Impacts of market-based climate change policies on the US pulp and paper industry

12
* Corresponding author. Tel.: 001-617-353-5741; fax: 001-617-353- 5866. E-mail address: mruth@bu.edu (M. Ruth) Energy Policy 28 (2000) 259}270 Impacts of market-based climate change policies on the US pulp and paper industry Matthias Ruth!,*, Brynhildur Davidsdottir!, Skip Laitner" !Center for Energy and Environmental Studies, and the Department of Geography, Boston University, 675 Commonwealth Avenue, Boston, MA 02215, USA "US Environmental Protection Agency (EPA), Ozce of Atmospheric Programs, 501 3rd Street NW, Washington, DC 20001, USA Received 13 July 1999 Abstract Much of the policy debates on climate issues have centered on the optimal level and timing of policy interventions that cost-e!ectively reduce greenhouse gas emissions. We brie#y review the range of policy issues in our paper and assess the impact that climate-change policies may have on energy use and carbon emissions in the US pulp and paper industry. We then present results from time series-based analyses of changes in technologies and fuel mix, and compare the results with engineering-based technology analyses of US pulp and paper production (for a copy of the model and software send e-mail to mruth@bu.edu.). Projections, based on information for eight paper and paperboard categories, are presented for the years 1995}2020. The "ndings indicate that, under a wide range of speci"cations and policy assumptions, carbon emissions from fossil fuel use per ton of product are likely to decline. When combined with investment incentives, an additional cost-e!ective reduction in carbon emissions per ton of product will be realized. However, expected increases in output from the industry are likely to be higher than the reductions in energy and carbon intensities. The policy implications of these "ndings are discussed. ( 2000 Elsevier Science Ltd. All rights reserved. 1. Key issues a4ecting climate change policy instruments Scienti"c uncertainty about global climate change has been dominating policy discussions for over a decade. At the same time, a number of studies indicate that if climate change is real and irreversible, and if greenhouse gas emission control costs are initially low, the choice of appropriate policy instruments can be decoupled from the uncertainties about the natural science (Pearce, 1991; Nordhaus, 1991). Much of the uncertainty about the presence of human-induced climate change is being re- solved (IPCC, 1990; IPCC, 1995) and a host of cost- e!ective energy e$cient technologies to reduce carbon emissions identi"ed (Energetics, 1990; OTA, 1993; Inter- laboratory Working Group, 1997; Energy Innovations, 1997). Increased attention needs to be given to the e!ects of various climate change policies on both the adoption of these technologies and reductions in greehhouse gas emissions (Koomey et al., 1998). Market-based policy instruments have long been fa- voured by economists over command-and-control in- struments for their potential to minimize compliance cost by industry. This, in turn, will minimize the cost of such policies to society (Baumol and Oates, 1988; Tietenberg, 1990, 1996; Hoeller and Wallin, 1991; OECD, 1995). Market-based methods are also believed to give continu- ous dynamic incentives to adopt ever cleaner and more e$cient technologies and thus continuously reduce emis- sions. However, while most empirical studies examine the potential macroeconomic impacts of climate change policies such as on GDP and aggregate employment (Jaccard and Montgomery, 1996; Laitner et al., 1998), few studies concentrate on the expected dynamic e$ciencies of such policies within the context of individual indus- tries. Among the market-based instruments to reduce car- bon emissions are tradable carbon emission permits and carbon taxes. Since di!erent fuels have di!erent carbon content, absolute and relative prices of fuels are changed by those policies. As "rms attempt to reduce the cost of climate change policies, they are expected to curb fossil fuel use. This will be done either by reducing output or by changing energy consumption patterns. The latter choice 0301-4215/00/$ - see front matter ( 2000 Elsevier Science Ltd. All rights reserved. PII: S 0 3 0 1 - 4 2 1 5 ( 0 0 ) 0 0 0 0 9 - 4

Transcript of Impacts of market-based climate change policies on the US pulp and paper industry

*Corresponding author. Tel.: 001-617-353-5741; fax: 001-617-353-5866.

E-mail address: [email protected] (M. Ruth)

Energy Policy 28 (2000) 259}270

Impacts of market-based climate change policies on the USpulp and paper industry

Matthias Ruth!,*, Brynhildur Davidsdottir!, Skip Laitner"!Center for Energy and Environmental Studies, and the Department of Geography, Boston University, 675 Commonwealth Avenue, Boston, MA 02215, USA

"US Environmental Protection Agency (EPA), Ozce of Atmospheric Programs, 501 3rd Street NW, Washington, DC 20001, USA

Received 13 July 1999

Abstract

Much of the policy debates on climate issues have centered on the optimal level and timing of policy interventions thatcost-e!ectively reduce greenhouse gas emissions. We brie#y review the range of policy issues in our paper and assess the impact thatclimate-change policies may have on energy use and carbon emissions in the US pulp and paper industry. We then present resultsfrom time series-based analyses of changes in technologies and fuel mix, and compare the results with engineering-based technologyanalyses of US pulp and paper production (for a copy of the model and software send e-mail to [email protected].). Projections, based oninformation for eight paper and paperboard categories, are presented for the years 1995}2020. The "ndings indicate that, under a widerange of speci"cations and policy assumptions, carbon emissions from fossil fuel use per ton of product are likely to decline. Whencombined with investment incentives, an additional cost-e!ective reduction in carbon emissions per ton of product will be realized.However, expected increases in output from the industry are likely to be higher than the reductions in energy and carbon intensities.The policy implications of these "ndings are discussed. ( 2000 Elsevier Science Ltd. All rights reserved.

1. Key issues a4ecting climate change policy instruments

Scienti"c uncertainty about global climate change hasbeen dominating policy discussions for over a decade. Atthe same time, a number of studies indicate that if climatechange is real and irreversible, and if greenhouse gasemission control costs are initially low, the choice ofappropriate policy instruments can be decoupled fromthe uncertainties about the natural science (Pearce, 1991;Nordhaus, 1991). Much of the uncertainty about thepresence of human-induced climate change is being re-solved (IPCC, 1990; IPCC, 1995) and a host of cost-e!ective energy e$cient technologies to reduce carbonemissions identi"ed (Energetics, 1990; OTA, 1993; Inter-laboratory Working Group, 1997; Energy Innovations,1997). Increased attention needs to be given to the e!ectsof various climate change policies on both the adoptionof these technologies and reductions in greehhouse gasemissions (Koomey et al., 1998).

Market-based policy instruments have long been fa-voured by economists over command-and-control in-struments for their potential to minimize compliance costby industry. This, in turn, will minimize the cost of suchpolicies to society (Baumol and Oates, 1988; Tietenberg,1990, 1996; Hoeller and Wallin, 1991; OECD, 1995).Market-based methods are also believed to give continu-ous dynamic incentives to adopt ever cleaner and moree$cient technologies and thus continuously reduce emis-sions. However, while most empirical studies examine thepotential macroeconomic impacts of climate changepolicies such as on GDP and aggregate employment(Jaccard and Montgomery, 1996; Laitner et al., 1998), fewstudies concentrate on the expected dynamic e$cienciesof such policies within the context of individual indus-tries.

Among the market-based instruments to reduce car-bon emissions are tradable carbon emission permits andcarbon taxes. Since di!erent fuels have di!erent carboncontent, absolute and relative prices of fuels are changedby those policies. As "rms attempt to reduce the cost ofclimate change policies, they are expected to curb fossilfuel use. This will be done either by reducing output or bychanging energy consumption patterns. The latter choice

0301-4215/00/$ - see front matter ( 2000 Elsevier Science Ltd. All rights reserved.PII: S 0 3 0 1 - 4 2 1 5 ( 0 0 ) 0 0 0 0 9 - 4

includes increased investments in energy e$cient techno-logy, substitution away from high carbon fuels, or a com-bination of each of these responses (Hoeller and Wallin,1991; Pearce, 1991).

In the short term, carbon policies will raise the price ofpurchased fuels and, ceteris paribus, reduce pro"tabilityin the industry. This may have the e!ect of lowering the"nancial capital available to invest in new equipment.Moreover, empirical studies and industry surveys furthersuggest that "rms generally resist energy saving invest-ments despite the potentially large rates of return fromenergy e$cient technologies (DeCanio, 1993, 1994, 1996).This is despite the fact that those investments can providea source of competitive advantage (Porter, 1990). A com-bination of market-based climate change policies withinvestment incentives may not only help overcome suchhurdles for investment but even tease out some otherwiseunrealized potentials for e$ciency improvement (Laitneret al., 1999).

Several studies base their analyses on energy e$ciencysupply curves that have been constructed for individualindustries by rank-ordering alternative technologies ac-cording to their marginal cost of conserved energy (Meieret al., 1983; Joskow and Marron, 1993; Ross et al., 1993;Stoft, 1995). On the basis of evaluations using e$ciencysupply curves, comparisons have been made of potentialenergy e$ciency improvements within industries (Rosset al., 1993) and across countries (Worrel, 1994). Signi"-cant potentials have been found across all major industrygroups and industrialized countries (Woodru! et al.,1997). E$ciency supply curves have been popular be-cause they explicitly capture the engineering perspectiveof individual technologies that might be chosen by di!er-ent industries. At the same time, a relationship is estab-lished between technical features and cost savings, thusproviding a direct connection to economic decision mak-ing at the "rm and industry level.

While e$ciency supply curves have found widespreadapplication in sectoral energy analyses, their use has notyet made major inroads into models of national energyuse in part due to the higher level of aggregation innational energy models. For example, speci"cations ofpotential e$ciency improvements and changes in fuelmix in the Industrial Module of the National EnergyModeling System (IM-NEMS) and derivative studies(Geller and Nadel, 1994; Train et al., 1995) are donewithout e$ciency supply curves. Instead, they quantifythe energy consumption per dollar of output by usingeconometrically estimated functions of energy prices andautonomous technical change (EIA, 1998). This ap-proach has been criticized for imperfectly capturing thedeclines in relative costs of technologies due to energyprice and learning-induced improvements. It also tendsto lock the models into technology choices that arebiased towards existing patterns, and relies on ratherarti"cal assumptions of pro"t and utility maximization

(DeCanio and Laitner, 1997). The latter critique holdsalso for the MARKAL model which is widely used in theframework of the International Energy Administration'sEnergy Technology Systems Analysis Program to cal-cuate least-cost system con"gurations using linear pro-gramming (Fishbone et al., 1983).

2. Dynamics of energy use and carbon emissions in the USpulp and paper industry

The US pulp and paper industry (SIC 26) is one of themore energy intensive manufacturing industries. It ac-counts for 12% of total manufacturing energy use, 95%of which is consumed by pulp, paper and paperboardmills (Bureau of the Census, 1995). The energy intensityof the industry has declined over the years, from 38.32million Btu per short ton of output in 1972 to 29.95million Btu per short ton in 1995 (AFPA, various years).E$ciency increases have been achieved mainly throughimprovements in existing technologies, the retirement ofless e$cient facilities, and better housekeeping practices.In contrast, little of the improvements have been theresult of major technological innovations in the industry.During that same period, the industry has also increasedits use of both natural gas and onsite generation ofelectricity powered by various forms of wood waste.Hence, the industry's emissions from fossil fuels per tonof output declined over the last two decades.

The US demand for paper products increased over thelast two decades to the highest per capita rates in theworld. Today, the US pulp and paper industry accountsfor approximately 30% of total world production capa-city (Gilbreath, 1995). Given continued growth in de-mand for its products, further capacity increases arelikely. This means that production of the US paper in-dustry will continue to expand (Freeman Miller, 1996).Even in the absence of climate change policies, energye$ciency improvements will be realized. Given theseindustry trends, the following question is of main interestto this study: How will climate change policies a!ect boththe industry's energy use and pattern of carbon emis-sions? To answer these questions, we concentrate on thee!ects of a policy-induced increase in the cost of carbon.We also examine the impacts of investment incentivesdesigned to promote energy e$ciency improvements. Inour analysis we make use of our own time series analysisand of key assumptions and data available from both theLIEF model (Ross et al., 1993) and IM-NEMS (EIA, 1998).

3. Model description

For the analysis of the impacts of a policy-inducedincrease in the cost of carbon and energy we developeda dynamic computer model that traces all major material

260 M. Ruth et al. / Energy Policy 28 (2000) 259}270

Fig. 1. (a) Historic and forecast production rate by product category;(b) historic and forecast production rate by product category.

and energy #ows through various stages of the pulp andpaper production process. That model consists of thefollowing interrelated modules:Module 1. Simulation of historic and future paper andpaperboard production for a total of eight di!erent prod-uct categories.Module 2. Simulation of historic and future energy use inthe industry by process step and fuel type.Module 3. Calculation of carbon emissions from the pa-per industry including carbon emissions from electricitygeneration.Module 4. Simulation of energy use in electricity genera-tion.Module 5. Estimation of energy and policy-induced ex-penditures.Module 6. Estimation of technological potentials andfuture energy use on the basis of LIEF.

Dynamic modeling (Ruth and Hannon, 1997) is usedto interrelate these six modules through various feedbackprocesses with each other. Many of these feedback pro-cesses exhibit time-lags and non-linearities. Parameterestimates that guide the relationships between variableswithin or between submodules either are estimated fromtime-series data or based on engineering information.Estimation techniques include multivariate regressionsusing both simultaneous and lagged independent vari-ables. Seemingly unrelated regression models (Zellner,1962) are used to capture interdependence in changes inthe fuel mix, and polynomial distributed lags are em-ployed to model time-delayed system responses (Pindyckand Rubinfeld, 1991). Model (parameter) speci"cation isbased on conventional hypothesis tests (t, F-test and R2)as well as extensive diagnostics such as Lagrange multi-plier tests for heteroscedasticity (Breusch and Pagan,1979) and serial correlation (Godfrey, 1978), and theDickey Fuller test for unit roots (Dickey and Fuller,1981). Each module is described in more detail below.Based on these modules, we assess industrial responses tothe increased cost of carbon for the years 1995}2020. Tocon"rm the reliability of the time series-based portion ofthe model, we simulated the industry's energy use pro"lefor the years 1972}1994 and compared the results to thehistorical data. The results indicated an absolute meanerror of 1.13%. This suggests, therefore, proper modelspeci"cation.

Module 1: Production rates. To generate productionforecasts, we disaggregated industry output into fourpaper categories (newsprint, tissue, printing and writing,packaging and industrial paper), four paperboard cat-egories (bleached and unbleached kraft, recycled, andsemichemical), and one construction paper category.This is the conventional breakdown into di!erent paperand paperboard categories used by the American Forestand Paper Association (AFPA) and the American PaperInstitute (API). However, for our model we excluded the

construction paper category because of vagueness in thede"nition of this category and the lack of comparabilityof data for that category across various data sources (e.g.Freeman Miller, various years; API, various years). Sincethe construction paper category is a minor part of thepaper and paperboard industry, accounting for approx-imately 2% of total industry output, we assume that thisomission will not signi"cantly in#uence our results.

Forecasts of paper and paperboard production rateson the basis of historic data from API (various years) andFreeman Miller (various years) were generated by usingvarious univariate ARIMA processes, depending on themost appropriate functional form to generate each fore-cast. To determine if or how often detrending was needed,we examined partial and autocorrelation parameters. Thechoice among di!erent models was made by using bothAkaike (1973) and Schwartz (1978) criteria as well asexamining partial and autocorrelation parameters. Theresults of the ARIMA forecasts are shown in Fig. 1a and b.

M. Ruth et al. / Energy Policy 28 (2000) 259}270 261

Consistent with industry practice, output of each productcategory is measured in short tons.

Module 2: Energy use in the industry. The module ofenergy use in the pulp and paper industry is designed tocapture changes in energy intensity and fuel mix. Energyintensity is measured in energy use (Btu) per ton ofoutput. Fuel prices are used as explanatory variables todescribe changes in energy intensity. This has the advant-age of better identifying the relationships between energyrequirements of a ton of output compared to e$ciencymeasures based on a dollar value of output. For example,many models of energy use evaluate energy intensities asthe number of Btus per dollar value of output (Patterson,1996). If that measure is regressed against energy price,then part of the explanatory power of energy prices maybe attributable to their impact on product price, not ontechnology choice.

Changes in energy intensity are endogenous to themodel. Decreased energy intensity is the result of collec-tive improvements that take place within companies asnew technologies are being implemented and as old onesare retired. To model the decrease in energy intensity, themodel assumes a logistic relationship between energy useper unit of output on the one hand, and cumulative paperand paperboard production (estimated in Module 1) andfuel prices on the other hand. As more paper and paper-board is being produced by the industry, the more ad-vanced its technology is likely to become. The logisticrelationship between cumulative production and energye$ciency re#ects decreasing returns to future rates oftechnological improvements (Yelle, 1979; Ruth, 1993).The regression results indicate that a doubling in cumu-lative production is expected to coincide with an 11.09%decrease in energy intensity. Fuel prices enter the equa-tion in lagged, logarithmic form, re#ecting the fact thattechnology adjustments to energy price changes do notoccur instantaneously and that there are decreasingreturns to the e!ect of price-induced technology changes(Table 1).

Using seemingly unrelated regressions (Zellner, 1962),changes in fuel mix are estimated as functions of rates ofproduction and absolute or relative fuel prices. The use ofthe seemingly unrelated regression model enables us tocapture interfuel substitution by estimating demandequations for each fuel simultaneously. The use of rela-tive fuel prices enables us to capture the impact of in-creased cost of carbon on changes in the fuel mix in theindustry (Table 1). The percentage share of three energytypes classi"ed in o$cial statistics as steam, `othera andliquid propane gas are held constant at the average of thehistorical percentages from 1990 to 1994. Those threecategories accounted in 1994 for approximately 2% oftotal fuel use in the pulp and paper industry. Hence, thissimpli"cation should not signi"cantly a!ect our "nalresults. The econometric analysis indicates signi"cantin#uence of lagged energy prices, production rates and

cumulative production on the industry's choice of fuel(Table 1).

In 1994, 56.52% of the industry's energy need wassupplied through self-generation. Bark, wood chips,spent liquor, other residuals and wastepaper are burnedto generate steam or heat. Most of these wastes accrue onsite and are not purchased by the industry. Regressionresults indicate that the share of self-generated energywill likely increase with increased production and as theaverage price of purchased fuels increases (Table 1).

Historical energy prices were obtained from the En-ergy Information Administration's Annual Energy Out-look 1998 (AEO98) (EIA, 1997). Future energy prices areexogenously given by the Energy Information Adminis-tration's Annual Energy Outlook base/reference caseforecasts that extend until the year 2020 (EIA, 1997). Weconstructed non-weighted averages of all fuel prices andof the prices of primary fuels, and used those averages incombination with individual fuel prices to estimate cha-nges in energy e$ciency and individual fuel use over time(Table 1). Our estimate of changes in energy intensities inthe pulp and paper industry is considerably more conser-vative than those estimated in the AEO98 scenario. Yet,the use of AEO98 fuel price forecasts are justi"ed becausethese estimates are not based on the performance ofindividual industries but on the performance of the na-tional economy and the industrial sector as a whole.Using AEO98 fuel prices in the model assumes that theseprices in#uence behavior by the pulp and paper industry.But we assume the reverse is not true. In other words, thepulp and paper industry's energy use does not noticeablyin#uence industrial energy prices as a whole. The higherthe share of self-generated energy in the pulp and paperindustry, the more realistic is this assumption.

Module 3: Carbon emissions. Carbon emissions fromthe pulp and paper industry are calculated by multiply-ing the industry's use of each fuel with the carbon contentper Btu of that fuel. Net carbon emissions are de"ned asemissions associated with the generation of purchasedelectricity and from the direct use of fossil fuels bythe industry, excluding emissions from residual fuels.Thus, in our calculation of net carbon emissions wein e!ect treat biomass fuels as carbon-neutral. Growth,harvest, transport and preparation of biomass fuels,however, may contribute to the atmosphere's carbonbudget. Additionally, self-generation of heat andsteam in pulp and paper production also includes burn-ing of fossil-fuel based chemicals that are used atvarious stages of the production process, such as thebleaching of pulp. Yet, those additions to the carbonbudget are small in comparison to the direct emissionsfrom fossil fuel use by pulp and paper mills (Mann andSpath, 1997).

To capture all carbon emissions from the productionof pulp and paper (excluding those from forestry andtransportation), we de"ne and calculate gross carbon

262 M. Ruth et al. / Energy Policy 28 (2000) 259}270

Tab

le1

Sum

mar

yof

regr

ession

resu

lts

(t-s

tatist

ics

inpar

enth

eses

)D

epen

dentva

riab

les

Reg

ress

ors

Funct

iona

lfo

rm&

lags

Ener

gye$

cien

cy(lo

gn,

Mill

ion

Btu

/t)

Nat

ura

lga

s(b

illio

nBtu

)C

oal(lo

g)(b

illio

nBtu

)R

esid

ual

fuel

oil

(billio

nBtu

)Ele

ctrici

ty(b

illio

nBtu

)Se

lfge

nera

ted

ener

gy(lo

g)(b

illio

nBtu

)

Distillat

efu

eloi

l(%

ofto

tal

Ener

gyus

e)

Con

stan

t4.

115

!65

2804

9.63

113

.690

3177

127.

123

!16

5519

7.17

56.

061

0.04

6(4

2.38

6)(!

10.0

25)

(72.

179)

(7.2

02)

(!33

.767

)(8

.614

)(2

.734

)C

umul

ativ

epro

duct

ion

index

Log

!0.

161

(!5.

841)

Cum

ulat

ive

pap

erpro

duct

ion

Log

!20

1155

.338

(!6.

2906

9)

Tota

lpap

erpro

duct

ion

(thous

and

tons)

Log

6258

53.8

00!

0.00

0004

0516

7456

.253

0.70

3!

0.00

3(1

0.60

6)(!

2.23

8)(3

4.97

5)(1

0.88

6)(!

2.29

7)

Coa

lpr

ice/

elec

tric

ity

price

(t!

2)!

4.26

2(!

12.0

19)

(t!

4)!

0.94

3(!

3.21

8)R

esid

ual

fuel

oilp

rice

($/M

illio

nBtu

)Log

(t!

2)!

1347

65.5

20(!

7.72

8)

Ave

rage

ener

gyprice

!($/

Mill

ion

Btu

)(t!

3)0.

028

(5.6

64)

Prim

ary

ener

gyprice

(PEP

)"Log

(t!

3)!

0.03

3(!

2.99

5)($

/Mill

ion

Btu

)(t!

3)!

7779

8.92

5(!

6.86

9)C

oalprice

/Nat

ura

lga

sprice

(t)

5616

8.93

8(2

.509

)Ele

ctrici

typrice

/PEP

(t)

!89

61.9

99(!

5.94

3)(t!

1)!

7664

.524

(!4.

751)

Distillat

efu

eloi

lprice

($/M

illio

nB

tu)

(t)

!0.

002

(!8.

496)

LM

s2#

0.65

60.

976

4.95

92.

113

2.37

32.

234

2.37

5(0

.883

)(0

.913

)(0

.292

)(0

.549

)(0

.667

)(0

.525

)(0

.498

)LM

s2(1

)$1.

450

0.56

40.

586

1.70

91.

704

0.40

30.

419

(0.2

28)

(0.4

53)

(0.4

44)

(0.1

91(0

.192

)(0

.526

)(0

.517

)A

djust

edR

20.

954

0.82

60.

959

0.93

20.

988

0.98

60.

939

!Ave

rage

ener

gypr

ice

isa

non

wei

gtht

edav

erag

eofal

lfu

elpr

ices

."Prim

ary

ener

gyprice

isa

nonw

eigh

ted

aver

age

ofpr

imar

yen

ergy

price

s.#L

agra

nge

multip

lier

estim

ate

ofhet

eros

ceda

stic

ity

(Godf

rey,

1978

),sign

i"ca

nce

leve

lin

pare

nthes

es.

$Lag

range

mul

tipl

ier

estim

ate

ofse

rial

corr

elat

ion

(Bre

usc

han

dPag

an,1

979)

,sig

ni"ca

nce

leve

lin

pare

nthes

es.

M. Ruth et al. / Energy Policy 28 (2000) 259}270 263

Table 2Carbon content of purchased fuels!

Fuel type Source Metric tons ofcarbon per billionBtu (1994, value)

Coal EIA, 1994 25.61Coal (electricity generation) EIA, 1994 25.71Natural gas EIA, 1994 14.47Residual fuel oil EIA, 1994 21.49Oil (electricity generation) EIA, 1994 19.95Liquid petroleum gas EIA, 1994 17.02Distillate fuel oil EIA, 1994 19.95

!All wood-based products are assumed to be carbon-neutral for pur-poses of this analysis.

Fig. 3. Base scenario energy use by source.

Fig. 2. Base scenario energy use in US pulp and paper production.

emissions as the sum of all fuel-speci"c emissions, includ-ing emissions from residual fuels (such as bark, spentliquor, and wastepaper). Carbon coe$cients by fuel typeare listed in Table 2 together with the sources of therespective data.

Module 4: Fuel use in the electricity sector. Part of theenergy used by the pulp and paper industry is in the formof purchased electricity. To properly re#ect the industry'sfuel requirements and carbon emissions, this modulecalculates fuel requirements by the electricity sector tosupply electricity to pulp and paper mills. The module isdriven by demand for electricity, as calculated in Module2, and uses actual and forecast data on fuel mix in theelectricity sector as reported in AEO98 up to the year2020 (EIA, 1997).

Module 5: Energy expenditures and policy-inducedpayments. Energy expenditures are estimated as the prod-uct of fuel use, as estimated by the model, and fuel pricesas forecast in AEO98 (EIA, 1997a) using EIA's base/ref-erence case forecasts. Policy-induced payments are sim-ilarly estimated through forecast fuel use and assumedcost of carbon. Net present value of annual and cumulat-ive cost of carbon payments and energy expenditures arecalculated using a 5% discount rate.

Module 6: LIEF technology adoption. Time-series-basedestimates of changes in energy intensities can be replacedin the model by using to various extents the assumptionsof LIEF discussed in more detail above and in Ross et al.(1993). Production forecasts, changes in fuel mix withinthe group of fossil fuels, and expansion of selfgeneratedenergy remain time-series based because insu$cient de-tail is provided in LIEF on these issues. Calibration hasbeen carried out to translate energy intensities measured inBtu per dollar value of shipments into purely physical units.

4. Impacts of increased carbon and energy cost

Parameters from the time-series analysis provide thespeci"cation of our base scenario against which alterna-

tive technology investment and policy scenarios are com-pared. Two sets of policy scenarios examine, respectively,the impact of an increase in the cost of carbon and theimpact of a policy that stimulates investment in cost-e!ective energy saving technologies (investment incen-tives). Results without climate change policy establishour base scenario and are shown in Figs. 1}5. Productiongrowth rates range from !0.34% for packaging andindustrial paper to 1.70% for printing and writing(Fig. 1a and b). The non-weighted, average annualgrowth rate of the industry is 1.65%. Output from theindustry as a whole (excluding construction paper)reaches 100 million short tons in 2002, and 129 millionshort tons in 2020. This is the last year of our simulations.Our production forecast is slightly more conservativethan USDA Forest Service's estimates of 100 million tons

264 M. Ruth et al. / Energy Policy 28 (2000) 259}270

Fig. 4. Base scenario carbon emissions by source.

Fig. 5. Base scenario net carbon emissions per ton of output and energyuse/t of output.

Table 3Cost of carbon for each fuel type (dollars per million Btu)

Fuel type 50$/t carbon 100$/t carbon

Electricity 2.39}2.59 4.79}5.18Coal 1.28 2.56Natural gas 0.72 1.45Residual fuel oil 1.07 2.15

by the year 2000 (Haynes et al., 1995; Ince, 1994) and isless aggressive than the AEO98 forecast of 134 millionshort tons in 2020 (EIA, 1997).

To simulate the e!ects of increases in the cost ofcarbon on energy e$ciency, fuel mix and carbon emis-sions, we assume an increase in the price of each pur-chased fuel proportional to its carbon content. The costof carbon used in the di!erent scenarios is either $50 or$100 per ton of carbon. Such increases fall within therange suggested by other studies on climate change pol-icies (e.g. Ekins, 1996; Nordhaus, 1993). Since, the carboncontent per Btu of coal is highest among fossil fuels, theprice of coal increases proportionally more than that ofoil and gas. The e!ect of the cost of carbon on electricity

prices depends on the fuel mix in electricity generationand hence #uctuates accordingly (Table 3). We assumethat the policy is implemented in the year 2000. Detailedresults for the alternative costs of carbon are listed inTable 4. The impacts of investment policies are alsocompared in Table 4 as discussed below.

5. Impacts of investment policies

The base and policy scenarios described above capturethe e!ects of learning by doing, price-induced e$ciencyimprovements, and price-induced fuel switching on en-ergy use and carbon emissions. These scenarios assumethat industry behavior with climate change policy isqualitatively consistent with observed, historic behavior.Increases in the cost of carbon that are modeled hererival the fuel price increases observed in the 1970s and1980s in response to the oil price shocks. However, unlikethose price shocks, the price increases in this analysis arepersistent over a 20-year time frame. As a result, theywould likely trigger signi"cant changes in the behavior ofthe industry. In the event that such changes do occur,smaller increases in the cost of carbon or energy may besu$cient to reduce emissions to acceptable levels. Gov-ernment policies that help disseminate technology in-formation or provide incentives for faster technologyadoption can help stimulate implementation of moreenvironmentally benign production processes. Voluntaryindustry initiatives can also improve environmentalperformance.

How would climate change policies a!ect net carbonemissions by the pulp and paper industry if there werea more rapid adoption of energy e$cient technology anda more rapid movement from high to lower-carbon pur-chased fuels? To answer this question we "rst substitutedassumptions from LIEF (Ross et al., 1993) for the time-series based forecasts of future energy intensities. Speci"-cally, we assumed an e$ciency gap between actual andbest practices in 1995 of 20%. The annual rate at whichnew technologies penetrate to close this gap was assumedto be 10%. These assumptions are based, in part, on theavailability of black liquor gasi"cation technology whichindustry believes can replace as much as 80% of thesigni"cantly less-e$cient Tomlinson recovery boilers

M. Ruth et al. / Energy Policy 28 (2000) 259}270 265

Table 4Summary of model results

Base scenario $50/t carbon $100/t carbon Investment-LedPolicy at $50/carbon

Total energy use (% change from 1990levels to 2020)

39.67 38.38 37.39 20.38

Total purchased fuel Use (% change from1990 levels to 2020)

33.24 15.86 3.23 0.36

Total self-generated energy (% changefrom 1990 levels to 2020)

44.58 55.59 63.48 35.68

Energy intensity, energy use/output (% changefrom 1990 levels to 2020)

!15.14 !15.92 !16.53 !26.86

Years gained in e$ciency improvement 3 4 17Net carbon emissions (% change from 1990levels to 2020)

30.10 18.04 9.64 2.93

Carbon emissions from selfgenerated energy(% change from 1990 levels to 2020)

44.58 55.59 63.48 35.68

Net carbon emissions per ton of output(% change from 1990 levels to 2020)

!16.13 !23.90 !29.32 !33.64

Energy MixCoal (% share of total energy use) 1990

202013.938.49

6.01 4.75 6.03

Residual fuel oil (% share of totalenergy use)

19902020

6.210.00

0.00 0.00 0.00

Electricity (% share of total energy use) 19902020

6.548.28

8.80 9.09 8.83

Natural gas (% share of total energy use) 19902020

16.5722.71

19.57 16.85 19.62

Self-generated energy (% share of totalenergy use)

19902020

56.6958.69

63.74 67.46 63.90

Financial ParametersCumulative present value of energyexpenditures, 2000}2020 (billion 1994 $,5% discount rate)

52.17 63.87 74.06 57.96

Cumulative present value of cost of carbonpayments, 2000}2020 (billion 1994 $,5% discount rate)

13.49 25.31 12.25

Energy expenditures in 2020 (billion 1994 $) 5.49 6.69 7.71 5.84Policy-induced payments in 2020(billion 1994 $)

1.51 2.79 1.31

over the next 20 years (Larson et al., 1998). The capitalrecovery factor for the indsutry was assumed to be 15percent rather than the more typical 33%.

When added together, these assumptions imply thepresence of policies that foster the accelerated adoptionof advanced technology. Such policies might include, forexample, allowing higher depreciation rates, providinginvestment incentives through tax credits, or providingregulatory #exibility with respect to demonstrating newtechnologies. When combined with the cost of carbon,the policies might result in carbon emission levels asreported in Table 4 and in Fig. 6.

6. Comparisons of results

A comparison across climate change policies (Fig. 7)reveals that investment incentive can cause signi"cant

reductions in net carbon intensities over and above thegains already achieved by a cost of carbon policy. Cost ofcarbon increases do not only lead to smaller energysavings and emissions reductions but industry responsesare time-delayed, following `normala procedures to retireand replace equipment.

Under the investment incentive scenario, the samelevel of energy e$ciency that is reached in the basescenario by the year 2020 is met 17 years earlier. Inthe cost of carbon scenarios e$ciency gains are acceler-ated by only 3 to 4 years, depending on the severity of thepolicy instrument. In all scenarios total energy use in-creases in the long run because industry output rises atrates that exceed its rates of reduction in energy inten-sities (Fig. 7). In the investment-led policy scenario, how-ever, total energy consumption temporarily declinessince the rate of energy e$ciency improvement is greaterthan the growth in output.

266 M. Ruth et al. / Energy Policy 28 (2000) 259}270

Fig. 6. Rates of net carbon emissions in 2020 for alternative cost ofcarbon with and without investment incentives.

Fig. 7. Total energy use per ton of output in the base scenario andunder selected climate change policies.

Fig. 8. Total energy use in the base scenario and under selected climatechange policies.

Base scenario energy expenditures are estimated toincrease from $4.25 billion in 1990 to $5.49 billion in2020. This implies that in 2020 the industry is expected tospend $42.47 per ton of output on purchased energy,down from $54.14 in 1990. All monetary values are inconstant 1994 dollars. In comparison, the cost of carbonscenarios lead to total energy expenditures (less cost ofcarbon payments) in 2020 of $5.18 and $4.92 billion forrates of $50 and $100 per ton of carbon, respectively. Thisimplies that the industry is estimated to spend on energy$40.15 to $37.99 for each ton of output. In the investmentincentive scenario, combined with an increase of $50 per

ton of carbon, total energy expenditures (less cost ofcarbon payments) in 2020 are $4.53 billion. This suggeststhat the industry is spending $34.98 per ton of output ifadoption of new technologies is fostered beyond the ratesthat are consistent with historic behavior. Energy expen-ditures per ton of product in the case of a more aggressivetechnology adoption are $5.17 lower than if no changesin behavior occurred.

Similar trends in energy intensities and total energyuse hold for net carbon emissions (Figs. 7 and 8). Declinesin net carbon intensities in response to energy e$ciencyimprovements and fuel switching are not su$cient to atleast balance increases in industry output (Figs. 9 and 10).

A comparison of time series-based results with resultsgenerated by IM-NEMS for the AEO98 indicate thata strict econometric approach gives a considerably morepessimistic outlook (see comparison in Table 5). Di!er-ences in results are mainly due to the assumption thatsomewhat more e$cient new capital in the paper indus-try replaces older capital only at historical rates. IM-NEMS thus does not allow for additional industry re-sponses beyond merely those triggered by an energy pricesignal. On the other hand, "ndings from the EPA/LBNL-NEMS Study (Koomey et al., 1997), which uses a combi-nation of LIEF and IM-NEMS, show how a combinationof a small carbon price with appropriate investment incen-tives can accelerate e$ciency improvements.

Under a business-as-usual scenario the EPA/LBNLstudy shows that the paper industry will decrease itsenergy intensity by 0.84% per year. Using LIEF policyassumptions with only a $23 carbon price, and assuminga capital recovery factor of 15%, the average annualdecline in energy intensity is increased to 2.1%. In thiscase, the energy savings to the industry are greater than

M. Ruth et al. / Energy Policy 28 (2000) 259}270 267

Table 5Comparison of time series-based scenarios with results from IM-NEMS (AEO98) and EPA/LBNL-NEMS analysis

Time-series basedbase scenario

AEO 98baseline

Time-series basedLC Scenario!

EPA/LBNLNEMS"

Paper and paperboard 2010 114.1 126.5 114.1 126.8production (Mshort tons) 2020 129.5 134.3 129.5 134.7Total energy consumption 2010 3.2 2.7 3.1 2.4(Quads) 2020 3.4 2.6 3.3 2.2Energy use/t of Output 2010 28.0 21.3 27.2 18.9(MMBtu/t) 2020 26.6 19.4 25.4 16.3

!Scenario with a $50 increase in cost of carbon."EPA/LBNL study's high e$ciency/low carbon scenario, with a $23 cost of carbon (Koomey et al., 1998).

Fig. 9. Net carbon emissions per ton of output in the base scenario andunder selected climate change policies.

Fig. 10. Net carbon emissions in the base scenario and under selectedclimate change policies.

the energy price increase that is driven by a $23 per toncost of carbon. From those results it is clear that allowingfor adoption and more rapid penetration of new tech-nologies beyond mere price-induced increases in energye$ciency and fuel substitution, signi"cant increases inenergy savings are possible.

The di!erence between the results presented in theAEO98 reference case, the EPA/LBNL study, and thetime series-based scenarios presented in this paper high-light the vast potential for energy savings and carbonemissions reductions in the paper industry. It also under-scores the importance of removing market barriers forenergy e$cient technologies.

7. Discussion

Climate change policy instruments such as increasedcost of carbon have frequently been proposed as tools to

stimulate cost-e!ective reductions in industrial energyuse and carbon emissions. At the same time, investmentincentives can noticeably speed up the adoption of e$-cient technologies to the bene"t of both industry andclimate. This conclusion indicates the importance of ad-dressing with climate change policies the mechanismsthat lead to adoption of new technologies, decisions thata!ect carbon sequestration, and changes in long-termindustry output.

A word of caution is warranted at this point. Contribu-tions to the atmospheric carbon budget continue to in-crease if industrial production rises at rates that exceedthe rates of technology improvement and carbon seques-tration. However, our results indicates that not imple-menting an investment-led climate policy would meanforegoing an opportunity to stimulate e$ciency improve-ments in the industry and to reduce net carbon emissionsper ton of product. For example, the results show thatan increase in the cost of carbon would leverage

268 M. Ruth et al. / Energy Policy 28 (2000) 259}270

learning-by-doing in the industry and speed up e$ciencyimprovements. In the absence of investment incentives,industrial responses occur most noticeably with a 3}4year time lag following implementation of climate changepolicy. With a series of `smart technology policiesa toencourage the investment in new technologies, industrialresponses can be even quicker and stronger. As notedearlier, the investment-led scenario achieved a 2020 busi-ness-as-usual energy intensity 17 years earlier. Thus, thesooner a policy is set, the faster can previously unrealizede$ciency potentials be explored. The longer it takes toformulate and implement a policy, the more uncertaintywill be generated in industry about its optimal policyresponses. Similarly, delaying policy means increasingthe likelihood that the industry will become locked in anultimately undesirable development trajectory which, inturn, will require stronger policy incentives to meet a goalof stable or declining carbon emissions.

Acknowledgements

This project was funded by the US EnvironmentalProtection Agency under Grant Number X 825828-01-0.The analysis and perspectives presented here do notnecessarily represent any o$cial views or policies of theUS Environmental Protection Agency.

References

AFPA, various years. Statistics of Paper Paperboard and Woodpulp.The American Forest and Paper Association, Washington, DC.

Akaike, H., 1973. Information theory and the extension of the max-imum likelihood principle. In: Petrov, B.N., et al. (Ed.), Proceedingsof the Second International symposium on Information Theory.AkadeHmi Kiado, Budapest, Hungary, pp. 267}281.

API, various years. Statistics of Paper Paperboard and Woodpulp. TheAmerican Paper Institute, Washington, DC.

Baumol, W., Oates, W., 1988. The Theory of Environmental Policy.Cambridge University Press, Cambridge, USA.

Breusch, T., Pagan, A., 1979. A simple test for heteroscedasticity andrandom coe$cient variation. Econometrica 47, 1287}1294.

Bureau of the Census, 1995. Census of Manufactures. GovernmentPrinting O$ce, Washington, DC.

DeCanio, S.J., 1993. Barriers within "rms to energy e$cient invest-ments. Energy Policy 21, 906}914.

DeCanio, S.J., 1994. Energy e$ciency and managerial performance:improving pro"tability while reducing greenhouse gas emissions. In:Feldman, D. (Ed.), Global Climate Change and Public Policy.Nelson-Hall Publishers, Chicago.

DeCanio, S.J., 1996. Understanding the di!usion process in modelingthe development of energy e$ciency technologies. Paper Preparedfor the Climate Change Analysis Workshop, Spring"eld VA, June6}7.

DeCanio, S., Laitner, J., 1997. Modeling technological change in energydemand forecasting: A generalized approach. Technological Fore-casting and Social Change 55, 249}263.

Dickey, D., Fuller, W., 1981. Likelihood ratio statistics for autoregres-sive time series with a unit root. Econometrica 49, 291}324.

EIA, 1994. Annual Energy Outlook. Department of Energy/EnergyInformation Administration, Washington, DC.

EIA, 1997. Annual Energy Review. Department of Energy/EnergyInformation Administration, Washington, DC.

EIA, 1997a. Annual Energy Outlook. Department of Energy/EnergyInformation Administration, Washington, DC.

EIA, 1998. Industrial sector demand module of the national energymodeling system. O$ce of Integrated Analysis and Forecasting,Department of Energy/Energy Information Administration,Washington, DC.

Ekins, P., 1996. How large a carbon tax is justi"ed by the secondarybene"ts of CO

2abatement?. Resource and Energy Economics 18,

161}187.Energetics, 1990. Industrial technologies: industry pro"les, Energetics

Inc. Columbia.Energy Innovations, 1997. Energy innovations: a prosperous path to

a clean environment. Alliance to Save Energy, American Council foran Energy-E$cient Economy, Natural Resources Defense Council,Tellus, Institute, and Union of Concerned Scientists, Washington, DC.

Fishbone, L.G., et al. 1983. UsermH s Guide for MARKAL: a MultiperiodLinear-Programming Model for Energy System Analysis. Brook-haven National Laboratory, Long Island USA and Kernforschun-gsanlage Julich, Germany.

Freeman Miller, (various years). The North American Pulp and PaperFactbook. San Francisco, CA.

Freeman Miller, 1996. Pulp and Paper. San Francisco, CA.Geller, H., Nadel, S., 1994. Market transformation strategies to pro-

mote end use e$ciency. Annual Review of Energy and Environment19, 301}346.

Gilbreath, K.R., 1995. draft. Life Cycle Assessment, Energy and theEnvironment from a Pulp and Paper Mill? Perspective, St. LaurentPaperboard, Inc., Montreal, Quebec, Canada.

Godfrey, L.G., 1978. Testing against general autoregressive and movingaverage error models when the regressors include lagged dependentvariables. Econometrica 46, 1293}1302.

Haynes, R., Adams, D., Mills, J., 1995. The 1993 timber assessmentupdate. US Department of Agriculture, Forest Service, RockyMountain Forest and Range Experiment Station, General Tech-nical Report RM-GTR-259, Fort Collins Colorado.

Hoeller, P., Wallin, M., 1991. Energy prices, taxes and carbon dioxideemissions. OECD Economic Studies 17, 91}105.

Ince, P., 1994, Recycling and long range timber outlook. US Depart-ment of Agriculture, Forest Service, Rocky Mountain Forest andRange Experiment Station, General Technical Report RM-242,Fort Collins CO.

Interlaboratory Working Group, 1997. Scenarios of US carbon reduc-tions: potential impacts of energy technologies by 2010 and beyond.O$ce of Energy E$ciency and Renewable Energy/US Departmentof Energy, Washington, DC.

IPCC, 1990. Intergovernmental Panel on Climate Change, ClimateChange: The IPCC Scienti"c Assessment. Cambridge UniversityPress, Cambridge.

IPCC, 1995, Intergovernmental Panel on Climate Change, ClimateChange: The IPCC Scienti"c Assessment. Cambridge UniversityPress, Cambridge.

Jaccard, M., Montgomery, W., 1996. Costs of reducing greenhousegas emissions in the USA and Canada. Energy Policy 24, 889}898.

Joskow, P., Marron, D., 1993. What does utility subsidized energye$ciency really cost? Science 260,281,370.

Koomey, J.G., Richey, R.C., Laitner, J.A., Sanstad, A.H., Markel, R.J.,Marney, C., 1998. Technology and greenhouse gas emissions: anintegrated scenario analysis using the LBNL-NEMS model. Inter-nal Review Draft, Energy Analysis Department. Lawrence BerkeleyNational Laboratory, Berkeley, California.

Laitner, J.A., Bernow, S., DeCicco, J., 1998. Employment and othermacroeconomic bene"ts of an innovation-led climate strategy forthe United States. Energy Policy 26, 425}432.

M. Ruth et al. / Energy Policy 28 (2000) 259}270 269

Laitner, J.A., Hogan, K., Hanson, D., 1999. Technology and green-house gas emissions: an integrated analysis of policies that increaseinvestments in cost-e!ective, energy-e$cient technologies. Paperpresented at the Electric Utilities Environment Conference, Tuscon,AZ, January 11.

Larson, E.D., Yang, W., Iisa, K., Malcolm, E.W., McDonald, G.W.,Frederick, W.J., Kreuz, T.G., Brown, C.A., 1998. A cost-bene"tassessment of black liquor gasi"cation/combined cycle technologyintegrated into a kraft pulp mill. Prepared for the TAPPI Interna-tional Chemical Recovery Conference, Tampa, FL.

Mann, M., Spath, P., 1997. Life cycle assessment of a biomass gasi"ca-tion combined-cycle power system. NREL/TP-430-23076. NationalRenewable Energy Laboratory, Golden, CO.

Meier, A., et al., 1983. Supplying energy through greater e$ciency: thepotential for conservation in CaliforniamH s residential sector. Univer-sity of California Press, Berkeley, CA.

Nordhaus, W.D., 1991. To slow or not to slow: the economics of thegreenhouse e!ect. The Economic Journal 101, 920}937.

Nordhaus, W.D., 1993. Rolling the DICE: an optimal transition pathfor controlling greenhouse gases. Resource and Energy Economics15, 27}50.

OECD, 1995. Environmental Taxes in OECD Countries. OECD, Paris.OTA, 1993. Industrial Energy E$ciency. Government Printing O$ce,

Washington, DC.Patterson, M., 1996. What is energy e$ciency?. Energy Policy 24 (5),

377}390.Pearce, D., 1991. The role of carbon taxes in adjusting to global

warming. The Economic Journal 101, 938}948.Pindyck, R., Rubinfeld, D.L., 1991. Economic Models and Forecasts.

McGraw-Hill, New York.Porter, M., 1990. The Competitive Advantage of Nations. The Free

Press, New York.

Ross, M., Thimmapuram, P., Fisher, R., Maciorowski, W., 1993. Long-term Industrial Energy Forecasting (LIEF) Model (18 Sector Ver-sion). Argonne National Laboratory, Argonne, IL.

Ruth, M., 1993. Integrating Economics, Ecology and Thermodynamics.Kluwer, Dortrecht, Netherlands.

Ruth, M., Hannon, B., 1997. Modeling Dynamic Economic Systems.Springer, New York.

Schwartz, G., 1978. Estimating the dimensions of a model. The Annalsof Statistics 6, 461}464.

Stoft, S., 1995. The economics of conserved-energy m̀supplym( curves. TheEnergy Journal 16 (4), 109}137.

Tietenberg, T.H., 1990. Economic Instruments for EnvironmentalRegulations. Oxford Review of Economic Policy 6, 17}33.

Tietenberg, T.H., 1996. Managing the transition to sustainable develop-ment: the role of economic incentives. In: May, P.H., daMotta, R.S.(Eds.), Pricing the Planet: Economic Analysis for Sustainable Devel-opment. Columbia University Press, New York, pp. 123}138.

Train, K. et al., 1995. Rebates, loans, and customer's choice of appliancee$ciency level: combining state-and revealed preference data. En-ergy Journal 16, 55}69.

Woodru!, M. et al., 1997. The potential for industrial energy e$ciencytechnology to reduce US greenhouse gas emissions. in: 1997ACEEE Summer Study on Energy E$ciency in Industry. Proceed-ings, ACEEE, Washington.

Worrel, E., 1994. Potentials of improved use of industrial energy andmaterials. Ph.D. Dissertation. Utrecht University, The Netherlands.

Yelle, L.E., 1979. The learning curve: historical survey and comprehens-ive survey. Decision Sciences 10, 302}334.

Zellner, A., 1962. An e$cient method of estimating seemingly unrelatedregressions and tests for aggregation bias. Journal of AmericanStatistical Association 57, 348}368.

270 M. Ruth et al. / Energy Policy 28 (2000) 259}270