The Pulp and Paper Industrial Sector

The U.S. pulp and paper industry is a diverse sector that utilizes a variety of processes, and manufactures hundreds of different grades of paper [4]. The industry is grouped under paper manufacturing in the North American Industry Classification System (NAICS) as code 322 [5] and includes pulp, paper, and paperboard mills (NAICS code 3221) and converted paper product manufacturing (NAICS code 3222). Facilities in the 3221 NAICS code classification are primarily engaged in producing pulp and/or paper and paperboard. A facility producing both pulp and paper, with paper as the primary product, is considered an integrated facility. Processes and emissions from these different types of facilities vary greatly, and in order to accurately represent the industrial sector, the model will need to account for them.

Integrated facilities generally produce paper in six main processing steps: wood preparation, pulping, pulp washing, pulp screening, bleaching (optional, depends on final product type) and paper making [6]. Facilities utilizing the chemical pulping process use chemical recovery to recover pulping chemicals and generate process steam. The wood preparation process involves wood cutting, transporting, debarking, chipping, and screening of the raw material. The main objective of this process is to produce uniformly sized chips to be used in the pulping process, a process in which wood is broken down into fibers that can be used for papermaking. There are three main types of pulping: mechanical, chemical, and semichemical. In the case of chemical pulping, cellulosic fibers are separated from the cellulose-hemicellulose-lignin matrix in wood using high temperature, pressure, and chemicals. In the case of mechanical pulping, logs or chips are mechanically broken into smaller pieces. Semichemical pulping uses a combination of both methods. From the pulping process, pulp and spent cooking liquor proceed to pulp washing. The pulp washing process is used to remove cooking chemicals and the dissolved wood components in the cooking liquor for recovery and for energy generation, respectively. The recovery of these materials also minimizes the addition of chemicals and solids to the effluent treatment plant. The pulp screening process separates cooked pulp fibers from uncooked fiber bundles and knots. The pulp proceeds from screening to bleaching if the final product requires bleached pulp, or to the papermaking process if the final product utilizes unbleached pulp. The bleaching process involves removing the lignin that still remains after cooking (chemically whitening) or breaking double bonds in the lignin without removing it (brightening), as the lignin contains the chromophoric groups which make the pulp dark. Bleached pulp proceeds to the papermaking process.

The papermaking process involves stock preparation, dewatering, pressing, drying and finishing [6]. First, pulp fibers are treated mechanically by refining to produce flexible fibers suitable for papermaking. The treated fibers are blended with product specific additives (e.g., fillers for printing grades, or wet strength agents for tissue) and are diluted significantly with water to create the papermaking slurry. This slurry is processed on a paper machine, which creates a fiber mat and removes water by gravity, suction, pressure, and heat. The rolls of paper produced on the paper machine can be converted to final products on site, or may be shipped to another location for conversion.

The chemical recovery process involves evaporation, combustion, causticizing, and calcining [6]. The weak black liquor from pulp washing is processed using a multiple-effect evaporator system to increase the black liquor solids content by removing water. This is done to improve the heating value of the liquor, because it will be burned in a recovery furnace to generate steam. The purpose of the recovery furnace is to burn the organics in the black liquor and recover the inorganics in molten form. These inorganics (known as smelt) are dissolved to create green liquor. Green liquor is then clarified and causticized using lime to create white liquor for the pulping process. Lime mud is collected from the white liquor clarifier and burned in a lime kiln to regenerate lime for the caustization process. All of the process steps discussed in this section are potential air emission points that are considered in regulatory review and development.

Current Economic Status of the Sector

Pulp and paper production has been prevalent in the U.S. since the early 1900s, and continues to contribute significantly to the U.S. economy. Market research indicates that the pulp, paper, and paperboard industry produced $80 billion in revenue in the U.S. in 2011 [7]. Historical production data were obtained from the Food and Agriculture Organization of the United Nations FAOSTAT database for “pulp for paper” and “paper and paperboard” [8]. In 2011, the most recent year with data available, the five highest world producers of pulp for paper were the U.S. (50.2M tonnes), China (21.1M), Canada (18.3M), Brazil (13.9M), and Sweden (11.7M). In 2011, the five highest world producers of paper and paperboard were China (103.1M tonnes), the U.S. (77.4M), Japan (26.2M), Germany (22.7M), and Canada (12.1M). Historically, the U.S. and China have led pulp and paper production. China production of paper began to significantly increase in 2002 and surpassed U.S. production in 2008, as shown in Fig. 1a. Pulp production did not follow this trend, as the U.S. maintained a significantly higher pulp production for all of the years analyzed (1980–2011), as shown in Fig. 1b. As expected with China’s large increase in paper production but minimal increase in pulp production, China’s pulp imports significantly rose from 2002 to 2011 (Fig. 1c). The U.S.’s exports of pulp also increased during this time (Fig. 1d) [8]. This interaction is important for understanding environmental impacts of pulp production.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Historical pulp and paper trends for the U.S. and China.

https://doi.org/10.1371/journal.pone.0120954.g001

Environmental Regulations

Air emissions from the U.S. pulp and paper industry have been federally regulated since the 1978 Kraft Pulp Mill New Source Performance Standards (NSPS) [9]. This regulation controlled the total reduced sulfur (TRS) emissions and particulate matter (PM) emissions from pulp mill and chemical recovery emission units. The National Emission Standards for Hazardous Air Pollutants (NESHAP) for Pulping and Papermaking [10] was promulgated in 1997 and controlled hazardous air pollutant (HAP) emissions from pulping and papermaking sources. Finally, the NESHAP for Combustion Sources [11] was promulgated in 2002 and controlled HAP emissions from the combustion sources at pulp mills.

The Clean Air Act (CAA) requires EPA to review, and if appropriate, revise the NSPS and the NESHAP every eight years. With the changing market and the potential to reduce emissions when reviewing the pulp and paper sector regulations, it would be beneficial to have a model that can analyze the impacts of potentially more stringent emission limits and new emission standards for pollutants not previously regulated on individual mills as well as the U.S. and the global market. This model would assist EPA in determining the impacts of revised federal regulations, as well as obtaining an estimate of how mills might comply with the new standards (e.g., reduced production, process changes, add-on controls, fuel switching, equipment upgrades, or mill closure).

Industrial Sector Integrated Solutions Model

The ISIS model was initially used to analyze multi-pollutant regulatory options for the cement industry [3]. Subsequently, EPA’s Office of Research and Development (ORD) and Office of Air Quality Planning and Standards (OAQPS) expanded the ISIS modeling framework to include the pulp and paper sector. The ISIS pulp and paper framework identifies pulp mills and paper mills as two independent units. As a result, the ISIS input database is divided into two major sections: pulp data and paper data. This renders the flexibility to independently optimize the pulp mill from the paper mill operation at an integrated facility, and vice versa. All mills with available data that have operated for at least one year since 2000 are represented in the database. Contrary to the cement model, the pulp and paper model does not project new mills or options for the addition of new production capacity by region, because no new integrated mills have been built since 1990, and the current trend is to reopen closed mills. Each facility modeled in ISIS is characterized by its location, pulping process, equipment utilized, annual product capacities, cost components, and retirement information when available. This information is discussed further in the following sections.

The current version of the ISIS pulp and paper model focuses on the combustion sources located at pulp and papermaking facilities; these sources are recovery furnaces, lime kilns, and boilers. Recovery furnaces are used to recover valuable pulping chemicals for re-use in the process and to generate process steam. Similarly, boilers are used to generate steam and electricity for the facility and its processes. Lime kilns are used to convert the recovered pulping chemicals into fresh pulping chemicals. It is necessary to relate control costs and potential emission reductions of these units to production of a given commodity. For pulp and paper, this is not a direct relationship because these sources do not produce the final product, but rather perform a supporting function of energy production (boilers) and chemical recovery (recovery furnaces and lime kilns). Recovery furnaces and lime kilns in particular are associated with the pulping process rather than papermaking. This article will discuss the data collection and database development needed to link emissions from boilers, recovery furnaces, and lime kilns to pulp and paper production.

Finished Product Data

The ISIS pulp and paper modeling efforts were focused on representing the entire population of U.S. integrated and non-integrated pulp and paper mills and their products. Subcategories of paper products provided by RISI were grouped into eight major categories, primarily classified on the basis of product end use. Similarly, there were two major categories of pulp products (softwood and hardwood pulp). Table 1 lists the major categories and subcategories of the pulp and paper products utilized in the model.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. ISIS product categories for the U.S. pulp and paper market.

https://doi.org/10.1371/journal.pone.0120954.t001

Mill Level Process Characterization

The number of mills in the ISIS database and their locations were determined using the North American Graphic Paper Capacity Report, the North American Paper Packaging Capacity Report, and the World Tissue Capacity Report purchased from RISI. The mill types (integrated vs. non-integrated) were determined using the ICR and CPBIS’s “Mills Online” database and the Mill Asset Database. For mills that are currently closed, a retirement date was determined using the Pulp & Paper Resource Council’s website and the capacity report (i.e., facilities with “0” capacity were assigned the first year of “0” capacity as their retirement year). Facilities in the ISIS database were identified based on a plant type (“PR” for paper mill or “PL” for pulp mill) and an ISIS ID number. These values were concatenated to form mill identification numbers and were then used to identify the facilities in the model.

The United States Mill Asset Database, a collection of process flow diagrams for 232 facilities, was used to extract process and production information. The data extracted included pulp production, product composition (i.e., grade recipes), facility type (i.e., whether or not the facility could purchase market pulp and/or recycled pulp), number of boilers and recovery furnaces, fuels utilized, steam and electricity generation, and electricity utilization. These data were extrapolated to the remaining 282 facilities that did not have flow diagrams available based on products produced at the facility, mill type, and region. Production process information for facilities without process flow diagrams (e.g., the ability to process recycled fiber) was determined using company websites.

The final product-specific capacity for each facility was determined using the maximum production value for the 10-year period for which data was purchased. It was assumed that a facility was operating at 85-percent capacity in the year with the highest production (based on utilization rates provided in the ICR). This value represents the maximum amount of each product a facility can make without making a process change or upgrading equipment. Capacity for a facility was reported as “0” for years which production values were “0,” as these periods represent periods after a mill was closed, stopped producing a particular product, or before obtaining equipment to produce the specific product. An example of the calculations are shown in Table 2.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Production capacity example.

https://doi.org/10.1371/journal.pone.0120954.t002

For integrated facilities, pulp production and pulp capacity were calculated using the process flow diagrams for those with available data. For pulp production, pre-digester and post-digester pulp values were extracted and used to calculate pulping yield. For example, a pre-digester value of 2400 bone dry short tons per day (BDST/D) of softwood chips and a post-digester value of 1170 BDST/D of softwood pulp indicate a pulping yield of 48.75-percent. This pulping yield was then used in conjunction with the values in the product summary table on the process flow diagrams for the amount of wood used for each product to determine the amount of pulp needed for each product. This was repeated for all products produced at a mill, and then summed to obtain the final pulp production value. Final pulp production values were averaged based on the major product category produced, and tons of pulp needed per ton of product values were assigned to similar facilities where process flow diagrams were unavailable. Pulp mill capacity was determined using the maximum production year and a utilization factor of 85-percent, the same method used to calculate final product capacity in Table 2.

Product composition was determined using the product summary table on the process flow diagram for facilities where flow diagrams were available. These values represented the amount of softwood pulp, hardwood pulp, additives, and recycled pulp in the final product. In the case of integrated facilities, the pulping yield, as calculated for the pulp mill production and capacity values, was also used to determine product composition. For non-integrated facilities, values for pulps and recycled papers from the flow diagram summary table were used to calculate product composition. All recycled paper values (e.g., deinked pulp, pulp substitutes, old corrugated containers, old newsprint, and mixed papers) were summed to create one “recycled pulp” value for each product. Market pulp values (e.g., northern bleached softwood kraft pulp and northern bleached hardwood kraft pulp) were maintained as separate values to determine softwood and hardwood percentages. Average product compositions were determined based on final product category and mill classification (e.g., non-integrated vs. integrated, recycle mill vs. virgin mill) and assigned to facilities without flow diagrams.

It was confirmed that the paper grade recipes matched the classification of the facility. For example, a facility producing boxboard from 100 percent recycled material was confirmed to be represented by a “1” in the “facility can purchase recycled pulp” column. It was also confirmed, for example, that a facility producing tissue from 50-percent hardwood pulp and 50-percent softwood pulp could produce the pulp (indicated by a “Y” in the “integrated facility” column) or could purchase the pulp (indicated by a “1” in the “can purchase market pulp” column) for use in the final product. In cases where these conditions were not met, product recipes were adjusted accordingly; that is, if a product composition was assumed for a facility and that facility could not produce or purchase virgin pulp, but could purchase recycled pulp, the product composition was changed to 100-percent recycled. Some facilities produce both products from recycled pulp and products from virgin pulp. In these cases, all products were compared to the mill classification to confirm that the facility was capable of making all of the products based on their characterization. It was also confirmed that all pulp mills had corresponding pulp production data. For example, an integrated facility producing hardwood pulp and making containerboard from hardwood and softwood pulp was confirmed to be represented by “Y” in the “integrated facility” column and a “1” in the “facility can purchase pulp” column. For the integrated mills, it was confirmed that hardwood pulp production and/or softwood pulp production values were available. If all years were reported as zero for both pulp types, the pulp mill classification was removed and the facility was classified as a stand-alone paper mill. An example of these classifications can be seen in Table 3, for all three types of facilities (i.e., stand-alone paper [PR540], stand-alone pulp [PL812], and integrated [PR291/PL291]).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Mill classification example.

https://doi.org/10.1371/journal.pone.0120954.t003

Pulp mill type was determined for the 163 pulp mills in the database. This information dictated the type of equipment utilized at the facility, as well as the type and quantity of emissions. Pulp mills were classified as mechanical or chemical based on the process flow diagrams, ICR information, and company websites. Chemical pulp mills were further classified as sulfate (kraft pulping), the predominant chemical pulping process utilized in the U.S., or other pulping process.

Emission unit data were determined using the ICR, the Boiler Maximum Achievable Control Technology (MACT) database [28], and the process flow diagrams. For integrated facilities, boilers were assigned to paper mills, while lime kilns and recovery furnaces were assigned to pulp mills. The average number of boilers, lime kilns, and recovery furnaces (three, one, and two, respectively) was determined for integrated facilities with data and were assigned to those without data. The average number of boilers (two) for stand-alone paper mills was assigned to facilities without flow diagrams. Lime kilns and recovery furnaces were not assigned to facilities using chemical pulping methods other than sulfate (kraft), unless ICR data showed that the facility had one or both.

Existing controls for the assigned equipment were determined based on the ICR and the Boiler MACT database for those with data available. Boiler controls were assigned based on fuel type and products produced at similar facilities with data. Lime kiln controls were assumed to be scrubbers for all kilns without data, and recovery furnace controls were assumed to be dry-bottom electrostatic precipitators, both based on the representative control for the majority of emission units with data.

Lime mud production and black liquor production were calculated using information collected in the ICR. Values in the ICR for lime mud production were based on a ton of calcium oxide produced per day basis, and a recovery rate of 90-percent was assumed to calculate the amount of lime mud burned per day. This value was then combined with the previously calculated daily pulp production for 2009, resulting in a tons of lime mud produced per ton of pulp produced (TLM/TP) value. An average value of 0.270 TLM/TP was assigned to mills without ICR data (this value was the average of all available data). Black liquor values in the ICR were based on million pounds of solids per day. This value was combined with the previously calculated daily pulp production for 2009, resulting in tons of black liquor solids produced per ton of pulp produced (TBLS/TP). An average value of 1.49 TBLS/TP value was assigned to mills without ICR data (this value was the average of all available data).

Landfill data were provided in the summary table of the process flow diagrams in the form of ton of waste per ton of final product. Waste materials generated at the facility and disposed of in a landfill for this database are typically sludges generated from paper recycling and wastewater treatment. For facilities without a process flow diagram, a landfill value was assigned based on mills with data of similar mill type and final product type. This data is needed to estimate the cost of disposing of waste materials, and could potentially be used to estimate emissions from the landfill.

Electricity data were extracted from the boiler section of the process flow diagrams for the available facilities. In this section, the number of power boilers was indicated for each plant, as well as the amount of electricity generated, used, purchased, and sold. These values were combined with the daily production values to determine the electricity values (i.e., generated, used, purchased, and sold) per ton of finished product. Facilities with electricity values were averaged based on all of the products produced at each facility, as well as the facility type and then applied to similar facilities producing similar products without process flow diagrams.

Boiler data was extracted from the boiler section of the flow diagrams for available facilities. The number of power boilers and the fuel used for each facility was extracted. Fuel intensities, amounts of each fuel used per ton of production, and electricity usage were also extracted from the flow diagram summary table. Some boilers had input fuels that were not assigned to product intensities. These fuel intensities were not populated; however, they were included as fuels available for use by the facility. Fuel intensities were averaged per product and assigned to facilities without flow diagrams. Controls and boiler fuel types were assigned based on the most frequently used for each product in the database. Boiler fuel data as well as the number of boilers were also extracted from the Boiler MACT database and compared to the flow diagram results. These additional data assisted in populating values for facilities without flow diagrams. Facility fuel availability was assigned to each facility based on fuel intensities from the flow diagram summary table and boiler input fuels (flow diagram and Boiler MACT). These values were assigned to facilities without process flow diagrams based on final products produced and facility type.

Cost Data

Cost data were obtained from RISI or calculated for six primary cost functions in the model. The primary cost functions were raw material cost, maintenance and repair cost, labor cost, fuel cost, electricity cost, and solid waste disposal cost. The development of these functions is discussed in this section.

Demand Modeling Data

Transportation and Interregional Trade.

In the ISIS pulp and paper model, a domestic transportation matrix was used to describe the costs for transporting pulp from pulp mills to paper mills, and from paper mills to demand centers (the U.S. market for paper products). Figs. 2 and 3 illustrate the ISIS network of domestic transport of pulp and paper in three regions within the United States. Fig. 2 describes the domestic trade of softwood and hardwood pulps from pulp mills to paper mills located in three designated ISIS regions (north, south, and west). Similarly, Fig. 3 shows the transport of the eight major paper products from paper mills to demand centers or U.S. market.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 2. Domestic transport of pulp from pulp mills to paper mills.

https://doi.org/10.1371/journal.pone.0120954.g002

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 3. Domestic transport of paper products from paper mills to demand centers.

https://doi.org/10.1371/journal.pone.0120954.g003

Figs. 2 and 3 demonstrate that there is no constraint in the supply route. In other words, the ISIS database allows all regional pulp and paper mills to supply all regional paper mills or demand centers, respectively. This signifies the flexibility of the ISIS model to select the most effective transportation route (based on cost) to obtain the optimal solution.

Transportation costs were calculated by adapting the dollar value published in the USDA “Recycling and Long-Range Timber Outlook” report [33]. However, transportation costs in the USDA report were reported as 1986 dollar value. Inflation was taken into account and the real 1986 dollar values were adjusted by the consumer price index published by the Bureau of Labor Statistics [34] to obtain real 2010 dollar values to represent the 2010 inter-regional transportation in the ISIS database.

International Trade Data

The U.S. pulp and paper markets trade pulp and paper products with a number of countries; however, the U.S. trades significant quantities of pulp and paper products with Canada. Considering Canada’s role in trade with the United States, international trade in pulp and paper products was simplified to represent trade between Canada, the rest of the world (ROW), and the U.S. These traded products arrive and leave from three geographical domestic regions: north, south and west, as illustrated in Fig. 4. In Fig. 4, pulp is imported from the Canada and/or ROW market to pulp mills via import districts and then distributed to paper mills. Pulp is exported to Canada and/or ROW through export districts. Fig. 4 also demonstrates that paper is imported from Canada and/or ROW through import districts to regional demand centers. Paper is exported from the paper mills to Canada and/or ROW through the export districts. The average quantities of traded pulp and paper products were determined based on the data reported by RISI Inc., for a period of 11 years (2000 to 2010). Transportation costs between regions and mills and/or demand centers were also reported in the database.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 4. Trade network of pulp and paper products from Canada and/or Rest of the World to pulp/paper mills and demand centers respectively.

https://doi.org/10.1371/journal.pone.0120954.g004

Emissions and Controls Data

The design of the ISIS pulp and paper model can accommodate any number of pollutants of interest. In the model, each boiler, recovery furnace and lime kiln can be characterized by its nitrogen oxides (NO_X), sulfur dioxide (SO₂), hydrochloric acid (HCl), mercury (Hg), PM, volatile organic compound (VOC), total hydrocarbon (THC), total HAP, and carbon dioxide (CO₂) emissions. The first phase of the model includes the following pollutants: NOx, SO₂, and PM. The second phase will consider HCl, VOC, THC, Hg, and total HAP.

In the ISIS pulp and paper model, pollution control technologies were related to boiler, recovery furnace, and lime kiln heat inputs and/or furnace gas flow rate. The NO_X emissions from the pulp and paper industry result primarily from the boilers and the recovery furnace, and reduction methods included in the model were low NOx burners (LNB), ultra-low NOx burners, over fire air (OFA), flue gas recirculation (FGR), selective non-catalytic reduction (SNCR), selective catalytic reduction (SCR), and low temperature oxidation [35].

Sulfur dioxide emissions from the recovery furnace are the product of sulfur in the smelt, and SO₂ emissions from power boilers, especially those firing coal or residual fuel oil, are directly related to the sulfur content of the fuel. Wet and dry scrubbing technologies may be applied in the model to control SO₂ emissions.

The ISIS pulp and paper model database contains information on abatement approaches for NO_X, SO₂, and PM emissions. The four categories of abatement approaches included are: add-on controls, combustion modifications and upgrades, fuel switching, and mitigation boiler/furnace technologies. Add-on control abatement approaches include flue gas desulfurization, spray dryer absorber, and in-duct dry sorbent injection methods for SO₂ control, and SCR and SNCR methods for NO_X control in power boilers. Flue gas desulfurization abatement method in recovery furnace is used for SO₂, and SCR and SNCR abatement methods in recovery furnace are used for NO_X. Combustion modification used in power boiler are LNB, FGR, OFA, oxygen trim and water injection, and upgrade combustion controls for NO_X. For each emission abatement approach, where possible, information on the following parameters were developed [35] and included in the database: capital cost, fixed operating cost, variable operating cost, emission reduction performance for all of the pollutants, impacts on fuel and/or raw material use, impact on electricity consumption, byproduct generation and cost, and impact on water use.

To estimate capital recovery factors for capital costs associated with control technologies, economic life values and interest rate were considered. Economic life for each of these measures can be taken to be the average of the technical life and the payback period. Again, an interest rate of seven percent can be used for capital recovery in the absence of more specific information. End users can select different values for economic life and interest rate as appropriate.

Policy and Economic Parameters

The ISIS model can be used to analyze the impact of new or modified air regulations on individual facilities and on the industrial sector. Traditional policy scenarios in the model can include unit-specific emission limits based on production rates or concentrations. For example, a revision to the PM emission limit for recovery furnaces could be imposed in the model to determine several parameters such as the number of units that could not meet the new standard, how units could comply (e.g., process changes, add-on controls, reduced production), impacts to the general population such as product price increases, and impacts to the global market. The model can also analyze cap-and trade policy scenarios, if needed. Currently in the ISIS database, the following economic parameters are defined: discount rate, future escalation rates of various energy, material, transportation, and labor inputs, demand elasticities for each major category of products, and ten years of demand projections (tons/yr) for each major category of products.

The demand projections for each major product are key components in the model. This is because the demand for products drive production levels at mills in the model, and changes in costs resulting from regulatory scenarios adjust demand levels via a price mechanism. Estimating future demand of pulp and paper products required a two-step statistical modeling process. First, supply and demand systems were estimated using the historical trends in the product data compiled for the ISIS model. These models permitted an examination of the sensitivity (or “elasticity”) of product demand to various potential explanatory variables. These models served as the bases for the second step of the future demand modeling in which the explanatory variables where forecast into the future and, along with the parameter estimated for the demand equation, a baseline forecast of each product type was discerned. In the model, as a result, product demand varies from the baseline forecast as consumers potentially face different market prices for products as mills adjust to the simulated regulatory impacts.

In estimating supply and demand equations, the convention in economic statistics that supply and demand systems are simultaneously determined was followed. In other words, the supply and demand equations for each given product must share the same market clearing price, therefore, a change in quantity supplied simultaneously affects the market price and, as a result, causes a change in demand. This process is non-recursive, meaning that feedback travels in both directions, and presents statistical challenges in that the supply and demand equations are correlated through their error terms. The simultaneity of the supply and demand equations also introduces an additional “identification’ challenge, which is the inability to identify individual supply and demand curves, as the observed data used in the models only depict market clearing prices under the finite set of observed conditions, not the range of possible conditions which may be of interest from a regulatory or policy perspective.

To resolve these challenges, a two-stage least squares (TSLS) approach using instrumental variables was utilized. The TSLS approach helps to avoid the biased estimation of the supply and demand model parameters by isolating the exogenous behavior of price in the pair of equations for each pulp and paper product. The statistical strategy additionally leverages an instrumental variable (IV), which is an exogenous variable that is correlated with price but not the error terms of either model. The IV or “shifter” affects demand conditions without affecting cost conditions or could affect cost conditions without affecting demand conditions. These “shifters” are used to trace out or identify the supply or demand curve.

After estimating the supply and demand models, these models were used to develop baseline forecasts for pulp and paper products. To do so, however, required forecasts of the explanatory variables in the demand models. For some variables, the data sources from which the explanatory variables were drawn already had forecasts, such as those available from macroeconomic models, such as Gross Domestic Product. For other variables, a technique from time series analysis called autoregressive integrated moving average (ARIMA) modeling was utilized. The ARIMA models rely upon trends in the variable of interest alone to obtain forecasts, as opposed to attempting to explain influences upon the variable and estimating important elasticities as done in the supply and demand modeling.