Professional Documents
Culture Documents
Copper Recycling and Scrap Availability: Fernando Go Mez, Juan Ignacio Guzma N, John E. Tilton
Copper Recycling and Scrap Availability: Fernando Go Mez, Juan Ignacio Guzma N, John E. Tilton
Abstract
According to existing estimates, available old copper scrap has more than tripled over the past 40 years. Secondary production (that is,
copper produced from recycling old scrap), however, has only doubled. Indeed, over the past 10 years, while copper consumption and
primary production have continued to expand briskly, while available old scrap has increased by over 35%, secondary production has
actually stagnated.
For a world concerned with sustainable development and the quality of the earth’s environment, this performance is disappointing and
in need of explanation. Other things being equal, one would expect the amount of recycling to increase with the availability of scrap, as
many econometric models of the world copper market developed over the past several decades explicitly assume.
The key to understanding sluggish growth in secondary production, this paper argues, is distinguishing carefully between (1) the flow of old
scrap that arrives each year from products reaching the end of their useful lives during the year and (2) the available stock of old scrap that was
not recycled during earlier years presumably because it was too costly to do so. Using an econometric model, the paper shows that old scrap
stocks, which have contributed most of the increase in available old copper scrap over the years, have a very modest impact on secondary
production. Old scrap flows have a much greater effect, but they account for only about 4% of the available old scrap for any given year.
r 2007 Elsevier Ltd. All rights reserved.
Like other metals, copper is produced from ores and For the most part new scrap is inexpensive to recycle,
scrap. It is customary to refer to the output from ores as since it is easy to identify, sort, concentrate, and transport.
primary production and the output from recycling scrap as As a result, almost all new scrap is recycled. Moreover, new
secondary production. Scrap can be either old scrap, which scrap is not usually considered a net addition to supply.
is the metal available for reuse in products that have come This is because whether 5% or 25% of the metal used to
to the end of their useful lives, or new scrap, which is the produce a given finished good ends up as new scrap and is
scrap generated in the production process as goods recycled does not affect the amount of the metal ultimately
containing copper are made.1 embodied in the product. As a result, the availability and
recycling of new scrap is not considered further, and
!Corresponding author. Division of Economics and Business, Colorado references below to secondary production refer only to the
School of Mines, Golden, CO 80401, USA. Tel.: +1 303 273 3485; recycling of old scrap.
fax: +1 303 273 3416. Over the past several decades various economic studies
E-mail addresses: Fernando.J.Gomez@bhpbilliton.com (F. Gómez), have contributed to our understanding of copper and other
jguzman@puc.cl (J.I. Guzmán), jtilton@mines.edu (J.E. Tilton).
1
For example, an aluminum beverage can once its contents are
metal markets. Among these works are a number of
consumed is old scrap, while the skeleton left after the tops for aluminum structural econometric models of the world copper
cans have been punched out of a flat sheet is new scrap. industry, including the works of Fischer et al. (1972),
0301-4207/$ - see front matter r 2007 Elsevier Ltd. All rights reserved.
doi:10.1016/j.resourpol.2007.08.002
ARTICLE IN PRESS
184 F. Gómez et al. / Resources Policy 32 (2007) 183–190
16000 300000
Primary
Production
14000
Secondary
Production 250000
12000 Available old
scrap
10000 200000
kton
kton
8000
6000 150000
4000
100000
2000
0 50000
1966
1968
1970
1972
1974
1976
1978
1980
1982
1984
1986
1988
1990
1992
1994
1996
1998
2000
2002
2004
Fig. 1. World primary and secondary copper production (left axis) and available old scrap (right axis), 1966–2005.
Slade (1980), Vial (1988, 2004), and Valencia (2005). A old scrap increases largely because of old scrap stocks, as
common element in the studies cited is the assumption that has been the case (see Table 1), this will have a very modest
an increase in available old copper scrap, all else being effect on secondary production, and over time the share of
equal, increases the supply of secondary copper production available old scrap recycled will fall.
from old scrap. Section 2, which follows this introduction, develops a
As Fig. 1 shows, however, old copper scrap has roughly model to test this hypothesis. Section 3 describes the data
tripled over the past four decades while secondary used to estimate the model, and Section 4 discusses the
production from old scrap has increased much more results. Finally, Section 5 highlights the conclusions and
modestly. In fact, for a number of years, secondary examines some implications.
production has actually declined. As a result, secondary
production as a percentage of copper consumption has
Modeling secondary copper production
fallen, from 18 to 13 over the 1966–2005 period (see
Table 1).
A common specification for modeling secondary copper
Presumably part of the explanation for the falling
production—used for example by Fischer et al. (1972), Vial
proportion of available old scrap that is recycled is the
(1988, 2004), and Valencia (2005)—assumes that secondary
secular decline in the real price of copper over the
copper production (QSt) depends on the real price of
past several decades (Tilton, 2006). As for the failure of
refined copper (Pt), the available old scrap (At), the lagged
secondary production to grow as rapidly as primary copper
value of the secondary copper production (QSt!1), and a
production, this may in part reflect the fact that primary
disturbance term (et) as shown in Eq. (1)
producers have been more successful over time reducing
production costs, as Tilton (1999) suggests. QSt ¼ f ðPt ; At ; QSt!1 Þ þ !t . (1)
This study proposes another explanation as to why
The price of refined copper (Pt) is included for the usual
secondary copper production has failed to match the
reasons. Since the supply of secondary copper produced
growth in available old scrap, an explanation that to our
from old scrap should rise with price, the coefficient on this
knowledge is not found in the existing literature. During
variable is expected to be positive.
any given year, available old scrap depends on old scrap
The lagged dependent variable (QSt!1) can be justified
stocks and old scrap flows. Old scrap stocks encompass the
on the grounds that the adjustment of supply to changes in
scrap available but not recycled during the preceding year,
price and available old scrap may not be completed within
which as a result remains available for recycling. Old scrap
a year, and so a partial adjustment model is necessary. In
flows encompass the scrap that becomes available during
this case, the estimated coefficient for this variable should
the year as the products in which it is embedded reach the
be between 0 and 1.2
end of their useful lives.
However, some authors, including Fischer et al. (1972)
Our hypothesis is that old scrap stocks for the most part
and Valencia (2005), believe the response of secondary
are more costly to recycle than old scrap flows, as
producers, unlike that of primary producers, is sufficiently
presumably they were uneconomic to recycle in the past.
An increase in old scrap stocks, as a result, will have a very 2
For more on the partial adjustment model and the reasons why the
modest impact on secondary copper production. The coefficient for the lagged dependent variable is expected to be between 0
impact of old scrap flows will be much greater. If available and 1, see Pindyck and Rubinfeld (1998, chap. 9).
ARTICLE IN PRESS
F. Gómez et al. / Resources Policy 32 (2007) 183–190 185
Table 1
Primary production, secondary production, consumption, LME price, old scrap flows, old scrap stocks, and available old scrap for the world copper
industry, 1966–2005
Year Primary Secondary Consumption LME price Pt Old scrap Old scrap Available old
production production QSt (kton) (US$ 2005/ flows AFt stocks ASt scrap At
(kton) (kton) ton) (kton) (kton) (kton)
Sources: International Copper Study Group (2006) for primary and secondary production, as well as for consumption; Chilean Copper Commission
(2006) for LME price; and International Monetary Fund (2006) for US producer price index. Old scrap flows, old scrap stocks, and available old scrap are
estimated as described in the text.
rapid that a partial adjustment model is inappropriate is economic to produce at any given price. The disturbance
particularly when using annual data to estimate the model. term (et) reflects the net influence of all other variables.
They justify the inclusion of a lagged dependent variable in Table 2 shows estimates from various studies for the
their models on the grounds that this variable serves as a elasticities of secondary copper supply with respect to price
proxy for the difficulty of collecting old scrap. The idea and available old scrap. Interestingly, in all those studies
here is that the more old scrap recycled last year, the less where both short-run and long-run elasticities are esti-
readily available old scrap is this year. As a result, they mated, the former are greater than the latter. This is not
anticipate a negative sign for the coefficient of this variable. what one normally expects. In the context of a partial
Available old scrap (At) is included as a proxy for the adjustment model, these results imply that the market in
costs of secondary production (Slade, 1980), which by their the short run over-adjusts to a change in price or available
nature are difficult to estimate directly. An increase in this old scrap, in the sense that the short-run increase in supply
variable, it is assumed, means that more secondary copper motivated by a change in price or available old scrap is
ARTICLE IN PRESS
186 F. Gómez et al. / Resources Policy 32 (2007) 183–190
Table 2
Estimated elasticities of world secondary copper production for the short and long runs with respect to available old scrap (eA) and price (eP) from selected
models
Fischer et al. (1972)a Slade (1980) Vial (1988) Intarapravich (1989)d Vial (2004) Valencia (2005)d
!ASR
0.69 0.40 0.79–1.25b – – 0.80
!ALR
0.50 – – – – 0.78
!PSR 0.42 0.28 0.33c 0.075 0.20 0.32
!PLR 0.31 – 0.26c – 0.11 0.31
– not available.
Sources: Studies cited in table.
a
Fischer et al. (1972) estimate three supply equations for secondary production: one for secondary production in the United States from new scrap, one
for secondary production in the United States from old scrap, and one for secondary production in the rest of the world from both new and old scrap.
Since elasticities of secondary production from new scrap and from old and new scrap combined are not of interest here, the elasticities shown for Fischer
et al. (1972) are only for secondary production from old scrap in the United States.
b
The range of figures shown covers the estimated elasticities for the particular areas considered, namely the United States, Japan, Germany, and the rest
of Europe.
c
The elasticities of secondary production for the world as a whole with respect to price for Vial (1988) are reported in Vial (2004).
d
Intarapravich (1989) and Valencia (2005) estimate linear equations, which implies that the elasticities vary as secondary supply and the other variables
vary. The figures shown in the table are the estimated elasticities at the average values of the variables over the 1966–2004 period for Valencia and over the
1950–1986 period for Intarapravich.
Table 4 Table 5
Empirical results using OLS and ridge regressions Short-run and long-run elasticities of secondary copper supply with
respect to price, old scrap flows, and old scrap stocks estimated by OLS
OLS RR (rHKB ¼ 0.029) RR (rMSE ¼ 0.051) and ridge regressions
5 6
For an explanation of how this is done, see Pindyck and Rubinfeld For more on ridge regressions, see Green (1998), Guilkey and Murphy
(1998). (1975), Marquardt and Snee (1975), and Vinod (1978). Principal
ARTICLE IN PRESS
F. Gómez et al. / Resources Policy 32 (2007) 183–190 189
following estimators: flows, and old scrap stocks are only 1.35 times greater than
their short-run values rather than nearly three times
b^ ¼ ½X 0 X þ rD'!1 X 0 Y , (5) greater, as shown in Table 5.
Of particular importance for our purposes, the results of
where D is a diagonal matrix whose diagonal elements are the ridge regressions continue to show a much greater
coincident with those of the matrix X0 X and r is a scalar response of secondary production to a change in old scrap
chosen arbitrarily. Compared with those for OLS, the flows than old scrap stocks, a difference that is now
resulting parameter estimates are biased but have smaller statistically significant at the 1% level. And, as noted
variances. above, the very large increase in old scrap stocks continues
Though there are many different ways to select the value to have only a very modest effect on secondary production.
of r (Judge et al., 1985), two are most often used. The first
is the Hoerl–Kennard–Baldwin rHKB, which is calculated Conclusions and implications
as shown below
The secondary industry accounts for only a small share
ks^ 2 (less than 20%) of total copper production. In addition, the
rHKB ¼ 0 , (6)
b^ b^ industry contains many small and secretive firms, making
comprehensive and reliable data difficult to obtain. For
where k is the number of parameters in the model (five in
these and other reasons, economists and others have
our model), and b^ and s^ are the estimates for these
devoted relatively little time to studying secondary
parameters derived using OLS. The resulting value of rHKB
producers. While this is now changing due in part to the
for our study is 0.029. The estimated coefficients using
growing concerns about the environment and sustainabil-
ridge regression with this value for r are shown in the
ity, there is still much that we do not understand about
second column of Table 4.
their behavior.
The second common method for determining r is to
One enigma—the focus of this analysis—is why over the
select the value that minimizes the mean square error of the
past four decades as the available old scrap has increased
estimation (Smith and Campbell, 1980), and thus obtain
substantially, the recycling of old scrap has not kept pace.
the most efficient set of estimators. The result is an rMSE
Other things being equal, we would expect this to be the
equal to 0.051. The results obtained when Eq. (4) was then
case, as is commonly assumed in econometric models of the
re-estimated using ridge regression with this value for r are
world copper industry. Of course, other things do change,
shown in the third column of Table 4.
and certainly part of the explanation for the failure of
Comparing the estimates derived from the ridge regres-
secondary production to grow as rapidly as expected can be
sions with those from OLS, we find that the estimates for
attributed to the decline in the real price of copper since the
b1, b2, b3, and b4 all continue to have values between 0 and
early 1970s, a decline driven largely by the highly successful
1. The estimated value for b1, the short-run elasticity of
efforts of primary copper producers to reduce their
secondary supply with respect to price, roughly doubles
production costs over this period (Tilton, 2006).
from its OLS value of 0.107, and increases in significance.
This study suggests another explanation has also played
The estimated value for b3, the short-run elasticity of
an important role. Available old copper scrap during any
secondary supply with respect to old scrap flows, declines
particular year consists of the old scrap stocks inherited
slightly from 0.243 to a low of 0.218, and becomes
from the preceding year and the old scrap flows returning
increasingly significant. The estimated value for b2, the
to the market during the year. Since old scrap stocks were
short-run elasticity of secondary supply with respect to old
not recycled in prior years, their recycling costs are
scrap stocks, rises from 0.003 to 0.016, a considerable
presumably relatively high, and hence one would expect
increase. However, even this higher value suggests that a
any given growth in old scrap stocks to increase secondary
doubling of old scrap stocks produces only a 1.6% increase
production less than the same growth in old scrap flows.
in secondary production.
Empirical analyses of secondary production over the
Finally, the estimate for b4 declines from 0.665 to a low
1966–2005 period strongly support this hypothesis. They
of 0.253, and becomes increasingly significant. The
also show that the influence of old scrap stocks, though
economic implication is that the response of secondary
positive and statistically significant, is very modest. Indeed,
production to a change in price, old scrap flows, or old
a doubling of old scrap stocks even in the long run
scrap stocks is much faster, with some 75% of any
apparently produces less than a 2% increase in secondary
difference between the current level of secondary produc-
production. A doubling of old scrap flows, however, may
tion and its long-run equilibrium level being realized within
in the long run produce a 73% increase. While the influence
a year. This in turn means that the long-run elasticities of
of both old scrap stocks and old scrap flows on copper
secondary supply with respect to changes in price, old scrap
recycling are positive and significant statistically, only old
(footnote continued) scrap flows are really important.
components is an alternative technique for dealing with multicollinearity These findings have several implications. First, they
(Green, 1998). provide some support for the view advanced by Gordon
ARTICLE IN PRESS
190 F. Gómez et al. / Resources Policy 32 (2007) 183–190
et al. (2006) as well as others that copper scrap if it is not Green, W.H., 1998. Economic Analysis, third ed. Prentice-Hall, Upper
recycled soon after returning to the market will rarely be Saddle River, NJ.
Guilkey, D., Murphy, J., 1975. Directed regression techniques in cases of
reused. While dramatic new technology may in the future
multicollinearity. J. Am. Stat. Assoc. 70 (352), 769–775.
make it profitable to reprocess the copper scrap in Intarapravich, D., 1989. A dynamic econometric analysis integrating long-
municipal and other waste sites, over the past 40 years and short-run adjustments in the mineral industry. Ph.D. Dissertation,
the amount of new copper produced from old scrap stocks West Virginia University, Morgantown, WV.
has been very modest. International Copper Study Group, 2006. World refined copper produc-
Second, the findings suggest that the future of the tion and usage trends. /http://www.icsg.org/markets/default.htmS
(accessed 26.07.2006).
secondary copper industry is much more closely tied to old International Monetary Fund, 2006. International Financial Statistics.
scrap flows. Over time these flows should follow with a lag /http://ifs.apdi.net/imf/logon.aspxS (accessed 26.07.2006).
the growth in total copper consumption and the allocation Jolly, J., Maryland, D., 2003. The US Copper-Base Scrap Industry and
of consumption among its principal end uses. its By-Products, fourth ed. Copper Development Association Inc.,
Finally, econometric models and other economic ana- New York, NY.
Judge, G., Griffiths, W., Hill, R., Lutkepohl, H., Lee, T., 1985.
lyses of the copper industry need to consider separately the The Theory and Practice of Econometrics, second ed. Wiley,
effects of old scrap stocks and old scrap flows on secondary New York.
production. When the two are lumped together as available Marquardt, D., Snee, R., 1975. Ridge regression in practice. Am. Stat. 29
old scrap the estimated coefficient is a weighted average (1), 3–20.
Pindyck, R., Rubinfeld, D., 1998. Econometric Models and Economic
hybrid of their individual influences, a hybrid that over-
Forecasts, 4th ed. Irwin McGraw-Hill, Boston, MA.
estimates the influence of an increase in old scrap stocks Slade, M., 1980. An econometric model of the US secondary copper
and underestimates the influence of an increase in old scrap industry: recycling versus disposal. J. Environ. Econ. Manage. 7 (2),
flows. 123–141.
Smith, G., Campbell, R., 1980. A critique of some ridge regressions.
J. Am. Stat. Assoc. 75 (369), 74–81.
Acknowledgments Tilton, J.E., 1999. The future of recycling. Resour. Policy 25 (3), 197–204.
Tilton, J.E., 2006. Understanding Cyclical and Secular Trends in Metal
The authors are grateful to Roderick G. Eggert and Prices. Mine Management Handbook. Australasian Institute of
Claudio Valencia for their helpful comments. Mining and Metallurgy, Carlton, Vic.
Valencia, C., 2005. An econometric study of the world copper industry.
Ph.D. Dissertation, Colorado School of Mines, Golden, CO.
References Vial, J., 1988. An econometric study of the world copper market. Ph.D.
Dissertation, University of Pennsylvania, Philadelphia, PA.
Chilean Copper Commission, 2006. Metal Prices. /www.cochilco.clS Vial, J., 2004. Modeling commodity markets in the global economy:
(accessed 26.07.2006). familiar findings and new strategies. CGSD Working Paper 18, Earth
Fischer, F.M., Cootner, P.H., Baily, M.N., 1972. An econometric model Institute, Columbia University, New York, NY.
of the world copper industry. Bell J. Econ. Manage. Sci. 3 (2), 568–609. Vinod, H., 1978. A survey of ridge regression and related techniques for
Gordon, R.B., Bertram, M., Graedel, T.E., 2006. Metal stocks and improvements over ordinary least squares. Rev. Econ. Stat. 60 (1),
sustainability. Proc. Natl. Acad. USA 103 (5), 1209–1214. 121–131.