Pollution Abatement Activities and Traditional Productivity Growth

Appendix to

Pollution Abatement Activities and Traditional Productivity Growth

in Germany, Japan, the Netherlands and the United States

(Not for publication)

OECD STAN Database for Industrial Analysis

The OECD STAN Database is the source of most of the value added and capital expenditure data used in this study. These data, which were labeled as Vol. 2005 - edition 03 (Released on May 24, 2005), were downloaded at the Library of Congress on June 9, 2005 (we will refer to these data as the 2005 STAN Database). Observations are available for 1970 through 2003. We supplement the 2005 STAN Database with observations downloaded at the Library of Congress on October 4, 2004 (referred to as the 2004 STAN Database in this Appendix). The URL for STAN information is http://www1.oecd.org/dsti/sti/stat-ana/stats/ . The following variables in the STAN database are used in this study:

Value added at current prices: VALU

Value added volumes (quantity index): VALUK

Gross fixed capital formation at current prices: GFCF

Gross fixed capital formation, volumes (quantity index): GFCFK

Multiplying the volumes (VALUK and GCFCK) by the reference year (volume index = 100 in 1995) values for the current price variable (VALU and GFCF) yields volumes expressed in national currencies (i.e., real value added and real gross fixed capital formation) (see OECD 2005, p. 14):

VALUR = VALUK × VALU1995 / 100

GFCFR = GFCFK × GFCF1995 / 100

Implicit value added deflators and implicit gross fixed capital formation deflators can also be derived from the 2005 STAN Database (see OECD 2005, p. 14):

π = (VALU × 100) / (VALUK × VALU1995)

π = (GFCF × 100) / (GFCFK × GFCF1995)

When missing observations exist, the 2004 STAN Database and 2005 STAN Database are supplemented with data from the OECD STAN Database for Industrial Analysis, 1974- 1993 (referred to as the 1974-1993 STAN Database in this Appendix) and OECD STAN Database for Industrial Analysis, 1978-1997 (referred to as the 1978-1997 STAN Database in this Appendix). The values in these two publications differ from those in the 2004 STAN Database and 2005 STAN Database. In addition, the 2004 STAN Database and 2005 STAN Database use ISIC (Rev. 3) classification codes, while the 1974-1993 STAN Database and 1978-1997 STAN Database use ISIC (Rev. 2) classification codes.

“Total annual hours worked,” which is our measure of labor, is from the “Groningen Growth and Development Centre (GGDC), 60-Industry Database, September 2006, http://www.ggdc.net” (see O'Mahony and van Ark, 2003). The Groningen database is comparable with the OECD STAN Database. In order to reach a greater degree of industry detail, and to provide a comprehensive dataset without gaps, the Groningen database complements the OECD STAN data with information from industry and services statistics and additional (historical) national accounts data for individual countries.

The following adjustments are made to the STAN data for Germany, Japan, the Netherlands, and the United States:

Germany

The STAN Database provides observations for West Germany from 1970 to 1991 and for Germany from 1991 to the present. Data for West Germany are used for 1975-1991, while values for Germany are used for 1992 to the present.

The GGDC Database provides observations for West Germany from 1979 to 1991 and for unified Germany from 1991 to the present. The unified Germany spreadsheet also includes 1979-1990 estimates for unified Germany. In order to maintain consistency with the STAN database, for 1985-1991 we use total annual hours worked data for West Germany, while values for unified Germany are used for 1992 to the present.

Japan

The 2004 STAN Database and 2005 STAN Database do not report GFCF or GFCFK for industries in Japan. However, GFCF expenditures by manufacturing industries in Japan are reported in the 1974-1993 STAN Database and 1978-1997 STAN Database. The values for the overlapping years are identical. As a result, we develop a database with observations from 1975 through 1993 for 3-digit ISIC (Rev. 2) industries. In 1994, only values for major (2-digit) industries are reported. For each 2-digit industry, we assumed that the 1993 3-digit shares were applicable for 1994

New Investment (Installation Base) expenditures are used to produce estimates of capital expenditures for 1995 to 2002 (see Japan Statistical Yearbook). First, we developed a concordance between the industries listed in the Japan Statistical Yearbook and the 3-digit ISIC industries in the 1978-1997 STAN Database. Next, we calculate the annual percent change in capital expenditures – in current prices - between 1994 and 2002 using values from the Japan Statistical Yearbook. Using these percent changes, we extrapolate the 1994 values in the 1978-1997 STAN Database to 1995 through 2002.

. The current prices estimates of capital expenditures were converted to constant price estimates using the implicit deflator for total Gross Private Fixed Capital (plant and equipment) for the calendar year. Deflator estimates for 1975 to 1979, which are from the 1995 Japan Statistical Yearbook (p. 140), are then linked to the index values for 1980 to 2002 (see 2005 Japan Statistical Yearbook, p. 140).

Netherlands

The 2004 STAN Database and 2005 STAN Database do not report GFCFK data for the wood industry (ISIC 20, Rev. 3) for 1970 – 1986. Instead, we use the 2005 STAN Database to derive the implicit GFCF deflator for manufacturing from 1975 – 1987: (GFCF × 100)/(GFCFK × GFCF1995). We then divide the GFCF of the wood industry by the implicit GFCF deflator for manufacturing to derive real GFCF for wood industry for 1974-1987. Next, we calculate the GFCFK index (1987 =100) for 1975 to 1986. Finally, we multiple the 2005 STAN Database wood industry GFCFK value for 1987 by these index values. This allows us to convert 1975 to 1986 estimates of GFCFK for the wood industry from 1987 as the reference year (GFCFK = 100 in 1987) to 1995 as the reference year (GFCFK = 100 in 1995).

The 2004 STAN Database and 2005 STAN Database do not report GFCFK data for the non-metallic mineral products industry (ISIC 26, Rev. 3) for 1970 – 1986. Instead, we use the 2005 STAN Database to derive the implicit GFCF deflator for manufacturing from 1975 – 1987: (GFCF × 100)/(GFCFK × GFCF1995). We then divide the GFCF of the non-metallic mineral products industry by the implicit GFCF deflator for manufacturing to derive real GFCF for non-metallic mineral products industry for 1974-1987. Next, we calculate the GFCFK index (1987 =100) for 1975 to 1986. Finally, we multiple the 2005 STAN Database non-metallic mineral products industry GFCFK value for 1987 by these index values. This allows us to convert 1975 to 1986 estimates of GFCFK for the non-metallic mineral products industry from 1987 as the reference year (GFCFK = 100 in 1987) to 1995 as the reference year (GFCFK = 100 in 1995).

Germany, Japan, Netherlands, and the United States

For machinery & equipment, (ISIC 29-35), GFCFK and VALUK are derived for the following two sectors: Machinery and Equipment (ISIC 29-33) and Transport Equipment (ISIC 34-35). This is accomplished in the following steps:

(1) We calculate separate price indexes for Machinery and Equipment (ISIC 29-33) and
Transport Equipment (ISIC 34-35),

(2) We combined the price indexes (from step 1) and nominal GFCF and Value Added to calculate Real GFCF and Real Value Added (in monetary units) for both sectors,

(3) We add Real GFCF and Real Value Added (from step 2) to create Real GFCF and Real Value Added for Machinery & Equipment sector (ISIC 29-35),

(4) Finally, we use Real GFCF and Real Value Added (from step 3) to derive GFCFK and VALUK, quantity indexes for Machinery & Equipment (ISIC 29-35).

Exchange rates and purchasing power parities

In order to calculate a single frontier for all four countries, value added and capital expenditure data are converted to a common monetary unit. Harrigan (1997, 1999) discusses the difficulties of converting industry-level value added and investment expenditures into a single currency.

Value Added

For each industry and country, we convert nominal value added to real value added using the procedure developed by the Groningen Growth and Development Centre (GGDC) when it calculates value added per hour worked in U.S. dollars. We use value added in current prices (VALU) and the index of value added (VALUK) to derive real value added. Next, we convert all real value added data to a single currency (i.e., U.S. dollars). These adjustments are summarized in the following equation:

ycjt = (Ycjt / π) / e

where ycjt is real value added (in 1997 U.S. dollars) for country c, industry j, and year t. Ycjt is nominal value added for country c, industry j, and year t. πis the implicit value added deflator for country c, industry j, and year t. e is the unit value ratio for value added for country c and industry j.

Ycjt is the VALU variable from the 2005 STAN Database. VALU values for the Netherlands and Germany are listed in Euros in the 2004 STAN Database and 2005 STAN Database. For pre-EMU years (typically before 1999), the OECD value added data for Germany and the Netherlands were converted from guilders and deutsche mark (DM) to Euros by applying the irrevocable conversion rate (see OECD 2001b, p. 6; Schreyer and Suyker 2002, p. 7). The derivation of π was explained earlier in this Appendix. Finally, we must determine the appropriate values for e

We convert real value added (Ycjt / π) into U.S. dollars using unit value ratios (UVR), The UVR values were developed by the GGDC as part of the International Comparisons of Output and Productivity (ICOP) 1997 Benchmark Database. The GGDC provides 1997 UVRs for total manufacturing and most 2-digit ISIC (Rev. 3) manufacturing industries for Germany, Japan, the Netherlands, and the United States. When aggregating the UVRs of two or more industries, we use 1997 value added (in current monetary units) from the GGDC as weights.

O’Mahony and van Ark (ed.) (2003) are the source of UVR values for Germany and the Netherlands, while Inklaar, Wu and van Ark (2003) provide the UVR values for Japan. For Germany and the Netherlands, the Groningen UVR data are in Euro per US dollar, while the UVR data for Japan are in yen per dollar. The UVRs for all industries in the U.S. are unity.

Gross Fixed Capital Formation (Capital Expenditure)

We use gross fixed capital formation in current prices (GFCF) and the index of gross fixed capital formation (GFCFK) to derive gross fixed capital formation in constant prices. After gross fixed capital formation is converted to constant dollars, it is necessary to convert all values into US dollars. These adjustments are summarized in the following equation:

icjt = (Icjt / π) / e

where icjt is real gross fixed capital formation (in 2000 U.S. dollars) for country c, industry j, and year t. Icjt is nominal gross fixed capital formation for country c, industry j, and year t. πis the implicit gross fixed capital formation deflator for country c, industry j, and year t. e is the purchasing power parity rate for investment expenditures for country c and year t.

Icjt is the GFCF variable from the 2005 STAN Database, while the derivation of π was explained earlier in this Appendix. Finally, we must determine the appropriate values for e.

The Penn World Table (PWT) Version 6.2, which is produced by the Center for International Comparisons at the University of Pennsylvania, (see Heston, Summers and Aten 2006) is the source of purchasing power parity data (PPP) for investment expenditures for 1975-2004 (base year is 2000). Although Harrigan (1997, 1999) acknowledged the weaknesses of these data for the purposes of industry-level productivity studies, they are the best available data.

According to the PWT, “the purchasing power parity in domestic currency per US dollar for GDP or any component, may be obtained by dividing the price level by 100 and multiplying by the exchange rate (domestic currency per US dollar)” (Heston, Summers, and Aten 2006, Data Appendix, p. 5). In order to use the PPP for investment values for Germany and the Netherlands, which are in domestic currency per US dollar, it is necessary to divide these PPP values by their respective irrevocable conversion rate (domestic currency per Euro). This adjustment is required because GFCF values for the Netherlands and Germany are listed in Euros in the 2004 STAN Database and 2005 STAN Database. The OECD GFCF data for Germany and the Netherlands were converted from guilders and DM to Euros for pre-EMU years (typically before 1999) by applying the irrevocable conversion rate (see OECD 2001b, p. 6; Schreyer and Suyker 2002, p. 7). The adjusted PPP values are in Euros per US dollar.

Pollution Abatement Capital Expenditures

In this study we use survey estimates of capital expenditures assigned to pollution abatement. Germany, Japan, The Netherlands, and the United States were the first nations to institute annual surveys of pollution abatement costs. Because the United States collected data on expenditures for air, water, and solid waste abatement, we select those media categories from the surveys of German, Japan, and the Netherlands that provide the closest match to the air, water, and solid waste categories in the United States survey.

Starting with data from 1975, pollution abatement capital stock data “Bruttoanlagevermögen für Umweltschutz in konstanten Preisen” (see Statistisches Jahrbuch 1992, p. 723) are available for eight manufacturing industries in Germany. In addition, the United States calculated pollution abatement capital stock based for selected industries (see Kappler and Rutledge 1982). Because we want to drive capital stock data using the same assumption across all four countries, these data are not used in this study.

Instead, information on the share of capital expenditures assigned to air, water, and solid waste abatement activities are used to derive estimates of capital stock assigned to good output production and capital stock assigned to pollution abatement. This is done to maintain consistency with the STAN capital expenditure data.

Germany