationality (Simon 1955,
1959), the behavioralist school (Hall and Hitch, 1939;
Cyert and March, 1963), the theory of agency (Stiglitz,
1974), slack and discretionary behavior (Marris, 1964;
Williamson, 1966), evolutionary theory (Nelson and
Winter, 1982), and X-efficiency (Leibenstein, 1966).
They raise the possibility that firms may have consider-
ably more opportunities to reduce costs once in a
recession than the conventional theory anticipates.
Costs presumably provide the link between labor pro-
ductivity and mine survival during a recession. Labor
productivity—both the level of labor productivity at the
start of a recession and changes once the recession is
underway—should affect a mine’s ability to maintain
operations by influencing its variable costs. Employee
compensation is an important component of variable
costs. For example, Codelco, the large state-owned cop-
per mining company in Chile, attributes nearly 40% of
its variable (cash) costs to worker compensation in
recent years. When payments for services to third parties
are added in, an unknown part of which is also for
worker compensation, this figure rises to 64%. For US
copper companies, the US Bureau of Census (1992)
reports that employee payroll varies between 55 and
90% of the cost of supplies (after subtracting capital
expenditures) over the 1972–92 period. This suggests
that labor accounts for a third to a half of variable costs.
As a result, when labor productivity rises over time
allowing a mine to produce a ton of copper with fewer
manhours, variable costs should decline. Similarly, at
any point in time, variable costs should be lower for
mines with relatively high productivity.
Labor productivity and survival: evidence
Table 1, as noted earlier, identifies the 24 significant
US copper mines in 1975, including those that survived
and those that did not. This table also provides for each
of these mines the following information:
앫 Mine output measured in thousands of tons of con-
tained copper equivalent for 1975 and 1990, along
with the percentage change in output over the 1975–
90 period,
앫 Mine labor productivity measured in tons of output
per thousand manhours of labor for 1975 and 1990,
along with the percentage change in productivity over
the 1975–90 period.
Many copper mines produce valuable byproducts,
such as gold, silver, and molybdenum, in addition to
copper. Contained copper equivalent output includes
both the copper contained in the ore mined plus the cop-
per that the associated byproduct output could purchase
assuming copper and byproducts are valued on the basis
of their average real prices over the 1975–90 period.4
The contained copper equivalent output data, provided
by Brook Hunt and Associates and the Rio Tinto Mine
Information System, are originally from company annual
reports and other sources. The labor productivity data
are from reports compiled by the US Mine Safety and
Health Administration (MSHA), a rich source of infor-
mation on individual mines in the United States.
Productivity figures for 1990 are not available for
mines that closed before that year, and tend to be seri-
ously distorted for mines producing very little output in
1990. The 1990 figures in Table 1 for these mines show
productivity during their last year of normal operation.
According to the economic reasoning reviewed earl-
ier, the copper mines that survived the recession of the
late 1970s and 1980s should on average have enjoyed
higher 1975 labor productivity and greater percentage
increases in productivity over the 1975–90 period than
those mines that largely or completely ceased operations.
This should be especially so for the mines that increased
their output over this period.
Table 2 confirms that this is the case. It divides the
24 mines producing in 1975 into three groups—the
expanding mines (the 10 mines that survived and
enjoyed an increase in output over the 1975–90 period),
the contracting mines (the three mines that survived but
suffered a reduction in output), and the non-surviving
mines (the 11 mines whose decline in output was so
great that they were no longer significant copper pro-
ducers by 1990). This table then shows for each group
their output growth over the 1975–90 period, their labor
productivity in 1975, and their increase in labor pro-
ductivity over the 1975–90 period. As expected, the
expanding mines had the highest labor productivity in
1975 and the greatest increase in this variable over the
following 15 yr. Labor productivity in 1975 for non-sur-
viving firms, somewhat surprisingly, was above that of
the contracting mines. The subsequent growth in labor
productivity for this group, however, was actually nega-
tive, and far less than either of the other two groups.
4
In determining the copper equivalence of byproduct output, some
studies use the real prices for copper and byproducts for each individ-
ual year rather than using real prices averaged over the period under
consideration. This approach is less appropriate for the analysis here
since year-to-year fluctuations in metal prices can significantly affect
copper equivalent output and labor productivity.
 J.E. Tilton / Resources Policy 27 (2001) 107–117 111

Table 2 PerQi⫽ −49.0 +0.93 PerPi +ei R2=0.50

Average output, labor productivity, and cost performance for (2)
expanding, contracting, and non-surviving US copper mines, 1975–90a (−1.97) (4.92)
PerQi⫽ −138.2 +2.38 P75i +0.97PerPi +ei R2=0.52
Performance Expanding Contracting Non-
mines mines survivorsd (−1.99) (1.74) (5.17)
Output growth 1975–90 (%) 81 ⫺18 ⫺96 (3)
1975 productivity (t/1000 h) 36 24 28
Where:

Productivity growth 1975– 125 124 ⫺19 PerQi percent change in output for the ith mine over
90 (%)d the 1975–90 period,
P75i Labor productivity in 1975 for the ith mine, and
1975 Cash costs (cents per 154 165 160 PerPi percent change in productivity over the 1975–
pound)b 90 period at the ith mine.
Cash costs growth 1975–90 ⫺42 ⫺19 23
For these and subsequent equations, ei is the observed
(%)b,d
error term, R2 is the adjusted coefficient of determi-
1975 breakeven costs (cents 116 146 116 nation, and the figures in parentheses below the coef-
per pound)c ficient estimates are their t-statistics. All equations were
estimated on the basis of 24 observations, one for each
Breakeven costs growth ⫺29 ⫺26 ⫺6
of the mines identified in Table 1.
1975–90 (%)c,d
These results, particularly those for Eq. (3), provide
1975 average output (1000 92 89 25 further support for the hypothesis that the level of labor
t)a productivity at the start of a recession and changes dur-
ing the recession are important determinants of a mine’s
1975 average reserves 558 126 34
ability to survive market downturns. Both variables in
(million t)
Eq. (3) have the expected sign. Labor productivity in
1975 Average reserve life 47 10 9 1975 (P75i) is statistically significant at the 95% prob-
(yr)e ability using a one-tailed test, while the change in pro-
ductivity (PerPi) is significant at the 99% probability
Sources: Output and productivity data: Table 1 and the sources
level. The adjusted coefficient of determination (R2) sug-
cited there. Cash costs, adjusted breakeven costs, reserves, and
grade of reserves: Rio Tinto Mine Information System. gests that together these two variables can explain over
a
See Table 1 for an explanation of how output and productivity 50% of the differences in mine growth.
are measured. This table also defines expanding, contracting, and non- Of equal or perhaps more interest, the results indicate
surviving mines, and identifies the mines in each of these groups. that a mine’s ability to increase its productivity during
b
Cash costs are in real (1997) US cents per pound. As noted in
a recession may be considerably more important for sur-
the text, they cover all the expenses of mining and processing through
to the refined metal stage minus capital costs (specifically, vival than its level of productivity at the beginning of
depreciation, amortization, and interest on external debt). Cash costs the recession. According to Eq. (1), productivity at the
typically include expenditures for labor, materials, energy, and contract start of the recession by itself can explain very little of
services of third parties. the differences in mine growth. Mines with high pro-
c
Breakeven costs are also in real (1997) US cents per pound. They
ductivity in 1975 did not expand significantly more over
are actually adjusted breakeven costs, which are cash costs minus any
revenues received for coproducts and byproducts, minus the difference, the 1975–90 period than mines with low productivity.
if any, between a mine’s reported revenues per pound of copper and Yet the results for Eq. (3) indicate that high pro-
the world copper price. ductivity in 1975 did actually contribute to growth, once
d
Data for 1990 reported for labor productivity, cash costs, and the effects of differences in productivity growth over the
breakeven costs for non-surviving mines are actually for their last nor-
1975–90 period were taken into account. The expla-
mal year of operation: 1975 for Ruth McGill and Bisbee, 1976 for
Battle Mountain, 1977 for Yerington, 1980 for Mineral Park, 1981 for nation for this apparent inconsistency between Eq. (1)
Silver Bell, Superior, Esperanza, and Continental, and 1983 for Ajo and Eq. (3) lies in the fact that mines with high pro-
and Sacaton. ductivity in 1975 tended to increase their productivity
e
Reserve life for each mine is calculated by dividing the product less over the next 15 yr than mines with low productivity
of its reserves and the grade of its reserves by its 1975 output.
in 1975. This is confirmed by the simple correlation
To explore the relationship between survival and labor coefficient between these two variables, which is a nega-
productivity further, we estimated with ordinary least tive 0.16. As a result, the estimated coefficient for the
squares the following equations: independent variable in Eq. (1) is not only picking up the
positive influence on mine growth of high productivity in
PerQi⫽ −25.1 +0.90 P75i +ei R2=−0.04 1975, but also the negative effect of relatively modest
(1)
(−0.26) (0.36) productivity growth thereafter.
112 J.E. Tilton / Resources Policy 27 (2001) 107–117
Why mines with high productivity in 1975 should
have subsequently experienced less productivity growth
is not entirely clear. Perhaps mines with low productivity
were under greater pressure, if they wanted to survive,
to increase their productivity. Another possibility is that
mines with low productivity were more lax in their oper-
ations prior to 1975, and so had more readily apparent
opportunities to enhance their productivity.
Although interesting, Eqs. (1)–(3) should be treated
with caution for several reasons. First, the sample set is
small, only 24 mines, and so the results may be unduly
sensitive to the performance of one or several mines.
Second, though the data suggest the assumed linear
relationship between the dependent variable (output
growth) and the independent variables (level and change
in productivity) is reasonable, the actual relationship
may in fact be more complicated.5
Third, ordinary least squares require that cause-and-
effect flow in one direction only, from the independent
variables to the dependent variable. If the dependent
variable influences any of the independent variables, the
results suffer from simultaneity bias. While growth in
mine output presumably does not influence mine pro-
ductivity in 1975, it may affect the growth in mine pro-
ductivity over the 1975–90 period. An increase in output,
in particular, may produce economies of scale that cause
labor productivity to rise.6
The role of costs
The critical link between labor productivity and mine
output, according to the economic theory reviewed earl-
ier, is variable costs. Mines with high labor productivity
should generally have lower labor costs and hence lower
variable costs of production. This, in turn, should reduce
the need to cease or curtail production when market
prices fall during a recession. This suggests that the
relationship between labor productivity and variable
costs on one hand and the relationship between variable
5
Squared values of the independent variables were added to Eqs.
(1)–(3) in an attempt to determine if the dependent variable (output
growth) rose at an increasing or decreasing rate as the independent
variables (labor productivity in 1975 and growth in labor productivity)
increased. The results provide little evidence of a non-linear relation-
ship. The coefficient for the squared value of P75 (labor productivity
in 1975) was statistically significant in both Eq. (1) and Eq. (3). How-
ever, the estimated coefficients for the modified equations indicate that
over almost half of the sample range an increase in labor productivity
in 1975 is associated with a decrease in mine output growth over the
1975–90 period, a most plausible result.
6
We did not apply two stage least squares (2SLS) or other instru-
mental variable (IV) techniques, which are often used in place of ordi-
nary least squares (OLS) when simultaneity is present, in part because
no appropriate instrumental variable is readily apparent for the growth
in mine productivity (PerPi) and in part because these techniques elim-
inate simultaneity bias only in large samples.
costs and mine growth on the other should be as strong
or stronger than the relationship we have already
observed between labor productivity and mine growth.
This section examines whether this is actually the case
for our sample of 24 US copper mines. The cost data
are from the Rio Tinto Mine Information System, a pro-
prietary source of annual information on metal mines
around the world going back to 1970. It contains two
cost series closely related to variable costs—cash costs
and adjusted breakeven costs.
Cash costs cover all the expenses of mining and pro-
cessing through to the refined metal stage minus capital
costs (specifically, depreciation, amortization, and inter-
est on external debt). Cash costs typically include expen-
ditures for labor, materials, energy, and contract services
provided by third parties.
Breakeven costs are cash costs minus any revenues
received for coproducts or byproducts, such as gold, sil-
ver, or molybdenum, produced along with copper.
Adjusted breakeven costs go one step further and sub-
tract from breakeven costs the difference, if any,
between a mine’s reported revenues per pound of copper
and the world copper price. Thus, adjusted breakeven
costs reflect the lowest market price for refined copper
at which mines can operate without cash losses.
Productivity and costs
Considering first the relationship between labor pro-
ductivity and costs, we expect this relationship to be
negative. The higher a mine’s productivity at any parti-
cular time compared to its competitors, the lower its
costs should be. The more a mine’s labor productivity
increases over time, the more its costs should fall.
Table 2 indicates that the US copper industry during
the 1975–90 period provides some support for these
expectations. The 10 expanding mines on average have
greater labor productivity and lower cash costs in 1975
than the two groups of mines whose output fell between
1975 and 1990. Somewhat surprisingly, however, there
is no difference between these 10 mines and the 11 non-
surviving mines in the case of adjusted breakeven costs.
The three contracting mines have higher cash costs and
adjusted breakeven costs in 1975 than the eleven non-
surviving mines. Though surprising, this is consistent
with the fact they had lower labor productivity than the
latter in 1975.
Support for our expectations is stronger in the case of
changes over time. The 10 expanding mines on average
more than double their productivity, and reduced their
cash and adjusted breakeven costs substantially. The
group of 11 mines that did not survive, on the other
hand, suffered a decline in productivity while their cash
costs rose and their adjusted breakeven costs fell only
modestly. The performance for the group of the three
 J.E. Tilton / Resources Policy 27 (2001) 107–117
contracting mines falls, as expected, between these
two groups.
To assess more rigorously the relationship between
productivity and cash costs, we estimated Eqs. (4) and
(5) below. Eq. (4) presumes that the relationship between
cash costs and productivity in 1975 is non-linear. This
is because cash costs are likely to decline at a decreasing
rate as productivity goes up, since cash costs cannot be
negative and so at some point must asymptotically
approach zero or some positive value. Similarly, as pro-
ductivity goes down, cash costs can increase without
limit, but productivity must at some point approach zero
or some positive value.
The relationship between the change in cash costs and
change in productivity over the 1975–90 period could
be linear or non-linear. As a result, Eq. (5) was estimated
with both a linear and quadratic specification. The results
for the quadratic specification are reported below since
its better fit suggests a non-linear relationship.
Both equations are again estimated by ordinary least
squares. For Eq. (4), we anticipate a negative coefficient
on the productivity variable and a positive coefficient on
the square of this variable for the reasons noted above.
For Eq. (5), we expect a negative coefficient for the
change in productivity and a positive coefficient for the
square of this variable, assuming the relationship is not
linear. If this were not the case, smaller and smaller
changes in productivity would be associated with larger
and larger reductions in costs.
CC75i⫽ 214 −2.01 P75i +0.0082 SqP75i +ei R2=0.18
(4)
(4.65) (−0.75) (0.23)
PerCCi⫽ 5.58 −0.56 PerPi +0.0012 SqPerPi +ei R2=0.33
(5)
(0.67) (−3.34) (2.60)
Where:
CC75i cash costs in 1975 of the ith mine,
PerCCi percent change in the cash costs of the ith
mine over 1975–90,
P75i labor productivity of the ith mine in 1975,
SqP75i square of the productivity of the ith mine in
1975,
PerPi percent change in labor productivity of the ith
mine over 1975–90,
SqPerPi square of the change in productivity of the
ith mine over 1975–90.
The results provide some support for our expectations
that cash costs tend to fall at a decreasing rate as labor
productivity increases, and that changes over time in
cash costs are similarly related to changes in labor pro-
ductivity. All of the estimated coefficients have their
expected signs. However, only the coefficents for Eq. (5)
113
are statistically significant at the 95% probability level.7
Moreover, in contrast to our expectations, the relation-
ship between productivity and cash costs is weaker than
the relationship discussed earlier between productivity
and mine output growth, even though the latter is an
indirect relationship presumably working through vari-
able costs.
Since labor productivity is measured in terms of out-
put per thousand manhours, where output includes both
the copper contained in the concentrate and the copper
equivalent of associated byproducts, we expect the
relationship between labor productivity and costs to be
stronger for adjusted breakeven costs than for cash costs.
To test for this, we estimated Eqs. (6) and (7), whose
specification are exactly the same as Eqs. (4) and (5)
except that the dependent variables are respectively 1975
adjusted breakeven costs (BC75i) and changes in
adjusted breakeven costs over the 1975–90 period
(PerBCi). The results suggest that the use of adjusted
breakeven costs in place of cash costs does not make a
big difference. While Eq. (6) does somewhat better than
Eq. (4), Eq. (5) does somewhat better than Eq. (7).
BC75i⫽ 214 −4.43 P75i +0.043 SqP75i +ei
(5.02) (−1.81) (1.30) (6)
R =0.23
2
PerBCi⫽ −3.29 −0.23 PerPi +0.00031 SqPerPi +ei
(−0.45) (−1.54) (0.79) (7)
R =0.13
2
Costs and mine output
A close relationship should also exist between vari-
able costs and changes in mine output during a market
recession. Indeed, the relationship between mine growth
and variable costs should be stronger than the relation-
ship between mine growth and productivity, in part
because productivity presumably affects mine growth
indirectly via costs, and in part because, as we have just
seen, productivity is not perfectly correlated with costs.
We also expect to find a stronger relationship between
costs and mine growth when costs are measured by
7
Eq. (4) was also estimated using a double log, rather than a quad-
ratic, functional relationship. The results are quite consistent with those
shown above. The adjusted coefficient of determination is 0.18 for both
equations. However, the coefficient for the independent variable (the
log of productivity in 1975) was significant at the 95% probability
level. It indicates that a one percent increase in labor productivity was
associated with a 0.30% decline in cash costs in 1975. Since the double
log functional form cannot be used to estimate other equations reported
here, as the latter contain variables with negative or zero values, the
results for the quadratic form are shown above for consistency.
114 J.E. Tilton / Resources Policy 27 (2001) 107–117

adjusted breakeven costs rather than cash costs, as the
former take account of coproduct and byproduct rev-
enues and any discrepancies a mine may receive from
the reported copper price for its output.
Table 2 provides some support for these expectations.
The group of ten expanding mines enjoyed lower cash
costs in 1975 and a greater subsequent decline in both
cash and adjusted breakeven costs than the other two
groups of mines. Surprisingly, the adjusted breakeven
costs in 1975 for the non-surviving mines at 116 cents
per pound were as low as those for the expanding mines,
and considerably lower than those for the contracting
mines. Perhaps of greater interest, Table 2 provides little
support for the hypothesis that mine growth is more
closely tied to adjusted breakeven costs than cash costs.
Nor does it demonstrate that mine growth is more
closely related to costs than to labor productivity.
In an attempt to assess more fully the relationship
between mine growth and costs, we estimated the equa-
tions shown below. These equations are comparable to
Eqs. (1)–(3), but use cash and adjusted breakeven costs,
rather than labor productivity, to explain differences in
output growth among mines over the 1975–90 period.
PerQi⫽ 59.5 −0.34 CC75i +ei R2⫽⫺0.04
(0.42) (−0.37)
PerQi⫽ −29.2 +0.32 BC75i +ei R2⫽⫺0.04
(−0.25) (0.33)
PerQi⫽ −6.31 −1.57 PerCCi +ei R2⫽0.16
(−0.22) (−2.31)
PerQi⫽ −12.5 −1.69 PerBCi +ei R2⫽0.10
(−0.39) (−1.87)
PerQi⫽ 54.6 −0.40 CC75i −2.29 PerCCi +ei
(0.42) (−0.69) (−2.29)
R ⫽0.13R
2 2
In a number of respects, these results fail to support
our expectations. First, the estimated relationship
between mine growth and costs in 1975 is tenuous at
best. Whether or not we control for changes in costs after
1975, whether we use cash costs or adjusted breakeven
costs, the coefficients for these variables are not statisti-
cally significant, and in one instance actually have the
wrong sign.
Reductions in costs between 1975 and 1990, on the
other hand, do affect mine output growth and survival.
The coefficients for this variable—whether measured in
terms of adjusted breakeven costs or cash costs, and
whether considered alone or with 1975 costs—consist-
ently possess the expected sign and are statistically sig-
nificant at the customary 95% probability level using a
one-tailed test. Overall, however, changes in costs
explain only a very modest amount of the differences in
mine growth—no more than 16%, according to the
adjusted coefficients of determination reported above.
Second, there is no tendency for changes in mine out-
put to be more closely associated with adjusted break-
even costs than cash costs. The opposite appears to be
the case.
Third, we expected mine growth to be more closely
associated with costs than with labor productivity, but
(8) the above equations suggest just the opposite. Pro-
ductivity—both its level in 1975 and changes over the
next 15 yr—explains some 52% of the differences in
(9) mine growth, according to Eq. (3). This is a considerably
higher percentage than accounted for by costs in any of
the above equations.
(10) These results and findings, it is important to stress
again, should be considered as preliminary and treated
with caution for reasons noted earlier. Nevertheless, they
(11) are of interest, as much for the questions they raise as
for the answers they provide.
(12)
Reserves and mine size

PerQi⫽ 56.9 −0.63 BC75i −2.00 PerBCi +ei The surprising finding that labor productivity and
(0.48) (−0.61) (−1.90) (13) changes in labor productivity over the period are better
able to explain mine survival and growth during a
R ⫽0.07
2
recession than variable costs at the beginning of the per-
iod, changes in variable costs over the period, or the two
Where as in earlier equations: together raises an intriguing question: Why were some
mines so successful in raising productivity once into the
PerQi percent change in output for the ith mine over recession, while others were not? Or, to pose the ques-
the 1975–90 period, tion in a different way: Does a mine’s willingness and
CC75i cash costs in 1975 of the ith mine, ability to increase productivity reflect more fundamental
PerCCi percent change in the cash costs of the ith factors that determine whether or not it survives during
mine over 1975–90, a recession?
BC75i adjusted breakeven costs in 1975 of the ith This section examines two such possibilities. The first
mine, and is mine size. The larger a mine is, the more infrastructure
PerBCi percent change in adjusted breakeven costs of and physical capital is in place, which often must be
the ith mine over 1975–90. written off if the mine closes. This may enhance the
 J.E. Tilton / Resources Policy 27 (2001) 107–117
incentives owners have to invest in efforts to raise labor
productivity in order to avoid shutting down. In addition,
the larger a mine, the more employees it is likely to have,
and thus the more human capital available to find ways
to increase productivity.
The second is the size of reserves. Reserves reflect the
amount of metal contained in the mine’s deposits that is
known to exist and that is economic to extract at current
costs and prices. Presumably it is more worthwhile to
invest the capital and effort needed to raise productivity
if a mine’s reserves are likely to last for many years as
opposed to only a few years.8
Tables 1 and 2 provide considerable support for the
first of these two propositions. All of the non-surviving
mines produced less than 45,000 t of copper equivalent
in 1975. Only Bagdad among the surviving mines had
an output below this level. On average, the non-surviv-
ing mines produced less than a third of the output of the
other two groups in 1975. Mines size, however, is far
less helpful in explaining the differences in productivity
growth and output growth between the expanding and
contracting mines, as the average size for the contracting
mines was only slightly less than that for the
expanding mines.
Table 2 also highlights the importance of reserves. On
average, the expanding mines in 1975 had 47 yr of
reserves, compared to only 10 yr for the contracting
mines and 9 yr for the non-surviving mines.
To assess more rigorously the extent to which mine
size and reserve life account for differences in pro-
ductivity growth among mines, we estimated Eq. (14),
again using ordinary least squares. For the reasons noted,
we anticipate a positive coefficient for both the mine size
and the reserve life variables.
PerPi⫽ −102 0.27 Q75i +5.07 YrsRsi ei R2=0.59
(14)
(3.05) (−0.67) (5.84)
Where
PerPi percent change in productivity over the 1975–
90 period at the ith mine
Q75i copper equivalent output in thousands of tons
for the ith mine in 1975, and
YrsRsi years of reserves remaining for the ith mine
in 1975 at its 1975 output level
Eq. (14) supports the proposition that reserve life is an
important explanatory variable for the differences among
8
It is important to note, however, that there is a cost associated with
proving reserves. As a result, once reserves reach a certain multiple of
current production—25–30, for example–firms may not find it worth-
while to develop additional reserves. For this reason, a mine’s reported
reserves at any particular time may underestimate the amount of metal
expected to be extracted over the rest of its life.
115
mines in productivity growth. The coefficient for this
variable has the correct sign and is highly significant.
The equation as a whole accounts for nearly 60% of the
differences in productivity growth among mines. Sur-
prisingly, however, mine size does not appear important.
The coefficient for this variable is not significant and
has the wrong sign. While again, these results should be
treated with caution for all the reasons noted earlier, the
association between mine size in 1975 and survival sug-
gested by Table 1 apparently arises simply because the
average life of reserves is much greater for large mines
than for small mines.
Conclusions and implications
To our knowledge, this is the first attempt in the pro-
fessional literature to assess empirically the role of both
labor productivity and variable costs in mine survival
and growth over a major market recession. The findings,
based on the performance of 24 copper mines in the
United States during the 1975–90 period, both confirm
many of our expectations and provide some surprises.
Mines that entered the recession with relatively high
labor productivity and that managed to increase their
productivity during the recession were more likely to
survive than mines that did not. Indeed, many even
increased their output and market share over the course
of the recession and subsequent recovery.
Also, as expected, we found an inverse relationship
between labor productivity and variable costs. Though
the relationship was not particularly strong, mines with
high productivity in 1975 tended to enjoy lower variable
costs. Mines that greatly improved their productivity
over the 1975–90 period tended to reduce their variable
costs more than other firms.
Variable costs also appear to be linked to survival.
The lower a mine’s cash or adjusted breakeven costs in
1975 and the more it reduced these costs over the follow-
ing 15 yr, the greater its expected growth in output. In
a number of respects, however, the relationship found
between variable costs and mine survival and growth
is surprising.
First, it is weaker when adjusted breakeven costs are
used to reflect variable costs than when cash costs are
used.
Second, even cash costs explain only a small portion
of the differences in mine growth. This suggests that
other factors may be more important determinants of
growth and survival.
Third, the relationship between mine growth and vari-
able costs is less strong than the relationship between
mine growth and labor productivity, even though the lat-
ter presumably influences mine growth indirectly by
affecting costs.
The importance of labor productivity, and particularly
116 J.E. Tilton / Resources Policy 27 (2001) 107–117
of the growth in labor productivity over the period, for
mine survival leads one naturally to inquire about the
forces or determinants behind the growth in labor pro-
ductivity. From the available literature9 we know that
the US copper industry relied heavily on innovation to
raise its labor productivity and lower its variable costs
during the 1975–90 period. Many innovations entailed
the use of more advanced and efficient technology, such
as larger trucks, mobile in-pit crushers, real-time com-
puter-controlled processing, and solvent-extraction elec-
trowinning (SX-EW). Such innovations required heavy
capital expenditures. Others were operational changes
that required little new capital, such as more flexible
worker-manning schedules, changes in stripping ratios,
the more aggressive pursuit of price concessions from
suppliers, and the elimination of non-essential activities
(including company housing, company hospitals, and
other community services).
What is less clear, but of critical importance, is why
certain mines were successful at fostering innovation and
productivity growth, while others were not. One parti-
cularly important factor is the expected life of a mine’s
reserves. Mines with many years of remaining reserves
have greater incentive to invest in new equipment and
facilities embodying the latest technology than mines
with limited reserves.
While these findings are of interest, perhaps the most
important and significant finding from our analysis of
the U.S. copper industry is that a mine’s labor pro-
ductivity and costs at the beginning of a recession were
far less important for subsequent survival and recovery
than its ability to increase productivity and reduce costs
once in the recession. This finding has several interest-
ing implications.
First, in the realm of economic theory, it provides
empirical support for the recent developments in the
theory of the firm that challenge the assumption of con-
ventional microeconomics that firms are always able to
maximize profits. If the mines in our sample operated
on their efficiency frontier with known and given tech-
nology, their costs and productivity at the beginning of
the recession should have largely determined their ability
to survive.
Second, in the less academic world of mine finance,
this finding raises a note of caution. In particular it ques-
tions the widespread belief that a mine’s expected pos-
ition on the comparative cost curve reflects its ability to
weather a severe market downturn. The experience of
the US copper industry suggests that, once a recession
is underway, many mines will find ways to substantially
reduce costs. To remain among the lowest cost pro-
9
See, for example, Aydin (1998), Campbell (1989), National
Materials Advisory Board (1990), Takeuchi et al. (1987), Tilton and
Landsberg (1999) and US Congress (1988).
ducers, even low-cost mines must innovate. Much more
important than a mine’s starting position, it appears, is
a mine’s ability to react once in a recession.
Finally, while every mine is finite and must ultimately
run out of reserves, the experience of the US copper
industry suggests that managers and workers are not
always hapless bystanders with little or no influence over
a mine’s survival. Innovation and other activities that
reduce costs and raise labor productivity can often
extend the path to extinction by years if not decades.
Acknowledgements
I am grateful to Hamit Aydin for compiling the data
on labor productivity by mine used in this analysis from
US Mine Safety and Health Administration records, to
Michael L. Bailey and David Humphreys for thoughtful
comments on an earlier draft, and to Rio Tinto plc for
providing mine cost data from their proprietary Mine
Information System.
References
Aydin, H., 1998. Labor productivity growth in the US copper industry.
Unpublished PhD dissertation, Colorado School of Mines,
Golden, CO.
Campbell, G.A., 1989. The response of US copper companies to
changing market conditions. Resources Policy 15, 321–337.
Crowson, P.C.F., 1984. Are production costs of copper determined by
prices of copper? Paper presented to the CRU/CDA Conference on
the Marketing and Trading of Copper, October 8.
Crowson, P.C.F., 1998. Inside Mining: The Economics of the Supply
and Demand of Minerals and Metals. Mining Journal Books, Lon-
don.
Cyert, R.M., March, J.G., 1963. A Behavioral Theory of the Firm.
Prentice Hall, Englewood Cliffs, NJ.
Hall, R.L., Hitch, C.J., 1939. Price theory and business behavior.
Oxford Economic Papers 2, 12–45.
Hotelling, H., 1931. The economics of exhaustible resources. Journal
of Political Economy 39, 137–175.
Leibenstein, H., 1966. Allocative efficiency vs X-efficiency. American
Economic Review 56, 392–415.
Marris, R., 1964. The Economic Theory of ‘Managerial’ Capitalism.
Macmillan, London.
Nelson, R.R., Winter, S.G., 1982. An Evolutionary Theory of Econ-
omic Change. Harvard University Press, Cambridge, MA.
National Materials Advisory Board, National Research Council, 1990.
Competitiveness of the U.S. Minerals and Metals Industry.
National Academy Press, Washington DC.
Simon, H.A., 1955. A behavioral model of rational choice. Quarterly
Journal of Economics 69, 99–118.
Simon, H.A., 1959. Theories of decision making in economics. Amer-
ican Economic Review 49, 253–283.
Stiglitz, J.E., 1974. Incentive and risk sharing in sharecropping.
Review of Economic Studies 41, 219–255.
Takeuchi, K. et al., 1987. The World copper industry: its changing
structure and future prospects. Commodity Working Paper No. 15.
The World Bank, Washington DC.
Tilton, J.E., Landsberg, H.H., 1999. Innovation, productivity growth,
and the U.S. copper industry. In: Simpson, R.D. (Ed.), Productivity
 J.E. Tilton / Resources Policy 27 (2001) 107–117 117

in Natural Resources Industries: Improvement through Innovation.
Resources for the Future, Washington DC.
Torries, T.F., 1988. Competitive cost analysis in the mineral industries:
the case of nickel. Resources Policy 14, 193–204.
Torries, T.F., 1995. Comparative cost of nickel sulphides and laterites.
Resources Policy 21, 179–187.
US Census Bureau, 1992. Census of Mineral Industries
(www.census.gov/prod/1/manmin/92mmi/ 92minif.html).
US Bureau of Mines, 1987. An Appraisal of Minerals Availability for
34 Commodities, Bulletin 692. US Government Printing Office,
Washington DC.
US Congress, Office of Technology Assessment, 1988. Copper: Tech-
nology and Competitiveness, OTA-E367. Government Printing
Office, Washington DC.
Williamson, J., 1966. Profit, growth and sales maximization. Econ-
omica, NS 33, 1–16.