An Econometric Model of The U.S. Secondary Copper Industry: Recycling Versus Dis - Posal '

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 19

JOURNAL OF ENVIRONMENTAL ECONOMICS AND MANAGEMENT 7, 123-141 (1980)

An Econometric Model of the U.S. Secondary Copper Industry:


Recycling versus Dis.posal’

MARGARET E. SL’ADE
Ofice of Resource Analysis, The U.S. Geological Survey, Reston, Virginia %‘092 and
Department of Economics, The George Washington University, Washington, D.C. 80062

Received September 19, 1979; revised December 19, 1979

In this paper, a theoretical model of secondary recovery is developed that integrates


microeconomic theories of production and cost with a dynamic model of scrap generation
and accumulation. The model equations are estimated for the U.S. secondary copper
industry and used to assess the impacts that various policies and future events have on
copper recycling rates. The alternatives considered are: subsidies for secondary produc-
tion, differing energy costs, and varying ore quality in primary production.

I. INTRODUCTION
A natural resource commodity is classified as nonrenewable if the process that
generates it is long, and renewable if the process is short. For example, although
most harvested crops can be replenished within a year, the geologic processes
that produce minerals take so long that we can, for all practical purposes, regard
the stock of minerals in the Earth’s crust as fixed. In addition, a commodity is
classified as nonrecyclable if the commodity’s valuable properties are dissipated
after use, and recyclable if they remain intact. For example, fuels are valued for
their energy content, which is dissipated in combustion. In contrast, such desirable
properties of metals as strength, conductivity, or noncorrosiveness persist in
discarded metal products. Because costs are associated with the disposal of
wornout products (that is, residuals may be harmful or perhaps just unsightly),
it is important to know what determines recycling rates for any commodity.
There is additional interest in the determinants of recycling rates for nonrenewable
resources, because, for given consumption, increased recycling forestalls the
exhaustion of these resources.
When a product has reached the end of its useful lifetime, the materials in it
can be either reused or discarded. The decision to reclaim or not is an important
one, because most products that are not reclaimed are burned, buried, or allowed
to rust or decay. Eventually their constituent materials will be in a form for
which reclamation costs are prohibitive. This paper examines the factors that
influence the decision to recycle. The model developed here is implemented for
one natural resource commodity, copper. However, the model has broader
r The author is grateful to Anthony Yezer, Raymond Kopp, Emil Attanasi, and two anonymous
reviewers for helpful discussions and constructive comments.

123

0995-0696/80/020123-19$02BO/O
.Copyrlpt0 198!, by. Academic Press, Inc.
All n&s o npmduchon m any form mewed.
124 MARGARET E. SLADE

implications and could be used to st*udy the determinants of recycling rates for
a large number of materials (both nonrenewable and renewable).
The published economic literature on recycling can be grouped into three
classes : theoretical, empirical, and descript,ive. The theoretical class of studies
(Smith [23], Schulze [ZO], Hoe1 [lS], for example) examines optimal pat.hs for
recycling under alternative assumptions about, the environmental effects of
waste disposal, the incent,ive structure for recycling, and the structure of the
markets in which secondary producers operate. Interesting theoretical results
have been derived using optimal control theory, but predictions from these
models are mainly qualitative. Little econometric work has been done on recycling.
Fisher et al. [13] looked at recycling in the cont,ext of a broader econometric
model of the world copper industry, Anderson and Spiegelman [2] investigated
the effects of tax policy on secondary material use, and Edwards and Pearce [12]
examined price incentives for secondary recovery. The econometric models of
secondary supply in these studies do not rest, on firm microeconomic foundations
and are based on inadequat,e specifications of the dynamics of scrap accumulation.
Descriptive studies (Bower [3], Butlin [5], Grace [15], and Grace et al. [14],
for example) discuss many of the important issues involved in recycling, and
tabulate information, such as recycling rates in different countries, international
trade in scrap materials, and secondary price movements. As with the theoretical
studies, most descriptive-study conclusions are qualitative.
In this paper, a theoretical model of secondary recovery is developed that
integrates microeconomic theories of production and cost with a dynamic model of
scrap generation and accumulation. The model also provides an explanation for
the low price elasticities of secondary supply observed in practice as well as in
most, empirical st,udies, and shows why policies designed to increase recycling
through price manipulation alone are bound to fail. The model equations are
estimated for the U.S. secondary copper industry and used to assess the impacts
that various policies and future events have on copper recycling rates. The
alternatives considered are : subsidies for secondary production, differing energy
costs, and varying ore quality in primary production.

II. THE SECONDARY COPPER INDUSTRY IN THE U.S.


The secondary copper industry can be broadly or narrowly defined. Although
the term “secondary producer” usually refers to the secondary refiners and alloy
ingot makers (the firms t,hat process scrap), the term “secondary industry” may
refer only to the secondary producers or it may also include the scrap dealers
and collectors. In this paper, the secondary industry is broadly defined to encom-
pass the scrap dealers, and all secondary metal statistics pertain to scrap used
directly as well as to scrap that is treated by the secondary producers.
Unlike the primary copper industry, which is highly concentrated, the second-
ary industry ranges from extremely to moderately unconcentrated, depending
on its definition. There are thousands of scrap dealers, and scrap markets are
highly competitive. In contrast, there are 50 to 100 secondary producers, and, if
the industry is defined to include only these producers, concentration is just
slightly less than average market concentration for manufacturing industries
in the United States. However, a more meaningful definition of t,he market would
include the output of all producers, primary and secondary. Using the latter
COPPER RECYCLING VERSUS DISPOSAL 125

FIG. 1. The flow of metal in the copper industry.

definition, we can describe the copper industry as an oligopoly (the primary


producers) that has a competitive fringe (the secondary producers).
In the secondary industry, metal production is from scrap, not ores. The three
types of copper scrap are: new, home, and old. New scrap is industrial and is
generated at various stages of the product8ion process by primary producers,
fabricators, and the manufacturers of end products. Home (or run-around) scrap,
a subset of new scrap, is recycled by the same establishment that generated it.
Because home scrap never leaves the firm and is never marketed, it is generally
not included in new-scrap statistics. Old scrap is recovered from discarded and
wornout products after they have ceased to be useful. Scrap can be virtually
pure metal or it can be one of many types of metal alloys that have as little as
30% copper content. Refinary slags, dross, skimmings, and ash, though not
technically scrap, are also treated by the secondary industry.
Some scrap is used directly by primary refineries, mills, foundries, and chemical
plants. Other scrap goes through the scrap dealers to the secondary refiners and
alloy ingot makers. Figure 1 shows the complicated flow of metal from the mines
to the producers, fabricators, and ultimate consumers. Scrap is generated at each
stage and is returned, either directly or via the secondary producers, to each
stage of the production process. For a more complete description of the U.S.
secondary copper industry, the reader is referred to Carillo et al. [6].

III. THEORETICAL MODELS OF PRICING AND SUPPLY


111.1. Secondary Price
We can distinguish three basic copper prices : the U.S. producer price, exchange
prices, and scrap prices. The producer price, which is set by the major American
and Canadian primary producers, changes only at discrete intervals and is less
volatile than exchange or scrap prices. Exchange prices, which are related to
price quotations on the major metal exchanges-the New York Commodity
Exchange (COMEX) and the London Metal Exchange (LME)-fluctuate with
short-run changes in supply and demand and are very unstable. Scrap prices,
though often not determined directly by the exchanges, are highly correlated
with but always lower than exchange prices. Figure 2 shows the fluctuations of
the U.S. producer price, the LME and COMEX prices, and the price of number
two copper scrap between 1966 and 1976.
126 MARGARET E. SLADE

Because refined primary and secondary copper are virtually indistinguishable,


secondary copper price is determined principally by primary price. However,
because the producer price is relatively stable, secondary price must fluctuate to
absorb some of the excess demand created by sticky primary prices, and the
level of aggregate economic activity is also a factor in its determination. If we
postulate that changes in secondary price are caused by changes in primary price
and by fluctuat’ions in aggregate demand, then a model for secondary price
formation is
In Ps = a0 + al In Pp + a2 In Y + a3 In Ps-~ (1)
where Ps is secondary copper price, Pp is primary copper price, and Y is some
measure of aggregate demand. The lagged value of secondary price was included
in the equation to test the hypothesis that secondary price is damped (a3 > 0
implies that prices respond slowly to changed conditions) or overly responsive
(a3 < 0 implies that the immediate response is greater than the equilibrium
response).
It should be noted that Eq. (1) implies that the secondary industry as a whole,
not just the individual firm, is a price taker (the elasticity of price with respect
to secondary output is zero). This formulation requires some justification. In
1977, secondary copper supplied 43% of U.S. copper consumption. Such a large
proportion might indicate that the secondary producers have a great deal more
price power than is implied by the price formation equation. However, the
supply of metal from new (indust,rial) scrap, which is a byproduct of primary
production and fabrication, is proportional to the amount of metal consumed and
is not determined by price (see Section 111.2). Metal reclaimed from old (post
consumer) scrap, which does respond to price, supplied only 16% of U.S. copper
consumption in 1977. In addition, the demand for copper is fairly inelastic in the
short-run.” Therefore, it is not unreasonable to assume that t.he secondary
industry can ignore its effect on price, and t’hat Eq. (1) is also the short-run
demand curve for secondary metal produced from old scrap.

III.2. Secondary Output


New, or industrial scrap, is generated by the production process itself. New
scrap is fairly good quality, recovery efficiency for new scrap is high, and recovery
takes place soon after t’he scrap is generated. Because the determinants of the
supply of met,al from new scrap are technological, not economic, new scrap supply
will not be discussed here.3
Old scrap is material t.hat becomes available when products reach the end of
their useful lifetimes. Old scrap processing ranges from simply sorting high-grade
scrap, which can be reused directly without further treatment, to resmelting and
refining the poorest quality material. Therefore, the cost of reclaiming scrap
depends on the form in which the scrap metal is found, and increasing secondary
metal production at a given time means reclaiming scrap material of lower
quality. Marginal cost in secondary metal production thus increases with recovery
* When the secondary producers increase production, the primary producers might lower price
if a small reduction would maintain their market share. However, if a large reduction is required,
they will probably not react with price changes.
8An equation of the form Qns = jf.? (where Qns is the supply of copper from new scrap and C
is copper consumption) was estimated. This simple equation explains 9770 of the variation in
production of copper from new scrap.
COPPER RECYCLING VERSUS DISPOSAL 127

6 /lb

- Producer

-- LME
- COMEX
- - Scrap

1966 67 68 69 70 71 72 73 74 75 76
YEAR

FIG. 2. Copper prices, 19661976.

efficiency (defined as the fraction of the stock of scrap available for collection that
is actually reclaimed).* For example, Chapman [7] showed that whereas, on the
average, secondary ‘copper production uses one-eighth the energy per ton needed
to produce copper from ores, marginal energy cost in secondary copper production
increases very rapidly in the range of present U.S. recovery efficiency.
The secondary producer reclaims metal from old scrap by a process that uses
the services of five inputs: capital (K), labor (L), energy (E), materials (M),
and t,ransportation (T). Dual to the production function is the secondary
producer’s minimum unit cost function, c, which, in general, depends on t’he level
of output (Q) and input prices
c = c(Q, Pi), i = K, L, E, M, T.
(3)
If we specify a cost function that is homothetic, then duality theory assures
us that the dual production function exists (Shephard [21]). For the purposes
of estimation, we must choose a specific functional form for c. Because average
and marginal costs vary with recovery efficiency (RE = Q/t%',where SS are the
stocks of scrap available for collection), let c be defined by
c = ao(Q/L~%3>~l-J Pi”‘, i = K, L, E, M, T (4)
*
with
a, ao, >o, 0 < ai < 1, C ai = 1. (51
*
’ In actuality, there are many different types of scrap, each with different productivity. Making
marginal cost a function of recovery efficiency is equivalent to assuming a continuum of scrap
types with continuously decreasing productivity.
128 MARGARET E. SLADE

Conditions (5) ensure that the cost, function is homothet.ic.5 The total and
marginal cost cost functions, TC and MC, corresponding to c are
TC = cyOQafl,‘SSa n P? (6)
and

Because t,he secondary copper industry is competitive, the individual producer’s


supply curve is his short-run marginal cost curve and the industry supply curve
is the horizontal sum of these marginal cost curves. Industry output is determined
by the intersection of t,he price (demand) and marginal cost, (supply) schedules
Ps = MC = ao(a + 1) (Q/S’S)” n PF.

Taking natural logarithms of both sides of Eq. (8) we have

lnPs=cO+alnQ- a In SS + C LYEIn Pi where co = In [LY~(~ + l)], (9)

which can be solved for In Q

In Q = -co/a + l/a In PS + In SS - C ai/a In Pi. (10)

When we estimate Eq. (lo), a practical difficulty arises. Scrap metal is the
principal input to secondary metal production, but, its price is highly correlated
with the output price. For example, the price of secondary refined copper should
be approximately equal to the COMEX price (because primary and secondary
refined copper are virtually indistinguishable). Figure 2 shows that the price
of scrap moves with the COMEX price but is always below it. If Ps is a represen-
tative price of the output (processed scrap), and Pm is the price of the input
(unprocessed scrap), and if changes in Ps give rise to changes in Pm, then we have
In Pm = X In Ps. (11)
We can substitute h In Ps for the input price in Eq. (10) to obtain

In& = -co/a + l/ff In Ps + ln SS - C LYE/~In Pi - Xcu,/ff In PS,

i = K, L, E, M’, T,7 (12)


or
In Q = -co/~ + l/cy[l - XCY~]In Ps + In SS - l/a Tori In Pi. (13)

In Section V, Eq. (13) will be used to estimate the supply curve of metal
5 A discussion of the cost function and its dual production function can be found in Appendix A
8 For Eq. (7) to be the marginal cost function corresponding to 6, we must assume that &W/aQ
= 0, which is not totally realistic. However we can assume that, at any time t, SSt is fixed and
production from scrap only changes the stockes of scrap in the next time period SSg+l = SSt
+ ASSl - Q1 where ASSt are addftions to the stocks of scrap in period t. This is the same as
defining recovery efficiency as the percentage of the stocks of scrap available at the start of the
period that are reclaimed during the period.
7 M’ stands for all material inputs other than scrap.
COPPER RECYCLING VERSUS DISPOSAL 129

$/Unit I MCI
MCI

MCo

PI

p0

I!i I
Units
a; aoa; a1
FIG. 3. Adjustment of secondary supply to a price increase.

produced from old scrap.8 Note, however, that restrictions 5 do not appear in
the equation in any obvious fashion. In addition, the parameters of Eq. (13) are
not identified. Instead of imposing the restrictions on the parameters before
estimation, the restrictions will be used to identify the parameters.
In Eq. (13), the output-price elasticity of supply is l/a - k/or. This elasticity
can be interpreted as follows. The first half, l/a, is the partial output-price
elasticity of supply obtained by holding the prices of all inputs constant. If
OL= 0, supply is perfectly elastic ; as (Yincreases, the elasticity of supply is reduced
(that is, LYdetermines the slope of the marginal cost curve). The second half,
-XO~/(IL, measures the reduction in output due to an increase in the price of the
principal input, and is determined by (Y,, the cost share of scrap in producing
secondary metal, (Y, the output elasticity of marginal cost, and X, a measure of
the sensitivity of the scrap price to changes in the output price. Therefore, when
secondary metal price increases, there is an increase in output as the demand
curve shifts upward, which is partially offset by a fall in output as the supply
curve rises. A graphical interpretation of the adjustment of secondary metal
supply to a higher price is found in Fig. 3. When secondary price increases from
PO to PI, if all input prices are held constant, output increases from Qo to &I.
However, at the same time, the marginal cost curve shifts from MC0 to MC1
as scrap metal becomes more costly to purchase, and output falls from &I to &I’.
Even if the partial output-price elasticity, l/cw, is fairly high, the net elasticity,
l/a - has/~, can be low, accounting for the observed poor response of secondary
supply to price changes.g In fact, it is theoretically possible for the net elasticity
to be negative (if Xa, > 1) as illustrated by marginal cost curve MC2 and corre-
sponding output Q1”. A negative elasticity will result if the price of unprocessed
scrap is extremely sensitive to changes in the price of processed metal (i.e., if
X is very large). Equation (13) leads us to expect that policies designed to
increase recycling through price manipulation alone (such as price supports)
will be ineffectual.
8 When Eq. (13) is estimated, the coefficient of SS will not be constrained to equal one. If, in
addition to increases in marginal cost with higher recovery efficiency, short-run marginal cost
rises &S output expands (due to capacity constraints), the exponent of Q in Eq. (7) will be greater
than the exponent of SS, implying that the coefficient of SS in Eq. (13) is less than one.
e A low price elasticity of supply has been computed, not only for secondary copper (Fisher
et al. [13]) but also for other secondary materials-aluminum (Charlea River Associates [9])
and waste paper (Miedma el al. [lS]), for example. The explanation advanced above applies to
any secondary material.
130 MARGARET E. SLADE

IV. DYNAMIC MODEL OF SCRAP GENERATION


Equation (13) contains SS, the stocks of copper scrap, as an explanatory
variable. SS is not just the level of inventories of scrap metal kept by secondary
establishments, for which data exist; it also includes the accumulated stock of
discarded metal in junk yards, an unrecorded item. It is, t,herefore, necessary to
devise a method for calculating SS.
In contrast to new scrap, which is recycled almost immediately after it is
generated, the metal that becomes old scrap may be embodied in a product for
30 years or more before it can be reused. Previous models of scrap generation
assume either that the supply of scrap is infinitely elastic or that there is a
constant lag between consumption and the availability of the metal for recycling
(usually set at 1 year-a very unrealistic length of time for copper).iO In actuality,
the time required for metal to be scrapped depends on the type of product
produced. If we assume, for simplicity, that each product type has a constant
average lifetime and that the mix of products is invariant over time, then the
metal that becomes available for collection in any year (the sum of the metal
contained in all products whose average lifetime has just expired) is

ASS1 = c fiCt-ti i=l f . . . 9 n,

where
ASS are additions to the stocks of scrap,
fi is the fraction of metal consumption that goes into product, type i,
n is the number of product types,
ti is the average lifetime of product type i, and
c is metal consumption.

Net additions to the stocks of scrap in any year are equal to additions minus
production from old scrap in that year. l* Therefore, scrap stocks can be computed
from

SST = SST--~ + (1 - ~)(ASST--I - QT--1)

= sst, + 2 (1 - S)‘(ASST-t - &r--t)

lo Fisher et al. [133 estimated an equation of the form In (Q/SS) = a0 + al In Ps + a2 In


(&/5%9.-I, al > 0, a, < 0, where SS is computed as cummulative consumption (i.e., products
are immediately available for recycling). Charles River Associates [S] discussed a model similar
to that proposed below but estimated an equation of the form Q = ao + alps + at&-t, ai > 0,
at < 0. Anderson and Spiegelman [2] estimated an equation (for scrap steel) of the form Q = aa
+ alps + a2t + a,Y + a>s-i, al, a,, al > 0, a, < 0. Stocks of scrap do not enter the second
and third equations, and input prices are ignored in all three secondary supply models. Moroney
and Trapani [I9] specified translog cost functions for metal users and estimated elasticities of
substitution between primary and secondary metals for both copper and steel. However, they
ignored any limits on the availability of scrap. In addition, because the prices of primary and
secondary metals move together, some of their estimates imply that primary and secondary metals
are complements.
11Net exports also subtract from scrap stocks. Copper Development Association (CDA) show
that net exports of scrap were 6% of scrap consumption in 1976. However, CDA do not distinguish
between new and old scrap exports, and, therefore, their data cannot be used in calculating SS.
COPPER RECYCLING VERSUS DISPOSAL 131

TABLE I
Parameters Used in Calculating the Stocks of Old Copper Scrap

Fraction of Average lifetime


consumption of product
(%) (Pears)

Building and construction 24 30


Transportation 14 1p
Consumer durables 22 5
Electrical and electronic 23 30
Industrial machinery 16 20

to = 1942 &St, = 4000 X lo3 tons

a An average lifetime of 17 years was computed by assuming that three-fourths of metal con-
sumed in transportation goes into automobiles, with an average lifetime of 10 years, while the
rest goes into products with an average lifetime of 40 years.

where
SS are stocks of scrap,
6 is the rate at which scrap decays,
Q is metal produced from old scrap, and
to is some base year.
If the stock of scrap at time t = to, S, the fi, and the ti were known, Eqs. (14)
and (15) could be used to calculate the stock of scrap in any future year.
Table I shows the values of the parameters of Eqs. (14) and (15) used in
calculating the stock of copper scrap .I2 The fractions of metal consumption going
into each type of product, the fi, were chosen to equal their 1976 proportions.
Several people (Brown and Butler [4], Carillo et al. [6], and Crampton [ll])
have estimated product average lifetimes, the ti. The product lifetimes from the
various sources do not differ greatly, and an average is used here. The base year
was chosen to be 1942, and a very rough estimate of scrap stocks in that year was
made.la
Equations (1) and (13)-(15) provide the models of secondary metal price and
production that will be quantified in the next section. For empirical implementa-
tion, Eqs. (1) and (13) will be embedded in a stochastic framework (that is, we
assume that deviations of prices and output from the specified models are due to
random errors). In contrast, Eqs. (14) and (15) are identities.

V. ESTIMATION OF THE MODEL EQUATIONS


Data for copper prices (primary and secondary) came from the U.S. Bureau
of Mines “Minerals Yearbook,” and most input prices used are indices calculated
by the Bureau of Labor Statistics in computing the Wholesale Price Index. All
prices (input and output) were deflated by the Wholesale Price Index (1967 = 1),
I* Values of 6 ranging from 0 to 0.5 were tested. However, the choice of 6 made little difference
to the estimates of the supply curve parameters or to the t-statistic of the coefficient of SS in
Eq. (13). Therefore a value of 8 = 0 was used. Scrap stocks are thus somewhat overestimated.
ISThe estimate used is approximately two times total consumption of copper in that year.
Fortunately, the estimates of the coefficients of the supply equation are not sensitive to the
choice of initial stocks, a result to be expected because metal that has been scrapped for more than
a decade is not very apt to be reclaimed.
132 MARGARET E. SLADE

and are thus in 1967 constant dollars. Secondary production figures are recorded
by Copper Development Association in its “Annual Data.” The measure of
aggregate economic activity used is the Federal Reserve System’s Index of
Industrial Production. Appendix B cont,ains a more complete description of the
data and documents all data sources.
The final specification for the equation for copper secondary price, based on
(1) was
lnPsl = - 8.75 + 1.15 In Ppt + 1.74 In IP, + 0.24 D66t
(4.0) (2.6) (2.5)
- 0.30 lnPsl-l, (16)
(-2.1)
R2 = 0.79, F = 16, DW = 2.1, p = 0.94, Years : 1954-1976,
where
Ps is the deflated price of secondary copper,
PP is the deflated price of primary copper,
IP is the Index of Industrial Production, and
D66 is a dummy variable equal to one in 1966 and zero ot,herwise.
The estimated coefficient of primary price is greater than one, implying that
fluctuations in the producer price are magnified in the scrap price. The estimated
coefficient of the Index of Industrial Product,ion is even larger than that of
primary price, implying that the scrap price is very sensitive to fluctuations in
economic activity. (A 1% increase in industrial production causes a 1.74%
increase in Ps). A dummy variable equal to one in 1966 and to zero elsewhere
was included in the equation because, in 1966, uncertain supplies in Zambia led
to unusually high copper prices worldwide. As would be expected, the estimated
coefficient of the dummy variable is positive. Finally, the estimated coefficient
of the lagged value of Ps is negative, implying that secondary price is overly
responsive to changed conditions (that is, the immediate response is greater
than the equilibrium response). The signs and magnitudes of all the coefficients
in Eq. (16) reinforce the conclusion that secondary price is very volatile and that
it takes up much of the slack caused by more sluggish primary prices.
Equation (16) has both a lagged dependent variable and a serially correlated
error term. Therefore, when estimated by ordinary least squares (OLS), the
coefficient of Ps-~ will be biased if the sample is small. To avoid this bias, instru-
ments were used for Ps-1. The instruments were once and twice lagged values of
the independent variables in the equation (Pp, IP, and 066). In this and in all
other equations reported, the numbers in parentheses under the estimated
coefficients are the t-ratios of the corresponding coefficients, and p is the estimated
autocorrelation coefficient of the error term.
When the supply curve for secondary metal produced from old scrap (based
on 13) was estimated by OLS, the resulting equation was
In o1 = 5.98 + 0.32 In Pst - 0.14 In SS, - 0.03 In PKt
(5.6) (-0.70) (-0.11)
+ 0.53 In PLt - 0.27) PEt + 0.08 In PC1+ 0.05 In PTt (17)
(1.1) (-1.5) (0.40) (0.50)
R2 = 0.90, F = 18, DW = 1.9, Years : 1954-1976,
COPPER RECYCLING VERSUS DISPOSAL 133

where
& is secondary copper produced from old scrap,
Ps is the deflated price of secondary copper,
SS are stocks of old copper scrap,
PK is a deflated index of capital costs,
PL is a deflated index of labor costs,
PE is a deflated index of energy costs,
PC is a deflated index of indust,rial chemical costs,*( and
PT is a deflated index of transportation costs.
Because there is a high degree of multicollinearity among the input prices,
most of the coefficients of the cost variables in Eq. (17) are statistically insignif-
icant, and many have the wrong signs. We know, a priori, that the a< [the input-
price elasticities in Eq. (4)] must be positive and should probably be about 0.1.15
In addition, CY(the returns to scale parameter minus one) must be positive. A
reasonable guess for or is 0.5 (that is, if (Y = 0.5, total cost increases with & at a
rate &I.“). To obtain better estimates of the coefficients of Eq. (13), we can
incorporate this prior information into the estimation procedure.
In the notation of Johnston [17, pp. 221-2221, if /3 is the vector of coefficients
in Eq. (13), then a set of stochastic a priori restrictions on fl can be expressed as
r = Rp + v, E[v] = 0, E[v'v] = q, WI
and the restricted system can be estimated by generalized least squares. Here
we assume that
\k = u,2Ig, (19)
where g is the number of restrictions, and Ig is an identity matrix of order g.
In our particular case, R is an indentity matrix and the elements of r are t,he
expected values of the coefficients of Pi in Eq. (13). If we assume that these
coefficients lie with near certainty in the range 0 to 0.5,
0 < coef Pi < 0.5,

TABLE II
Estimated Elasticities of Marginal Cost in Secondary Copper Production
with Respect to the Input Prices

Input Elasticity

Scrap metal 0.50


Capital 0.10
Labor 0.14
Energy 0.21
Chemicals 0.05
Transportation 0.03

I4 Industrial chemicals are the principal nonmetal material inputs to secondary copper pro-
duction.
iK The ai are the cost shares of K, L, E, C, and T in producing secondary metal. Input-output
tables indicate that purchases of scrap and primary metal constitute about 350% of total cost,
and that the cost shares of the remaining five inputs are approximately equal (with chemicals
and energy each accounting for somewhat less than 10% of total cost, while capital and labor each
account for somewhat more).
134 MARGARET E. SLADE

then 3.u, will be 0.25 and an estimate of CT,,&,, is 0.0833.16


Equation (13) was reestimated with a priori restrictions on the coefficients of
the cost variables.” The result was’s
In o1 = 1.72 + 0.28 In Psi + 0.40 In SS1 - 0.23 In PKt
(6.86) (8.61) (-2.95)
- 0.33 In PL, - 0.49 In PE, - 0.11 In PCt - 0.07 In PT,, (20)
(-4.13) (-7.39) (-1.50) (-1.26)
RZ = 0.80, F = 9, Years : 1954-1976.

In Eq. (20), all of the coefficients have their expected signs and most are
statistically significant. If we assume that (Y~= 0.5 (see footnote 15), then
restrictions 5 can be used to solve for the parameters of the marginal cost function
defined by Eq. (7).ls The result is shown in Table II.
Theil[24 J proposed a test for the compatibility of prior and sample information
when the prior information consists of a set of stochastic linear restrictions as
defined by Eqs. (18). Theil’s compatability statistic is

y = (r - Rp)‘[s%(X’X)-1R’ + 9-J-‘(r - R/9), (21)


where 6 and s2are the estimates of p and u,,~obtained by applying OLS to Eq. (13)
[with error term u k N(0, uU2) appended]. Theil showed that, for large samples,
y will be distributed approximately as x2 with g degrees of freedom. The computed
value of y for Eq. (20) is 5.7, much less than 12.6, the value that would lead us to
reject the compatibility hypothesis at the 95% level of confidence. y is well within
the acceptable range; therefore, even though our sample is not large (T = 23),
we can feel reasonably confident in accepting the hypothesis that the prior
restrictions are compatible with the observed data.
The estimated output-price elasticity of the supply of secondary copper from
old scrap is 0.28. This elasticity can be separated into its constituent parts: l/a
and --ha,/cr. The value of LYobtained by applying restrictions 5 to the parameters
of Eq. 20 was 0.42. Therefore, supply is elastic if the prices of all inputs are held

r6 Suppose that ai = at/a = coefficient of Pi in Eq. (13), and that E[a] = 0.5, Eta<] = 0.12.
Then Erai] = E[ai/a] = 0.25 (Qi and (Y are independent by the homotheticity assumption).
If oi lies between zero and 0.5 with very high probability, then ai = 0.25 f 3u,,; = 0.25 rt 30,,
0.25 - 30, = 0, 0.25 + 3a, = 0.5, and &$ = 0.25/3 = 0.0833.
I7 No prior restriction was placed on the coefficient of Ps, the price elasticity of supply, which,
unlike the other elasticities, can be either positive or negative.
r* Various forms of distributed lags were tested (to see if the short- and long-run elasticities
differ), but none improved the equation. For example, when the lagged value of Q was introduced
(a geometric distributed lag), its estimated coefficient was 0.01, implying that 99% of desired
adjustment takes place within a year. In an industry where capacity has not grown, desired
changes in output can usually be accomplished fairly quickly by changing capacity utilization
rates.
I9 The restrictions result in three equations in four unknowns :
-J./U~Uis - 1.2 = x coefficient of Pi in (20) i = K, L, E, C, T

l/o[l l Xa.) = 0.3 = cdefbcient of Ps in (20)


= the output-price elasticity of supply
l?l cli + aa = 1, i = K, L, E, C, T (restriction 5).

If one of the unknowns (CU.)is assumed known (or set equal to its best guess), these equations can
be solved for the remaining unknown parameters.
COPPER RECYCLING VERSUS DISPOSAL 135

constant ((Y < 1 -+ l/a > 1). However, even though the partial output-price
elasticity is high, the net elasticity is very 10w.~O

VI. SIMULATION RESULTS


In this section, results from model simulat’ions for the 1977-1986 period will be
reported and used to test the impacts that various policies and future events have
on copper recycling rates. To perform the simulations, we incorporated Eqs.
(14)-(16) and (20) into a larger econometric model of the U.S. Copper industry
in which the pricing policies, costs, and output of the primary producers, as well
as total copper consumption, are endogenous. A complete description of the larger
model is available in Slade [22] ; only simulation results will be reported here.21

VI.1. Price Subsidies to Secondary Production


Because the disposal of wornout copper products is not costless, it might be
socially advantageous to subsidize secondary production at a rate that reflects
disposal costs. No simulation is needed to predict the effect of a per-unit subsidy
for secondary copper production. For example, suppose that the subsidy is set
at a level of five cents per pound of secondary copper produced. The price that
the secondary producer receives for copper will be raised by five cents, but,
because the price to the consumer is not changed, consumption will be unaffected.22
The average price of number two copper scrap between 1974 and 1976 was forty-
six cents per pound. Five cents thus represents an 11% increase in price. Because
the price elasticity of secondary supply is 0.28, the subsidy should result in a 3%
increase in secondary production (0.11 X 0.28 = 0.03), which in turn means a
0.5y0 increase in the market share of copper produced from old scrap.23 A model
simulation incorporating a price subsidy produces the same results as the simple
argument advanced above (a 0.5% increase over the base case in the market
share of copper produced from old scrap). The end result is predictably disappoint-
ing and may not be not worth the cost of the subsidy (approximately forty five
million dollars per year).24 See the second line of Table III.

VI.2. Alternate Energy Price Forecasts


Secondary copper production uses one-tenth to one-fifth the energy per ton
needed to produce copper from ores. We might, therefore, predict that an increase
20This net elasticity is in line with the results reported by Fisher et al. [13] who estimated the
price elasticity of the supply of secondary metal produced from old scrap to be 0.25.
t1 The results are reported as deviations from a base case of the market share of copper produced
from old scrap. The base case corresponds to linear growth in energy prices at 1974-1976 rates
and to a drop in the average yield of copper ores mined in the primary industry from 0.47% in
1975 to 0.39% in 1986. The two market shares (base case and alternative) are compared in 1986.
**For optimality, consumption should be taxed at a rate equal to the environmental cost of
disposal, and recycling should be subsidized at the same or a lesser rate (depending on whether
or not recycling removes harmful residuals at the same rate as consumption creates them) (Hoe1
[lS]). Such a policy discourages consumption, but it does not have a significant impact on the
market share of recycled copper (for the same reason that a subsidy alone is ineffectual).
*a Percentage changes in market share were calculated as (&‘/Cl - Q&&/(&&b), where Q
is production from old scrap in 1986, C is copper consumption in 1986, the subscript b refers to the
base case, and the primes correspond to the policy being tested.
*( Forty-five million dollars is the direct cost of the subsidy (the distribution effect). The social
cost of the subsidy (the allocation effect due to allocation costs of raising government revenue)
will be much less.
136 MARGARET E. SLADE

TABLE III
Simulated 1986 Market Share of Copper Produced from Old Scrap

Market share Percentage


Case (%I change

Base 14.1
Price subsidy 14.6 0.04
High-energy price 13.3 -6
Low-energy price 15.7 11
High-grade ores mined 12.3 -13
Low-grade ores mined 15.9 13

in energy price would result in a higher market share for recycled copper. However,
the issue is much more complex than just stated. As noted earlier, the supply of
copper from old scrap is a function of secondary copper price, the stocks of copper
scrap, and input prices
Q = &U's, SS, Pi), i = K, L, E, M, T, (22)
and secondary price is determined by primary price and aggregate economic
activity
Ps = Ps(Pp, Y). (23)
Substituting (23) into (22) we have
Q = Q(Ps(Pp, Y), SS, Pi>, i = K, L, E, M, T. (24)
The change in secondary supply due to a change in energy price can be obtained
by differentiat,ing Eq. (24) with respect to PEAR
dQ/dP* = aQ/aPs. aPs/dPp . dPp/dPE + aQ/aPs. aPs/aY. dY/dPE
+ aQ/a SS.dSS/dPE + aQ/aPE. (25)
When energy price goes up, the first part of Eq. (25) provides a positive
incentive to recycle-an increase in energy price causes secondary price to rise,
as primary price rises, and stimulates supply. However, all other terms in Eq.
(25) are disincentives to recycling. As industrial production falls, secondary
copper price falls (more than proportionally). The stocks of copper scrap are
smaller (because past consumption was smaller). And finally, energy is itself a
cost to the secondary producer. Because the various effects just noted work in
opposite directions, the net effect on secondary copper production of an increase
in energy price must be determined empirically.
The base case corresponds to linear growth in deflated energy prices at 1974-
1976 rates. Two alternatives were tested-constant deflated energy prices after
1977 and exponential growth in deflated energy prices at 1974-1976 rates-the
“low” and “high” energy price forecasts. The third and fourth lines of Table 111
show the results of these two simulations. Counter to the initial argument
advanced, that a higher energy price might improve the market share of recycled
copper, secondary copper’s market share is actually inversely related to energy
price (that is, the disincentives predominate as energy price rises).
z6Because we are considering deflated input prices, we will assume, for simplicity, that dPi/dPn
= 0, i f E. This assumption, of course, is not totally realistic because all prices change somewhat
with changes in energy prices.
COPPER RECYCLING VERSUS DISPOSAL 137

VI.3. Di$ering Ore Grades in Primary Production

In the past few decades, the average grade of copper ores mined in the U.S.
has declined. In the future, this trend will probably continue, but the exact
rate of decline is unknown. Unlike higher energy prices, which are a cost to the
secondary producer as well as to the primary producer, declining ore quality
does not affect secondary industry costs. We might, therefore, predict’ that the
market share of recycled copper will be inversely related to the average grade of
ores mined by the primary producers.
Equation (24) can also be used to show the effect that. changing ore quality
has on secondary production from old scrap. Differentiating Eq. (24) with respect
to ore grade, G, we have
dQ/dG = aQ/aPs’aPslaPp.dPpldG + aQ,‘aSS.dSS,ldG. (261

When grades fall, primary price rises, as primary production becomes more
costly, and the induced change in secondary price provides an incentive to
recycle. However, the fall in the stocks of copper scrap (a high copper price
discourages consumption) is a disincentive to secondary production. We can
assume that industrial production and other input prices are unaffected by higher
copper prices, and that, therefore, two of the disincentives to recycling en-
countered with higher energy prices have no counterparts with declining ore
grades.
In the base case, the average yield of ores mined by the primary producers
was assumed to fall from 0.47% in 1975 to 0.39% in 1986.26 Two alternative fore-
casts will now be considered-yields falling continuously to a low of 0.35% in
1986, t#he “low” ore-quality case, and yields falling to 0.45% in 1986, the “high”
ore-quality case. The last two lines in Table III show the results from these
simulations. As predicted, recycled copper’s market share is inversely related to
the average yield of ores mined by primary producers. The increase in secondary
price more than offsets the higher costs due to smaller stocks of copper scrap
available for collection.27

VII. SUMMARY AND CONCLUSIONS

Three sets of simulations were run-one involving subsidies for secondary


copper production, and the other two concerned with different assumptions about
future U.S. energy costs and ore quality in primary copper production. The
results from the simulation involving subsidies were predictably disappointing.
Most empirical studies of secondary materials have concluded that price elas-
ticities of supply are low. The model discussed here provides an explanat#ion for the
poor response of secondary production to price incentives-subsidizing the use of
scrap merely bids up the price of scrap. The simulations confirm these findings.
The initial hypothesis advanced about energy prices, that higher energy prices
would stimulate recycling (because secondary copper production is much less
energy intensive than primary production), was not confirmed. The simulations
26Yield is defined as metal produced divided by ore mined and is approximately 70% of grade
for copper.
w Because copper products are long lived, there will be little change in scrap stocks before 1986,
and then only in the last few years of the period.
138 MARGARET E. SLADE

indicate that the market share of recycled copper is actually inversely related to
energy price.28 In retrospect, this is not a surprising conclusion. Secondary
copper price is cyclical because the U.S. producer price is based to a large extent
on long-run average cost and does not, fluctuate enough to equate supply and
demand in the short run. Therefore, secondary price is extremely sensitive to
changes in aggregate demand, and the secondary producer is more disadvantaged
by slow economic growth (whether caused by high energy prices or some other
factor) than the primary producer. In addition, marginal energy costs for recycled
copper rise very steeply in the relevant range, implying that large copper price
increases are needed to improve recovery efficiency.
In contrast, to higher energy prices, which mean higher costs for both primary
and secondary producers, a decline in the yield of ores mined by the primary
industry does not affect secondary industry costs. The simulations involving
alternate assumptions about ore quality indicate that the market share of recycled
copper is inversely related to the yield of ores mined by the primary producers.
Although this result could be anticipated, its verification is reassuring, because
it means that, as a mineral commodity becomes scarce (more costly to extract and
process), recycling is stimulated automatically.
These conclusions do not seem to leave much room for government policy
(that is, policies designed to manipulate secondary supply are ineffectual whereas
processes bhat occur naturally-declining yields, for example-produce the
desired results). However, this should not be taken as a recommendation for
government inact,ion. The U.S. government has intervened in the past in several
ways to create inequities in metal markets-tax advantages for primary producers
(such as depletion allowances) and more favorable freight rates for ores than for
scrap are examples. Wherever these inequities exist, they should be removed,
even it’ the removal is not expected to result in large increases in secondary
production.

APPENDIX A: THE COST FUNCTION AND


ITS DUAL PRODUCTION FUNCTION
A homothetic production function is defined as
Q = J'CfGVl,
where f(X) is linear homogeneous and concave and F is monotonically increasing,
F[O] = O,and F[ z] --t m as z -+ 00. The dual total cost function is
TC = F-'[Q]c(P),
where c(P) is dual to f(X) (Shephard [21-J). With total cost function 6,
F-'[Q-J = Q"+'/SSa and c(P) = a0 n Pp.
t
Therefore, the production function dual to 6 is
Q = [i n &aif#a]l/(a+l) = [E n XiPi]l/ (rr+l)~p/ (o+lI,
, t

where Xi are the inputs whose prices are P;.


*8The conclusion that a high-energy price discourages recycling is not general. For example, in
a similar study of aluminum recycling (Made [B]), the opposite conclusion was reached-high-
energy prices stimulate secondary aluminum recovery.
COPPER RECYCLING VERSUS DISPOSAL 139

One implication of this production function is that an increase in the stocks


of scrap enables a larger output to be produced using the same input’s. For
example, if 01 = 0.5, then a 1OO~o increase in scrap stocks results in a 33%
increase in Q, when we hold the quantity of input,s constant.
TC is not a flexible cost function (a local second-order approximation to an
arbitrary cost function with the appropriate properties). The unit cost function
c could assume the translog form (Christensen et al. [lo)) which is flexible.
However, with five inputs, the equation that must be estimated [analogous to
Eq. (13) in the text] has 17 independent parameters, a very large number relative
to the number of observations (23). In addition, TC will still not be flexible (even
though c is) because Qa+l/&!+ is not a flexible form for F-I[&]. To make the
estimation of the supply equation t,ractable, the second-order terms of t,he
translog cost function were omitted.

APPENDIX B: DATA SOURCES


Copper prices, primary and secondary, were obtained from the U.S. Bureau
of Mines’ “Minerals Yearbook.” The primary price is an annual average of t,he
daily weighted average of producers’ delivered price quotations for refined
copper published in “Metal’s Week.” The secondary price is the price of number
two heavy copper scrap.Bl
Secondary copper production figures (production from old scrap) are published
by Copper Development Association in its “Annual Dat,a.” The Index of Indus-
trial Production is computed by the Board of Governors of the Federal Reserve
System and reported in “Industrial Production.”
Most of the input prices used are indices computed by the U.S. Department
of Labor’s Bureau of Labor Statistics in calculating the Wholesale Price Index
(WPI) and are reported in “Wholesale Prices and Price Indexes.” The series used
for the price of capital equipment (WPI 1134) is the index for industrial process
furnaces and ovens. The price of industrial chemical that was used (WPI 0611
0334 for 1954-1973 and 0613 0281 for 1974-1976) is the index for sulfuric acid.B2
The price of energy used in the estimation is a weighted average of price
indices for natural gas, residual fuel oil, middle distillate, and electricity, with
weights approximately,proportional to the Btu value consumed (0.4, 0.1, 0.1,
and 0.3, respectively). The WPI index for natural gas is 0531 0101, for the years
1958-1976. For the years before 1958, the natural gas price index from “Gas
Facts,” published by the American Gas Association, was used.B3 The WPI price
indices for residual fuel oil and middle distillate are 0574 and 0573, respectively.
Edison Electric Institute’s index for commercial and industrial large light and
power, published in its “Statistical Yearbook,” was used for the electricity price
index.
The Bureau of Labor Statistics reports wage rates for various industries in
“Employment and Earnings.” However, it does not collect data on wages in
secondary metal production. Therefore, the wage rate in primary metal production
was used as a proxy. Finally, the price of transportation services is a copper
B1Number two copper scrap is very high quality scrap that requires little further processing.
Therefore, this price is closer to an output price than an input price (seeSection 111.2).
B8Sulfuric acid is used a~ an electrolyte in refining.
B* For compatibility between the two series, the WPI natural gas price index (which starts in
1958) was regressed on the American Gas Association’s price index and ba&ca&ed.
140 MARGARET E. SLADE

transportation cost index compiled by Synergy, Inc., for the U.S. Bureau of
Mines.B4 The index was constructed from the Bureau of Labor Statistics’ transpor-
tation commodity code 10 (metallic ores), which goes back only to 1969, and
from Freight Commodity Statistics, Class one steam railways, compiled by the
U.S. Interstate Commerce Commission’s Bureau of Economics and St,atistics.

REFERENCES
1. American Metal Market, “Metal Statistics 1977,” Fairchild Publications, New York (1977).
2. R. C. Anderson and R. D. Spiegelman, Tax policy and secondary material use, J. En&m.
Econ. Manag. 4, 68-82 (1977).
3. B. T. Bower, Economic dimensions of waste recycling and reuse: Some definitions, facts, and
issues, in “Resource Conservation, Social and Economic Dimensions of Rmycling” (D. W.
Pearce and I. Walter, Eds.), New York Univ. Press, New York (1977).
4. M. S. Brown and J. Butler, “The Production, Marketing, and Consumption of Copper and
Aluminum,” Praeger, New York (1968).
5. J. A. Butlin, The prices of secondary materials and recycling effort, in “Resource Conservation,
Social and Economic Dimensions of Recycling” (D. W. Pearce and I. Walter, Eds.), New
York Univ. Press, New York (1977).
6. F. V. Carrillo, M. H. Hibpshman, and R. D. Rosenkranz, Recovery of secondary copper and
zinc in the United States, U.S. Bur. of Mines Information Circular 8822, Washington,
D.C. (1974).
7. P. F. Chapman, Energy conservation and recycling of copper and aluminum, Metals Materials
8, 311-319 (1974).
8. Charles River Associates, An economic analysis of the copper industry, prepared for General
Services Admin., Washington, D.C. (1970).
9. Charles River Associates, Policy implication of producer country supply restrictions: The
world aluminum-bauxite market, prepared for National Bur. of Standards, Washington,
D.C. (1976).
10. L. R. Christensen, D. W. Jorgenson, and L. J. Lau, Conjugate duality and the transcendental
logarithmic production function, Econometrica 39, 255-256 (1971).
11. D. K. Crampton, Production and use of copper, in “Copper” (A. Butts, Ed.), Reinhold,
New York (1954).
12. R. Edwards and D. Pearce, The effect of prices on the recycling of waste materials, Ree.
Policy 4, 242-248 (1978).
13. F. M. Fisher, P. H. Cootner, and M. N. Baily, An econometric model of the world copper
industry, Bell J. Econ. Manag. ~5%. 3, 568-609 (1972).
14. R. P. Grace, R. K. Turner, and I. M. Walter, Secondary materials and international trade,
J. Environ. Econ. Manug. 5, 172-186 (1978).
15. R. P. Grace, Metals recycling, a comparative national analysis, Res. Policy 4,249-256 (1978).
16. M. Hoel, Resource extraction and recycling with environmental costs, J. Environ. Econ.
Manog. 5, 220-235 (1978).
17. J. Johnston, “Econometric Methods,” 2nd ed., McGraw-Hill, New York (1972).
18. A. K. Miedema et al., The case for virgin material changes: A theoretical and empirical
evaluation in the paper industry, prepared for Environmental Protection Agency, Washing-
ton, D.C. (1976).
19. J. R. Moroney and J. M. Trapani, Options for conserving nonfuel, nonrenewable resources in
the United States, presented at the Resources for the Future/Electric Power Research
Institute Conference on Natural Resources Prices, Washington, D.C. (1979).
20. W. D. Schulze, The optimal use of non-renewable resources: The theory of extraction, J.
Environ. Ecm. Manag. 1, 53-73 (1974).
21. R. W. Shephard, “Cost and Production Functions, ” Princeton Univ. Press, Princeton, N.J.
(1953).
22. M. E. Slade, An econometric model of the domestic copper and aluminum industries: The

x4 This index can be found in the Final Report for the Joint Aluminum/Copper Forecssting
and Simulation Model, prepared by Synergy, Inc., for the U.S. Bureau of Mines, Washington,
D.C., 1977.
COPPER RECYCLING VERSUS DISPOSAL 141

effects of higher energy prices and declining ore quality on metal substitution and recycling,
unpublished Ph.D. dissertation, The George Washington University, Washington, D.C.
(1979). ,
23. V. L. Smith, Dynamics of waste accumulation: Disposal versus recycling, Quart. J. &on. 8,
600-616 (1972).
24. H. Theil, On the use of incomplete prior information in regression analysis, J. Amer. Statis.
Assoc. 54,401414 (1963).

You might also like