The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
FEDERAL RESERVE BANK OF SAN FRANCISCO
WORKING PAPER SERIES
A State Level Database for the Manufacturing Sector:
Construction and Sources
Robert S. Chirinko
Emory University and CESifo
Daniel J. Wilson
Federal Reserve Bank of San Francisco
October 2, 2009
Working Paper 2009-21
http://www.frbsf.org/publications/economics/papers/2009/wp09-21bk.pdf
The views in this paper are solely the responsibility of the authors and should not be
interpreted as reflecting the views of the Federal Reserve Bank of San Francisco or the
Board of Governors of the Federal Reserve System.
A State Level Database For The Manufacturing Sector:
Construction And Sources
Robert S. Chirinko
Emory University and CESifo
and
Daniel J. Wilson*
Federal Reserve Bank of San Francisco
October 2, 2009
* The authors gratefully acknowledge the helpful comments and assistance offered by Jane
Gravelle, Sharon Kozicki, and Huntley Schaller and the expert research assistance provided by
Ann Lucas and Charles Notzon. Robert Chirinko acknowledges critical financial support
provided by the Federal Reserve Bank of San Francisco. All errors and omissions are the sole
responsibility of the authors, and the conclusions do not necessarily reflect the views of the
organizations with which they are associated.
A State-Level Database For The Manufacturing Sector:
Construction And Sources
Appendix
This document describes the construction of and data sources for a state-level panel data
set measuring output and factor use for the manufacturing sector. These data are a subset of a
larger, comprehensive data set that we currently are constructing and hope to post on the FRBSF
website in the near future. The comprehensive data set will cover the U.S. manufacturing sector
and may be thought of as a state-level analog to other widely used productivity data sets such as
the industry-level NBER Productivity Database or Dale Jorgenson’s “KLEM” database or the
country-level Penn World Tables, but with an added emphasis on adjusting prices for taxes. The
selected variables currently available for public use are nominal and real gross output, nominal
and real investment, and real capital stock. The data cover all fifty states and the period 1963 to
2006.
2
A State-Level Database For The Manufacturing Sector:
Construction And Sources
This document describes the construction of and data sources for selected variables used in
Chirinko and Wilson (2008). These data are a subset of a larger, comprehensive data set that we
currently are constructing and hope to post on the FRBSF website in the near future. The
comprehensive data set will cover the U.S. manufacturing sector and may be thought of as a statelevel analog to other widely used productivity data sets such as the industry-level NBER
Productivity Database or Dale Jorgenson’s “KLEM” database or the country-level Penn World
Tables, but with an added emphasis on adjusting prices for taxes. The selected variables currently
available for public use are nominal and real gross output, nominal and real investment, and real
capital stock. The data cover all fifty states and the period 1963 to 2006.1
The state data described in this document measure economic activity in the manufacturing
sector. The primary raw source data for the state-level totals of output, investment, labor and
establishments counts is the Annual Survey of Manufacturers (ASM) conducted by the U.S. Census
Bureau. State-level totals (which the Census Bureau refers to as “AS-3” data) are reported in the
yearly volumes of the ASM publication. From 1994 onward, these data also can be found in the
yearly ASM Geographic Area Statistics (ASM-GAS) publications. Hereafter, we will refer to the
ASM data on state-level totals for all years as the ASM-GAS data. The ASM data are collected
from a large, representative sample of manufacturing establishments with one or more paid
employees. The 2004 ASM (Appendix B, p. B-1) defines the manufacturing sector as follows,
“The Manufacturing sector comprises establishments engaged in the mechanical,
physical, or chemical transformation of materials, substances, or components into
new products. The assembling of component parts of manufactured products is
considered manufacturing, except in cases where the activity is appropriately
classified in Sector 23, Construction. Establishments in the manufacturing sector are
often described as plants, factories, or mills and characteristically use power-driven
machines and materials-handling equipment. However, establishments that transform
materials or substances into new products by hand or in the worker’s home and those
engaged in selling to the general public products made on the same premises from
which they are sold, such as bakeries, candy stores, and custom tailors, may also be
included in this sector. Manufacturing establishments may process materials or may
1
The data used in Chirinko and Wilson (2008) covered only 1982-2004. Our data set does not include the District of
Columbia (D.C.) as the ASM-GAS data for D.C. appears to have been noticeably affected by the switch in 1997 from a
SIC to NAICS basis for defining the manufacturing sector.
3
contract with other establishments to process their materials for them. Both types of
establishments are included in manufacturing.”
The ASM manufacturing sector corresponds to NAICS sectors 31 to 33.
1. OUTPUT – Ys,t
Output is measured by real value added, and it is defined as nominal value added divided by
a price deflator,
BT,Y
Ys,t = Y$s,t / Pmfg,t
,
BT,Y
where Y$s,t is nominal value added output and Pmfg,t
is the price index for manufacturing output
BT,Y
series is
net of sales and excise taxes but before corporate income tax adjustments. Since the Pmfg,t
based on producer price indices, it measures average prices received by domestic producers (PPI).
Our database presents Ys,t in billions of constant 2000 dollars.
The Y$s,t series is obtained from ASM-GAS (e.g., in 2004, the data are published in Table
1, column F). Because the ASM was not conducted for the years 1979 to 1981, Y$s,t is missing for
all states for these years. In future versions of this data set, we intend to estimate these missing
values, but for now we leave these data points as missing. The ASM-GAS data for this series were
undisclosed for Minnesota for the years 1970 and 1971. We filled these two data points in via linear
interpolation between Minnesota’s value added data for 1969 and 1972. Our database presents
Y$s,t in billions of dollars.
BT,Y
series is obtained from INDUSTRY, the table labeled "Chain-Type Price Indexes
The Pmfg,t
BT,Y
as an index number with a
for Value Added by Industry," Line 12. Our database presents Pmfg,t
base year value in 2000 of 1.0.
2. CAPITAL -- K s,t
4
Capital input is measured by the real (constant-cost) replacement value of equipment
(excluding software) and structures, and this series is constructed from the following perpetual
inventory formula,
K s,t = K s, τ (1 − δmfg,t ) t −τ + Is,t
t = τ + 1,..., T ,
where K s, τ is the initial value of the real capital stock (where the index τ represents the initial
period), δmfg,t is the rate of economic depreciation (hence (1 − δmfg,t ) is the survival rate), and Is,t
is real total capital expenditure. The capital stock is dated end-of-period (EOP). Our database
presents K s,t in billions of constant 2000 dollars. Each component determining the capital stock is
discussed in the following subsections.
2.1. The Initial Value Of The Capital Stock -- K s,τ
The K s, τ series is measured by the book value of the capital stock adjusted for inflation,
(
)
BV
CoC
HC
K s, τ = K s,
τ * K mfg, τ / K mfg, τ ,
CoC
BV
where K s,
τ is the book value (historical-cost) of the capital stock for state s, K mfg, τ is the constant-
cost value of the capital stock for the manufacturing sector, and K HC
mfg, τ is the historical-cost value
of the capital stock for the manufacturing sector. All capital stock series are EOP. Inflation drives a
wedge between book value capital stocks (based on the original purchase cost of investment) and
(
)
HC
real capital stocks useful in economic analyses. The K CoC
mfg, τ / K mfg, τ ratio provides an
approximate adjustment for the inflation wedge based on national manufacturing industry data. Our
database presents K s, τ in billions of constant 2000 dollars.
We compute initial values of the real capital stock EOP for τ = 1962 and τ = 1981 . Note we
“re-initialize” the capital stock in 1981 (as opposed to simply using the perpetual inventory formula
starting with the 1962 initial stock estimate) for two reasons. First, the 1962 initial stock is
5
estimated (as described below) rather than observed, since book value was not collected by the ASM
prior prior to 1975, and so we do not want to rely too heavily on the 1962 initial value estimate.
Second and more importantly, data on capital expenditures are missing for 1979 to 1981. Thus, the
initial capital stock for 1981, based on book value data, likely is a better measure of the true capital
stock in 1981 than a capital stock measure based in part on imputed investment data from 1979 to
1981.
#
A provisional estimate of K s,1962 , K s,1962
, is estimated by solving backward using the
perpetual inventory formula, beginning with the 1975 data on the book value of capital (adjusted for
inflation), subtracting investment data from 1963 to 1975, and weighting these terms by survival
rates,
#
#
K s,1962
= K s,1975
(1 − δ1975 )−(1975−1962)
⎛ (1975 −1962 −1)
−⎜
(1 − δ)−(1975−1962 − j) Is,1975− j
∑
⎜
j= 0
⎝
(
#
BV
HC
K s,1975
= K s,1975
* K CoC
mfg,1975 / K mfg,1975
⎞
⎟
⎟
⎠
)
The first part of the first of the equations above starts with the 1975 book value of capital (adjusted
for inflation) and adds back all of the 1962 capital stock that has depreciated between 1962 and
1975. The second part then subtracts all of the investments made from 1963 to 1975, after adding
back to each year’s investment the portion that has depreciated between 1962 and when the
investment was made. In essence, this formula undoes all of the additions to and depreciation from
the original capital stock of 1962 and subsequent investments from 1963 to 1975. Note we choose
1975 as the year from which to work backwards since it is the earliest year in which book value data
are available from the ASM.
The final estimate of K s,1962 is then obtained by rescaling the provisional state estimates by
the national real capital stock total in 1962 from the BEA, K CoC
mfg,1962 . Specifically,
51
⎛
⎞
#
#
K s,1962 = K s,1962
* ⎜ K CoC
K
⎟.
⎜ mfg,1962 ∑ s,1962 ⎟
s =1
⎝
⎠
6
A potential inconsistency exists in using the BEA data to rescale our provisional estimate based on
ASM data. Software investment is included in the BEA data but excluded in the ASM data. During
the early 1960's, the discrepancy introduced by software investment is negligible. In 1963, software
investment was 1.3% of manufacturing investment (though software embedded or bundled in
computers and other equipment is not reflected in this figure). The impact of software investment is
likely less than this figure for two reasons. First, for the older vintages of investment entering the
1962 capital stock, their share is likely to be even smaller than 1.3%. Second, software depreciates
more rapidly than other capital. It would seem safe to conclude that that the discrepancy owing to
the different treatment of software investment is less than 1% of the 1962 capital stock.
BV
The K s,
τ series is obtained from ASM (e.g., in 1975, the data are published in Table 4, row
CoC
BV
5). Our database presents K s,
τ in billions of dollars. The K mfg, τ series is the product of a quantity
index and a base year value that converts the index into a real stock,
CoC
CuC
K CoC
mfg, τ = INDEXK mfg, τ * K mfg,t = 2000 ,
CuC
where INDEXK CoC
mfg, τ is the chain-type quantity index for the real capital stock and K mfg,t = 2000 is
the base year value for the current-cost value of the capital stock for the manufacturing sector. Our
CoC
database presents K CoC
mfg, τ in millions of dollars. The INDEXK mfg, τ is obtained from FIXED, Table
4.2, line 7, and this series is divided by 100. Our database presents INDEXK CoC
mfg, τ as an index
number with a base year value in 2000 of 1.0. The K CuC
mfg,t = 2000 datapoint is obtained from FIXED,
Table 4.1, line 7. Our database presents K CuC
mfg,t = 2000 in millions of dollars.
The K HC
mfg, τ series is obtained from FIXED, Table 4.3, line 7. Our database presents
K HC
mfg, τ in millions of dollars.
2.2. The Rate Of Economic Depreciation -- δmfg,t
7
The δmfg,t series is measured by the flow of annual depreciation divided by the capital stock
existing at the beginning of the year,
δmfg,t =
DCuC
mfg,t
K CuC
mfg,t −1
,
CuC
where DCuC
mfg,t is the current-cost flow of depreciation in manufacturing industries and K mfg,t −1 is
the current-cost capital stock in manufacturing industries. Our database presents δmfg,t in
percentage points.
CuC
The DCuC
mfg,t series is obtained from FIXED, Table 4.4, line 7. Our database presents D mfg,t
in millions of dollars.
The K CuC
mfg,t −1 series is obtained from FIXED, Table 4.1, line 7. Our database presents
K CuC
mfg,t −1 in millions of dollars.
2.3. Real Total Capital Expenditure -- Is,t
Real total capital expenditure is defined as nominal capital expenditures deflated by a price
index,
Is,t =
I$s,t
I
Pmfg,t
,
NEW
USED
,
I$s,t = I$s,t
+ I$s,t
NEW
USED
, and I$s,t
are total, new, and used nominal capital expenditures,
where I$s,t , I$s,t
I
is the price deflator for investment for the manufacturing sector. Our
respectively, and Pmfg,t
I
database presents Is,t in billions of constant 2000 dollars. The I$s,t and Pmfg,t
series are discussed
in the following subsections.
8
2.3.1. Total Nominal Capital Expenditure -- I$s,t
The I$s,t series is obtained in three different ways each of which are based on the ASMGAS and depend on disjoint time periods. (This mixture of direct and indirect estimates is forced
upon us because of some anomalies in the ASM-GAS.) The series represents nominal expenditures
on equipment (excluding software) and structures. Our database presents I$s,t in billions of dollars.
For 1977, 1978, and 1982 to 2004, the series is obtained directly from ASM-GAS (e.g, in
2004, the data are published in Table 2, column I). For 1963 to 1976, the ASM-GAS only publishes
NEW
NEW
. For these years, I$s,t is derived based on a state's mean ratio of I$s,t
to I$s,t ,
data for I$s,t
{
NEW
NEW
I$s,t = I$s,t
* MEAN s I$s,v / I$s,v
}
,
t = 1963,...,1976
v = 1977, 1978, 1982,..., 2004.
where the MEAN s {.} is computed separately for each state and over all available observations
represented by the index v.
For 1979 to 1981, the ASM was not conducted, and hence no ASM-GAS source data are
NEW
, nor Ys,t . The missing investment data for these three years are
available for I$s,t , I$s,t
estimated with the following three-step procedure. First, we rely on the availability of alternative
output data from BEA for these three years and the workhorse of investment modeling, the
accelerator model, to estimate the missing total capital expenditure data. Output is defined as real
Gross State Product (GSP) for the manufacturing sector.2 With these data and the available data for
Is,t , we estimate the following flexible accelerator model,
2
For all intents and purposes, Gross State Product is conceptually identical to Gross Domestic Product,
though small differences exist in some minor categories.
9
'
'
'
'
'
Is,t / Ys,t
= αs + βs,0 ( ΔYs,t
/ Ys,t
−1 ) + βs,1 ( ΔYs,t −1 / Ys,t − 2 )
'
'
+ βs,2 ( ΔYs,t
− 2 / Ys,t − 3 ) + εs,t
,
t = 1977, 1978, 1982, ..., 2004
where αs is a state-specific constant capturing state fixed effects, the βs 's are state-specific slope
'
'
parameters, εs,t is an error term, and Ys,t
is real manufacturing GSP. The Ys,t
series is nominal
manufacturing GSP divided by a price deflator. Nominal manufacturing GSP is obtained from the
BEA’s Regional Economic Accounts (REA) data. (In 1997, the data are reported on both SIC and
BT,Y
discussed in Section 1.
NAICS bases; we use the SIC figures.) The deflator is Pmfg,t
Second, we use the estimated parameters (represented by ^ 's over the α and the β 's), data
'
I
and Ps,t
, and a transformed version of the above equation to generate a provisional
for Ys,t
#
) for the missing nominal capital expenditure observations,
estimate of I$s,t ( I$s,t
#
'
'
'
'
'
'
ˆ
ˆ
⎤
I$s,t
/ Ys,t
= ⎡ αˆ s + βˆ s,0 ( ΔYs,t
−1 ) + βs,1 ( ΔYs,t −1 / Ys,t − 2 ) + βs,2 ( ΔYs,t − 2 / Ys,t − 3 ) ⎦
⎣
'
I
* Ys,t
* Ps,t
.
t = 1979, 1980, 1981
Third, for each year (1979, 1980, 1981), we rescale states’ nominal investment so that it
equals the national total, I$ASM
mfg,t , which we estimate by applying the growth rate of the BEA’s
nominal private nonresidential fixed investment (net of software) for the manufacturing sector,
I$mfg,t, to the previous year’s value of national investment reported in the ASM. Specifically, we
multiply each state’s provisional estimate by the ratio of national manufacturing investment to the
national sum of the provisional estimates,
10
⎛
#
ASM
I$s,t = I$s,t
* ⎜ I$mfg,t
⎜
51
⎞
#
⎟
∑ I$s,t
⎟,
s =1
⎝
⎠
t = 1979,1980,1981
51
⎛ I$mfg,1979 ⎞
I$ASM
I$
=
⎟
mfg,1979 ∑ s,1978 ⎜⎜
⎟
I$
mfg,1978
s =1
⎝
⎠
⎛ I$mfg,1980 ⎞
ASM
I$ASM
I$
=
⎟.
mfg,1980
mfg,1979 ⎜⎜
⎟
⎝ I$mfg,1979 ⎠
⎛ I$mfg,1981 ⎞
ASM
I$ASM
⎟
mfg,1981 = I$ mfg,1980 ⎜⎜
⎟
⎝ I$mfg,1980 ⎠
The I$mfg,t series is obtained from FIXED, Table 4.7, line 7 less the sum of software investment
over all manufacturing industries (NAICS sectors 31 to 33) from DETAILED, row 9
The ASM-GAS data for I$s,t need to be adjusted for additional missing values and an error.
The additional missing values occur because the ASM-GAS did not report data for Minnesota for the
years 1970 and 1971. We use the relation between BEA data for the manufacturing sector and state
data for Minnesota on investment expenditures to impute the missing values with the following
relation,
⎧⎪ I$s = minnesota,v ⎫⎪
I$s = minnesota,t = MEAN ⎨
⎬ * I$mfg,t
I$mfg,v
⎩⎪
⎭⎪
t = 1970, 1971
v = 1967, 1968, 1969, 1972, 1973, 1974
where I$mfg,t is nominal capital expenditures on new and used capital by the manufacturing sectors
defined above and the mean of the ratio is computed for three years before and after the missing
values. The I$mfg,t series was discussed previously in this subsection.
The error occurs for I$s = ohio,t =1996 . In 1996, ASM-GAS shows a 400% jump in nominal
total capital expenditures in Ohio from about $8 billion in 1995 to $40 billion in 1996 and then back
down to $9 billion in 1997. This enormous jump can be traced to the motor vehicles sector ($35
11
billion), which suggests a huge capital investment – equal to 85% of the sector’s national capital
expenditures – for the building of an auto plant(s) in Ohio in 1996. We dismiss this number for three
reasons. First, the magnitude of this investment is implausible. By comparison, DaimlerChrysler's
jeep plant expansion in Toledo in 1998 was $1.2 billion of total investment over several years.
Second, correspondence with experts on the Ohio manufacturing sector (including one at the Ohio
Department of Economic Development) could not confirm any massive capital expenditure
programs in 1996. Third, the 1996 value for national total capital expenditures reported in the ASMGAS is inconsistent with and about $32 billion higher than a comparable figure reported in a
separate ASM publication, Statistics for Industry Groups and Industries (ASM-SIGI). These two
publications disagree on national capital expenditures only in 1996, suggesting an error is present.
We thus conclude that I$s = ohio,t =1996 = $40 billion is erroneous.
We fill in the 1996 Ohio data point by simply taking national manufacturing capital
expenditures from the alternative ASM publication, ASM-SIGI, and subtracting the sum of capital
expenditures from all other states.
I
2.3.2. Price Deflator For Investment -- Pmfg,t
The price deflator for investment is constructed as an implicit deflator,
I
Pmfg,t
=
I$mfg,t
I mfg,t
,
where I$ mfg,t and I mfg,t are nominal and real total capital expenditures, respectively, for the
I
as an index number with a base year value in
manufacturing sector. Our dataset presents Pmfg,t
2000 of 1.0.
The I$mfg,t series was discussed in the preceding subsection (Total Nominal Capital
Expenditure).
The I mfg,t series is the product of a quantity index and a base year value that converts the
index into real investment expenditures,
12
I mfg,t = INDEXI mfg,t * I$mfg,t = 2000 ,
where INDEXI mfg,t is the chain-type quantity index for real investment expenditures and
I$mfg,t = 2000 the base year value for current investment expenditures. Our database presents I mfg,t
in billions of dollars. The INDEXImfg,t is obtained from FIXED, Table 4.8, line 7, and this series is
divided by 100. Our database presents INDEXImfg,t as an index number with a base year value in
2000 of 1.0. The series containing the I$mfg,t = 2000 datapoint was discussed in the preceding
paragraph.
3. LEGEND/REFERENCES
ASM:
CENSUS, Annual Survey of Manufactures, Complete Volume (Various
Years). Complete volumes are available until 1993. In later years, the data
are presented in separate pamphlets, which include ASM-GAS and ASM-SIGI
mentioned below.
ASM-GAS:
CENSUS, Annual Survey of Manufacturers, Geographic Area Statistics
(Various Years). Publications for the years 1994 to 2004 (except 1997
and 2002) are available online. These data are published on an
establishment basis. The data are obtained from electronic or paper
documents depending on the time period: 2004 (Census website);
2003 to 1972 (CD's purchased from Census); 1971 to 1963 (paper copies).
URL: http://www.census.gov/mcd/asm-as3.html.
ASM-SIGI:
CENSUS, Annual Survey of Manufacturers, Statistics for Industry Groups
and Industries (1996).
URL: http://www.census.gov/mcd/asm-as1.html.
BEA:
Bureau of Economic Analysis, U.S. Department of Commerce.
URL: http://www.bea.gov.
BLS:
Bureau of Labor Statistics, U.S. Department of Labor.
URL: http://www.bls.gov.
BOP:
Beginning-Of-Period t.
CENSUS:
Bureau of the Census, U.S. Department of Commerce.
URL: http://www.census.gov.
13
Chirinko, Robert S., and Daniel J. Wilson. 2008. “State Investment Tax Incentives: A Zero-Sum
Game?” Journal of Public Economics, 92(12): 2362–2384.
DETAILED:
BEA, Detailed Fixed Assets Tables, Nonresidential Investment,
Historical-Cost.
URL: http://www.bea.gov/bea/dn/FA2004/Details/xls/detailnonres_inv1.xls
EOP:
End-Of-Period t.
FIXED:
BEA, Standard Fixed Asset Tables.
URL: http://www.bea.gov/bea/dn/FA2004/SelectTable.asp.
INDUSTRY:
BEA, Gross-Domestic-Product-by-Industry Accounts. URL:
http://www.bea.gov/bea/industry/gpotables.
PPI:
BLS, BLS Handbook of Methods, Chapter 14 Producer Prices.
URL: http://www.bls.gov/opub/hom/pdf/homch14.pdf.
REA:
BEA, Regional Economic Accounts: Gross State Product.
URL: http://www.bea.gov/bea/regional/gsp/default.cfm?series=SIC.