For compiling a comprehensive database on the long-term development of material and energy flows in the U.S. economy,1 standard methods of economy-wide material flow accounting were applied (Eurostat 2009). This study quantified used extraction, imports, and exports of materials and used data from a range of national and international statistical sources. Estimation procedures were applied for materials that were not reported in statistical sources. The MFA database follows the materials classification used by Eurostat (2009) and distinguishes between 50 and 60 material groups for (used) domestic extraction, imports, and exports. Unused extraction or indirect flows (upstream material flows of traded products) were not considered. In this article, data and indicators are presented for four main material groups: biomass, fossil energy carriers, ores, and nonmetallic minerals. Furthermore, a distinction between renewable (biomass) and nonrenewable (mineral and fossil) materials is made.
We calculated the following material flow indicators (see Fischer-Kowalski et al. 2011): domestic extraction (DE); imports and exports; domestic material consumption (DMC), which is defined as DE plus imports minus exports; physical trade balance (PTB), which is defined as imports minus exports; and material intensity (MI), which is defined as DMC per unit of GDP and is the inverse of material productivity. We also refer to per capita values of domestic material and energy consumption as metabolic rates.
We based our long-term historical reconstruction of material flows in the United States on three major sources. First, we used the Historical Statistics of the United States (HSUS; U.S. Bureau of the Census 1975),2 which provides a comprehensive collection of statistical time series data covering a wide range of socioeconomic variables, including resource extraction from colonial times to present. Second, in the 1960s and 1970s Resources for the Future3 commissioned a number of studies on U.S. historical resource trends. These yielded two comprehensive data compilations—Potter and Christy (1962) and Manthy (1978)—that provide annual data on production of and trade for a large number of raw materials and commodities since the late-nineteenth century. Third, the U.S. Geological Survey (USGS) maintains a material flow database and provides time series data on most mineral materials from 1900 to recent years (Kelly and Matos 2008). In addition to these national sources, we used a range of international databases that provide data for more recent decades. As a general rule, international data were cross-checked with national sources for selected overlapping years in order to ensure consistency between sources and to avoid statistical breaks. Cross-checks have shown that while the pre-1960 sources generally cover a smaller number of materials, the difference in overall mass flows covered is negligible.
The extraction of used crop residues was estimated on the basis of time-dependent harvest factors and recovery rates for major crops. These were based on information derived from Wirsenius (2000) and Cunfer and Krausmann (2009). Grazed biomass was estimated on the basis of a simplified feed balance (Krausmann et al. 2008a). Data on livestock, livestock production, and feed availability were taken from the HSUS (1975) and FAO (2009). Feed demand (kilograms [kg]4 dry matter per head per day) for cattle was calculated on the basis of ruminant production (milk, meat) and assumptions on changes in livestock productivity were derived from Krausmann and colleagues (2008a) and Cunfer and Krausmann (2009). Feed demand factors for all other grazers were kept constant over time. In order to determine the amount of grazed biomass, available fodder crops, market feed, and crop residues were subtracted from the calculated total dry matter feed demand. For early years, no information on market feed was available. Based on information available for the 1960s (FAO 2009), we assumed a market feed share of 15% of total dry matter feed demand. Data on fish catch (excluding aquaculture, which is considered an internal transfer in MFA) were taken from the HSUS (1975), Manthy (1978), and FAO (2006).
Data on the extraction of industrial roundwood were derived from the HSUS for years prior to 1960 and from FAO 2009 for the period 1961–2005. For the years 1870–1900, the HSUS reports lumber and pulpwood only. The inclusion of other wood items in statistical reporting after 1900 results in a statistical break; therefore, numbers for industrial roundwood production increase by 25% after 1899. Fuel wood data are available from energy statistics (Schurr and Netschert 1960; IEA 2007) and production statistics (Manthy 1978; Howard 2007). While both sources show a very similar development of fuel wood use over time, the values derived from energy statistics are up to 40% higher than those from production statistic. To avoid double counting due to reuse of wood, we followed a conservative approach and used the lower values from the production statistics.
Data on the extraction of fossil energy carriers are well covered in statistical sources. For the period 1870–1960, data on the extraction of coal, oil, and natural gas were taken from the HSUS (1975). For the more recent years (1961–2005), data are from IEA (2007). Data on peat extraction were taken from Kelly and Matos (2008).
Data on the extraction of ores are very well documented by the USGS (Kelly and Matos 2008) from 1900 on. Data for the years prior to 1900 are from the HSUS and were complemented with data from Manthy (1978). Only for iron and bauxite do statistical sources report production in terms of gross ore. All other metals are reported in terms of metal content. To arrive at gross ore production, as required in MFA, we used information on coupled production and metal content, derived from the USGS (e.g., U.S. Bureau of Mines 1987). Gross ore values were calculated only for the main metals mined through coupled production to avoid double counting. For copper, which is the largest mass flow of non-iron ores, we assumed that ore grades declined from 2.5% in 1880 to 0.5% in 1975 based on ore grades given in Ayres and colleagues (2004). For all other metals, current ore grades were used, which results in an overestimation of gross ores in earlier periods when grades of some domestic ores were higher (cf. Mudd 2009; West 2011).
We used data on the extraction of nonmetallic minerals from the HSUS (1975) and Kelly and Matos (2008). Data on the extraction of natural aggregates (sand, gravel, crushed stone) reported in the HSUS are not consistent with the much lower numbers provided in Kelly and Matos (2008). Therefore we estimated the demand for sand, gravel, and crushed stone based on the production and use of cement, concrete, and asphalt (see Eurostat 2009; Krausmann et al. 2009; Schandl and West 2010). We applied standard coefficients on the ratio of sand and gravel to cement and bitumen in concrete and asphalt in order to extrapolate natural aggregates use. Data on the production and consumption of cement and bitumen were taken from Abraham (1945), the HSUS (1975), IEA (2007), and Kelly and Matos (2008). The results of this estimate match well with data reported in the HSUS (1975) and those estimated by Matos (2009); trends over time are similar in both estimates (see supporting information available on the Journal's Web site for details).
Comprehensive trade data are difficult to obtain. Most sources only cover trade with raw materials and semimanufactured products, but mass flows of manufactured products are rare. Data on the trade of biomass, fossil energy carriers, and products thereof were derived from Potter and Christy (1962) and Manthy (1978) for the years 1870–1960. From 1961 on we used FAO (2009) data for trade of agricultural and forestry products and IEA (2007) data for trade of fossil energy carriers and petrochemical products. For wood, data on net trade were only available for the years 1870–1950. Trade data on nonmetallic minerals and ores and semimanufactured metal products were taken from Potter and Christy (1962) for the period 1870–1899 and from Kelly and Matos (2008) for the period 1900–2005.
We used data from the United Nations (UN) Comtrade database (United Nations Statistical Division 2008) to cross-check our results and to quantify underestimations due to incomplete coverage of trade of manufactured products like furniture, textiles, machinery, and vehicles. Comtrade data are available from 1962 to the present, but for most years data on physical trade flows are fragmentary. Comprehensive data on imports and exports are available only for the years 1978, 1985–1988, and 2005. An analysis of Comtrade data for the years 1978 and 2005 revealed that manufactured products not considered in our accounting amount to 10% to 20% of total imports and 4% to 11% of exports (see the Supporting Information on the Web for details). In terms of net trade (physical imports minus physical exports), the underestimation is much smaller. Underestimation is highest for imports and exports of metal products (vehicles, machinery), petrochemical products (organic chemicals, plastics) and so-called other products, which are not assigned to a specific material group.
Data on material flows were used to calculate the total primary energy supply (TPES) for the entire time period. We converted flows of fuel wood and fossil energy carriers into energy units using material-specific gross calorific values. Energy flow data derived from MFA were supplemented with energy inputs from hydropower, nuclear heat, and geothermal sources. Data on electricity output from hydro and nuclear power plants (U.S. Bureau of the Census 1975 and IEA 2007) were converted into primary energy input by applying coefficients derived from information on average conversion efficiencies (Warr et al. 2010) (see the supporting information on the Web for details).