An Annual Index of US Industrial Production, 1790

An Annual Index of U. S. Industrial Production, 1790-1915
Author(s): Joseph H. Davis
Reviewed work(s):
Source: The Quarterly Journal of Economics, Vol. 119, No. 4 (Nov., 2004), pp. 1177-1215
Published by: Oxford University Press
Stable URL: http://www.jstor.org/stable/25098716 .
Accessed: 01/08/2012 14:07
Your use of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at .
http://www.jstor.org/page/info/about/policies/terms.jsp
.
JSTOR is a not-for-profit service that helps scholars, researchers, and students discover, use, and build upon a wide range of
content in a trusted digital archive. We use information technology and tools to increase productivity and facilitate new forms
of scholarship. For more information about JSTOR, please contact [email protected].
.
Oxford University Press is collaborating with JSTOR to digitize, preserve and extend access to The Quarterly
Journal of Economics.
http://www.jstor.org
THE
quarterly
of
journal
economics
November
Vol. CXIX
Issue 4
2004
INDEX OF U. S. INDUSTRIAL
PRODUCTION, 1790-1915*
AN ANNUAL
Joseph
a remedy
As
this
data,
consistently
quantity-based
ing industries
industrial
for the notorious
introduces
study
denned
from
annual
an
1790
series
in a manner
production
of the U. S.
volatility
index is that
index.
of pre-Civil War
deficiency
annual
index of American
until
The
economy
tionary
depressions
1873 were
actually
I. The
changes
World
War
also
the
that
financial
expressed
Board's
of
I. A direct
demonstrates
followed
purportedly
when
rather mild
recessions
that
production
index
view
in industrial
differences
S. macroeconomic
Reserve
our
index
before
U.
industrial
43
incorporates
in the manufacturing
and min
War
World
(most entirely
new)
to the Federal
similar
antebellum-postbellum
The
index
indistinguishable.
tically
H. Davis
the
monthly
and
growth
implication
are
volatility
the pernicious
of the
statis
defla
in 1837
panics
in real output.
and
I. Introduction
data are an indispensable
tool in accurately
evalu
Reliable
In reality, economic historians,
of an economy.
ating the evolution
are routinely
much
like archeologists,
forced to interpret
past
events from the statistical
at hand. Since contemporary
artifacts
federal agencies
and private
collected
organizations
infrequently
*
This paper
is based upon my dissertation
at Duke University.
I
completed
to thank Paul Rhode,
Gianni
Milton
Toniolo,
Irwin, Peter
Douglas
Temin,
Charles
Michelle
Friedman,
Hanes,
Clotfelter,
Christopher
Connolly,
Jeffrey Wil
Claudia
Charles
Kenneth
Winifred
liamson,
Goldin,
Sokoloff,
Kindleberger,
an anonymous
Lawrence
Katz
and
referee
(the editor),
Rothenberg,
(especially),
seminar
from Duke University,
Harvard
the Economic
participants
University,
Association
Annual
the NBER
of the American
History
Meeting,
Development
Summer
Institute
and the Triangle
Economic
Work
Economy
meetings,
History
The companion
Data Appendix
is avail
Technical
shop for valuable
suggestions.
able electronically
from the author
upon request
([email protected]).
The views
in this article
do not necessarily
reflect
of The Van
those
expressed
Inc.
guard Group,
wish
? 2004 by the President
and Fellows
Technology.
The Quarterly Journal
of Economics,
of Harvard
College
November
2004
1177
and the Massachusetts
Institute
of
1178
QUARTERLY JOURNAL OF ECONOMICS
economic
definitive
facts on the historical
and
statistics,
growth
fluctuations
ofthe early U. S. economy are scarce. Existing
bench
mark
estimates
of national
income back to 1790 are very ques
to evaluate,
tionable and difficult
David
[1967] to aptly
prompting
coin the first half of the nineteenth
"statistical
century America's
dark
age."
Annual
far
are
estimates
even
less
reliable.
of U.
S. economic
Currently,
activity
macroeconomists
going
back
have
that
ac
cess
to two annual
series for the pre-Civil War
output
period:
Robert Gallman's
annual
estimates
for the 1834
unpublished
1859 period compiled
in the 1960s, and Berry's
[1988] real GNP
was never sufficiently
series from 1789. Gallman
confident
ofthe
to publish
of his annual estimates
reliability
them, and chastised
researchers
to use them in an analysis
who attempted
of early
American
business
employed
regression
analysis
cycles.1 Berry
on a hodge-podge
of industrial,
and price data in order
financial,
to estimate
annual real GDP for the 1789-1889
period. However,
researchers
and Gallman
[1983] and Rhode
(e.g., Engerman
series as far removed
from reality.
[2002]) have dismissed
Berry's
reason
The primary
is that Berry's
final data are an ad hoc
drawn from hundreds
of overlap
of select extrapolations
average
on
a
that
rest
back-casts
sparse set of
ping regression
ultimately
In short, the vast majority
price indexes and nominal
aggregates.
nor Gallman
of economic
the Berry
historians
regard neither
series to be of sufficient
to confidently
infer the historical
quality
con
evolution
of the early American
economy.
Quite
frankly,
a trustworthy
U. S. GDP
series
is simply
structing
impossible
before the Civil War
of an
owing to the comparative
deficiency
nual data on agriculture,
merchant
and wholesale
and
trades,
As it stands,
industries.
fluctuations
business-cycle
today
as those who
constraints
studied
service
than
a generation
researchers
of historical
U. S.
face essentially
the same data
more
these same phenomena
ago.
an
This
void by building
study aims to fill this statistical
measure
the
entire
annual
that
pre
spans
output
consistently
I economy. The paper begins by discussing
the meth
World War
to construct
data employed
the new index
odology and component
1. See Rhode
out that a 1963
data. Rhode
[2002] for Gallman's
[p. 12] points
from Robert Gallman
with
the annual
data circulated
the
containing
mimeograph
as reliable,
not be regarded
disclaimer:
should
"NOTE: These
figures
following
aver
for the purpose
of computing
decade
annual
estimates.
derived
They were
as
not for analysis
to interested
for testing,
technicians
ages and are supplied
series."
annual
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1179
an annual mea
in Section
II. Specifically,
the paper assembles
sure analogous
to the Federal
in methodology
and interpretation
In doing so,
industrial
index.
Board's monthly
Reserve
production
data on 43 manufactur
I have collected
annual physical-volume
is
The paper's quantity-based
industries.
sample
ing and mining
sense
in
that
its
the
components
indirectly
quite comprehensive
of the value added produced
close to 90 percent
represent
by the
sector during
the nineteenth
U. S. industrial
century.
some implications
III then considers
of the new in
Section
of American
before
dex. Since our knowledge
(and,
production
is severely
this new annual
indeed, after) the Civil War
limited,
new insights
index of industrial
may give researchers
production
even allow new questions
to be
and should
into old questions,
the
For one, is the conventional
wisdom
answered.
surrounding
of early American
pace of secular development
industry accurate?
take off and catch up
did American
industrial
When
productivity
in
to its European
business
Were
contractions
counterparts?
as
severe
as
observers?
America
portrayed
by contemporary
early
differ fundamentally
before and af
Did business-cycle
volatility
ter the Civil War? Are the early business
set down
dates
cycle
the
National
Bureau
of
Economic
Research
(NBER)
long ago by
to these
in particular,
latter questions,
realistic? Answers
gain
in light of the growing
relevance
literature
additional
regarding
in U. S. economic
the structural
break
since the mid
volatility
1980s (e.g., McConnell
and Quiros
[2000] and Stock and Watson
some of these questions,
III addresses
and Sec
[2002]). Section
the paper. A Data Appendix
tion IV concludes
briefly describes
of the index components.
the sources and quality
II. Data
II.A.
and
Index Methodology
Overview
index of industrial
for the 1790
My new annual
production
1915 period
with
Federal
the
Reserve
compares
conceptually
Board's historical
industrial
available
index
monthly
production
since 1919. Both
indexes
to measure
the same funda
attempt
the level of physical
in the nation's
namely
production
and
industries.
the primary
manufacturing
mining
Naturally,
attribute
of an index of industrial
is
that
it
is devoid of
production
in the index reflect purely fluc
nominal
data, so that changes
mentals,
tuations
in real
teenth
century,
output. And of particular
index often captures
my
for the nine
relevance
not only the factory
QUARTERLY JOURNAL OF ECONOMICS
1180
but also the wares manufac
businesses,
output of incorporated
and "industrial"
tured by private
at
businesses
goods produced
home
the
contract
under
for
sale
later
(say,
putting-out
system)
on the open market.
an index of industrial
Yet how should we interpret
produc
tion for the early U. S. economy
that could best be characterized
as a largely agrarian
it is true that more
emerging market? While
of national
than one-half
United
States
output in the antebellum
was agricultural,
I strongly maintain
that the new index is ap
to define
the historical
evolution
of American
propriate
growth
if for no other reason
and business
than the fact that
cycles,
America's
emergence
as
an
economic
power
is commonly
equated
with
its industrialization.
More generally,
of
the new index should be broadly
indicative
economic
conditions
because
the industrial
the nation's
broader
sector has historically
derived
demand
from nonindus
directly
and the con
trial occupations,
farmers, merchants,
particularly
the demand
for
struction
trades. The processing
of foodstuffs,
to
and
the
equipment
required
machinery,
capital
agricultural
are
to
all
market
commodities
intimately
transport
agricultural
tied to farm output and the relative
goods,
price of agricultural
even though
as
is often characterized
agricultural
production
the state of the nineteenth-century
Likewise,
shipbuild
acyclical.
of the mari
upon the health
dependent
ing industry was heavily
the Em
time trades. Indeed, the nonintercourse
period following
impact on the nation's
shipbuild
bargo of 1807 had a devastating
the manu
[2003]). Likewise,
(see Irwin and Davis
ing industry
were
and transportation
of lumber products
facture
equipment
in
construction
the
to
business
conditions
sensitive
acutely
sectors.
and inland transportation
the railroad
trades,
industry,
nonindustrial
between
such
Indeed,
relationships
synchronous
and others
the NBER
sectors
is precisely
and industrial
why
as
of
indicators
coincident
indexes
industrial
production
classify
of
in
the
share
the
drop
cyclical turning points despite
precipitous
In
to industrial
the nation's
labor force dedicated
production.
and mined
of manufactured
quantities
short, an annual measure
eco
a much-needed
metric
of prewar American
should provide
nomic
activity.
II.B.
Component
Series
of 43
data on the physical
annual
I have assembled
output
on two principal
criteria
advocated
based
industries
by Romer
series
and Hanes
[1994]. First, any annual
[1991] and Calomiris
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1181
or to a
to actual output,
either directly
had to pertain
employed
This
selection
first
related
proxy.
guideline
physical-quantity
or financial
business
of general
indicators
eliminated
annual
Al
not explicitly
associated
with genuine
conditions
production.
wholesale
before
Civil
the
available
War,
conveniently
though
variables
such as bank
prices, equity prices, and other financial
from
the
The new
omitted
index.
have
been
entirely
clearings
stands
in sharp
focus on physical
index's exclusive
quantities
"business
to various
condition"
late-nineteenth-century
income estimates.
ad hoc national
indexes, as well as to Berry's
raw source data had to be
selection
As a second
criterion,
in order to preserve
index
for long stretches,
available
annually
over time. Specifically,
I omitted
and comparability
consistency
did not run at least
coverage
aggregate
products whose
existing
30 years before and after the Civil War. The second rule excluded
before the Civil War?most
five products
that were manufactured
whose
annual
alcohol
and
tobacco
output
products?but
notably
The 60-year
cutoff was not
data were only collected
thereafter.2
an index whose
set to avoid building
but rather was
arbitrary,
over
time.
reliability
changed
I presents
the share of index components
Panel B of Table
com
sources. Familiar
and private
that come from government
as
or
come
coal
from
the
lead
such
series,
modity
production,
contrast
of
States
Statistics
[U. S. Department
of the United
of Census
Historical
Sta
Bureau
1975]; henceforth,
Commerce,
tistics. More than one-half
of the quantity-based
series, however,
are novel in the sense that the data have previously
been unavail
in an aggregate,
format. This
able to researchers
user-friendly
sources
series
from
the
quantity-based
private
study develops
trade
such as published
company
unpublished
publications,
firm
records, historical
research,
society collections,
antiquarian
in
and
Historical
societies,
par
studies,
private
correspondence.
ticular, proved a rich source of base data that have only recently
the culmination
of years
of research
emerged
following
by its
sources
a
as
act
members.
these
historical
secondary
Collectively,
for the various
that currently
statis
surrogate
agencies
provide
for inclusion
in its monthly
tics to the Federal
Reserve
index.
New annual production
series have been compiled for an array
Historical
of final manufactured
2. The
five excluded
and would
era.
index,
present
the postbellum
products,
including
fire engines,
naval
ships,
series would
less than 5 percent
to the
annual
contribute
not meaningfully
alter the index's cyclical properties
for
1182
QUARTERLY JOURNAL OF ECONOMICS
TABLE I
Data
Quality
and
for
Sources
Various
Industrial
New
Panel
Actual
annual
production
manufacturing
index
1790-1915
A. Share
production
Indirect
23.3
proxy
Panel
Government
Private
B. Share
60.5
monthly
G.17
index
Since 1919
pertaining
to:
54.8
42.5
39.5
sources
components
57.5
(%) of index
sources
Federal
Reserve's
index
1860-1914
(%) of index
76.7
Indexes
Frickey's
industrial
Production index
Data coverage
Production
45.2
components
obtained
from:
75.0 72.5
27.5
25.0
Sources: Author's calculations based on information in Frickey [1947], U. S. Board of Governors of the
Federal Reserve System [1986, pp. 34-35, Tables 3.1 and 3.2] and Davis [2002].
Designations have been standardized to adhere to current Federal Reserve classifications. Series that
represent a hybrid of official and secondary sources have been assigned to the source that contributes the
majority of observations. Government sources include "official"data collected by national and state agencies.
Private sources pertain to all other reporting bodies, including trade groups, individual companies, firm
archives, and historical societies.
and scientific
newspapers,
firearms, musical
instruments,
watches,
and minor
items. Furthermore,
the data sets for locomo
apparel
tives, merchant
ships, and pig iron extend or refine the conventional
in Historical
series currently available
The Data Appen
Statistics.
a brief description
an
dix provides
of the 43 index components;
Technical
Data
with
unpublished
Appendix
considerably
companion
more detail is available
from the author upon request.
time series on U. S. locomotives
in Figure
I is a fine
My
of the reliability
of the privately
and comprehensiveness
example
source data. I have successfully
collected
traced the year of con
struction
for more than 120,000
the
by cross-referencing
engines
rosters of thousands
and locomotive
build
of railroad
companies
re
assembled
have
been meticulously
ers, which
by various
over
the present
searchers
the past 80 years.
Consequently,
locomotive
series is longer and more
comprehen
study's annual
loco
antebellum
sive than either Fishlow's
[1965] unpublished
or
motive
Burns's
series.
[1934] postbellum
series,
The index also draws on largely
Inspection
ignored sources.
to believe
is good reason
records are a case in point. There
that
an
were
was
in
since
records
it
accurate,
reasonably
inspection
so.
in
to
best
interest
make
them
States
mandated
authority's
wares
on
a
not
wide
of
manufactured
array
only to
spections
maintain
product
quality
(thereby
bolstering
the area's
reputa
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1183
7,500 f-?-T
4 ^
m 2,500
7,500
\ /sAn. / /Y \ i
\JW\rA\ 4- 2 500
\J
U
Newsenes
1
axis)~ ^
left
/ 2
{
ioo4"of
Aa
/
25 4-
// V\ /
o>
c
W//
|i
I
j
V
\/
\
Birth of industry'sdomesticmanufacture
1840
1825
1855
1870
Figure
U.
Ratio
S. Locomotive
See Davis
scales
1 100
Fishlow series
j
(1833-1860;
rightaxis) |
ill_
Sources:
V
y f 1,000
/
4-250
Burns seriesT
\/
/ (since1880;rightaxis)
T3/ /V.
/ a/
j
/
\r\r\J
*" \j
I\ /
]Ai j
g 2504 J
o
/\
\f
(since 1825;
^j 1,000t
have
1885
1900
1915
I
Construction?New
and Old
Data
[2002, 2004a].
been offset for clarity.
tion in the trade), but also to ensure that they received
the proper
the State of New York collected
Because
roy
royalty payments.
of
of salt produced
alties based on the quantity
by the hundreds
is
its leased Onondaga
it
that worked
establishments
reservation,
accu
to maintain
not surprising
inclined
that state officials were
rate inspection
to
often applied
records. Although
inspections
or
states
localized
cities
established
these
production,
typically
of the industry's
economic
because
systems
precisely
regulatory
with
the result that local or state inspection
records
importance,
often accounted
For
for a sizable
share of national
production.
in Massachusetts
officials
example,
routinely
inspected
quanti
ties of mackerel
cured in its ports, and Massachusetts
accounted
90 percent
for roughly
of the nation's
salted mackerel
output
during the nineteenth
century.
As they continue
to do so today, local trade organizations
and
industry groups in the nineteenth
century also tracked the physical
movement
of select manufactures
to and from their cities. In most
and flow of manufactured
items from multiple
cases, the production
to gauge national
regions have been aggregated
output. The broad
QUARTERLY JOURNAL OF ECONOMICS
1184
sets for lumber shipments
(eleven series),
and hog packing
the
(four series) parallels
incor
and thereby explicitly
expansion
porates profound
spatial shifts in timber and farming.
are not as comprehensive
as
measures
Still, some production
one would
and gunpowder,
like. For agricultural
pro
machinery
out
at the dominant
firms substitutes
for industrywide
duction
to
these two series are inherently more susceptible
put. Although
the approach
is ten
bias and idiosyncratic
shocks,
survivorship
structure
industries.
of the respective
able due to the oligopolistic
Harvester
and the International
1900s, DuPont
By the early
over 80 percent
of the powder
and har
controlled
Corporation
were
firms
di
and
both
vester markets,
eventually
respectively,
and other data limitations
violations.
These
vested
for antitrust
are discussed
at the end of the paper.
in the Data Appendix
one-fourth
of
As shown in Panel A of Table
I, approximately
conven
In
the components
represent
production.
only indirectly
ening of the regional data
flour receipts (four series),
nation's
ongoing westward
for
of primary
tional output
indexes, quantities
inputs substitute
data
suitable
when
the output of a manufactured
good
production
are unavailable.
and Romer
For instance, Miron
[1990] infer the
at Chicago
of
cattle
from
the
head
of
dressed
beef
arriving
output
soon after
were
on
that
cattle
slaughtered
grounds
stockyards
of
this approach,
my index uses the quantity
receipt. Following
fleet to substitute
oil returned
by the U. S. whaling
unprocessed
of sperm and whale
since virtually
all barrels
for oil processing,
the present
refined
dockside.
oil were
Similarly,
immediately
of raw
the
standard
index follows
quantity
by using
practice
as a surrogate
of cotton tex
for the production
cotton consumed
on more
because
is necessary
tiles. This substitution
quantities
were
not consistently
yarn or apparel)
Federal
In
Reserve's
the
monthly
fact,
reported
of
to measure
the output
index continues
industrial
production
of fiber consumed.
cotton and wool fabrics by the quantity
in
can also accurately
reflect relative
Trade volume
changes
a
the
when
domestic
represents
commodity
imported
production
in
have
material.
Researchers
routinely
nonindigenous
primary,
of silken goods from raw silk imports or
ferred the manufacture
finished
products
for the
(e.g.,
1800s.
of green coffee beans. Unfor
from the importation
roasting
in
data on U. S. imports are tainted by breaks
tunately,
pre-1870
and often pertain
definitions
only to imported values.3
fiscal-year
coffee
from
3. The U. S. Treasury
fiscal years
ending
Department
September
switched
30 to years
trade statistics
reporting
30.
June
ending
in 1843
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1185
In the absence
of reliable
import deflators,
imported
quantities
cannot be imputed with confidence. Moreover,
many U. S. imports
that were
in quantities
before
delineated
the Civil War were
reason
to
which
is
often
the
domestic
officials
tariffs,
subject
in the first place.4
tracked certain quantities
This paper
circumvents
these
complications
by consulting
British
custom records for three imported quantities:
copper, silk,
and tin. Trade
statistics
for Great Britain's
House
of
published
in the Sessional
Commons
Pa
(a.k.a. the Parliamentary
Papers
for the present
the British
data
pers) are attractive
study because
are available
much
earlier
and are of higher
vis-a-vis
quality
their American
counterparts.
consid
My approach may be illuminated
by a more detailed
of implied copper and tin consumption.
eration
de
Consistently
fined over the prewar era, Great Britain's
customs
records list the
of unwrought
tin shipped from
copper and unprocessed
quantities
to
all British
the
in
United
States
both American
ports directly
and foreign vessels
the
calendar
In
the
year.
during
addition,
are
Sessional
Papers
detailed,
extremely
delineating
quantities
to which
and by country
The
by type of manufacture
exported.
me
measure
detailed
Sessional
to
allow
the
domestic
Papers
of copper and tin products
manufacture
(e.g., pewter)
by the
transatlantic
of
British
and
copper
shipments
foreign unwrought
and tin departing
British
States.
ports for the United
Indeed,
these transatlantic
flows accounted
for nearly
two-thirds
of the
ores imported by the United
States
the nine
unwrought
during
teenth century, given the predominance
of the Bristol,
Cornwall,
and Devon mines,
and given the fact that London
and Liverpool
were the last ports-of-call
before crossing
the Atlantic
Ocean.5
II.C.
Component
Weighting
arrive at an index of industrial
individual
production,
series
must
be
their
relative
component
weighted
by
importance.
census reports have historically
Federal
such informa
provided
tion in the form of an industry's
or the difference
value added,
To
4. Consider
the case of coffee. Unroasted,
or green,
coffee beans
imported
by
were
the United
to duties
States
1832. Coffee
before
duties were
lowered
subject
in 1797 and doubled
the War
of 1812. From
1832 until
the
temporarily
during
Civil War,
in part because
the federal
sub
imports were
duty-free,
government
stituted
coffee for rum
rations.
(!) in military
5. Merchants
in Boston,
New
and elsewhere
York, Philadelphia,
routinely
commissioned
London merchants
to deliver
cargoes.
specific metal
By 1790 ship
ments
took less than a month
to traverse
the Atlantic
Ocean
and regularly
arrived
stateside
the spring
seasons.
and fall shipping
during
1186
QUARTERLY JOURNAL OF ECONOMICS
between
and the costs of raw materials
consumed
gross product
in production.
In the past,
the Federal
Reserve
Board has re
formed
its industrial
index to reflect newer
value
production
so as to better
added data as new census data became
available
account
for relative
the emergence
of new prod
price changes,
structural
the lack of
ucts, and broader
However,
developments.
across
in scope,
and reliability
nine
concept,
correspondence
a less ambitious
industrial
necessitates
surveys
teenth-century
when
historical
in
series.
For
output
approach
constructing
and
Miron
and
Romer
each
[1947]
[1990]
stance, Frickey
adopt
census
as their sole value-added
the 1899 U. S. manufacturing
base year in compiling
their fixed-weight
indexes.
postbellum
census
In deciding
the
that
would
the
dictate
upon
year(s)
new
I
two
relative
of
considered
my
index,
importance
principal
as accurately
as
I wished
to model
Most
objectives.
importantly,
structure
the industrial
of the pre-Civil War
economy
possible
no annual measure
I
because
exists for this period. Furthermore,
to account
out
wished
for the evolving
of
industrial
composition
from structural
the Civil War
put that resulted
changes between
I. Indeed,
distinct
antebellum
and World War
and postbellum
of additional
base years permit
the incorporation
manufactures
not commercially
that were
before
the
and minerals
produced
in 1851,
enumeration
of the 1850 Census
(i.e., pocket watches
in 1852, zinc mining
in 1858, petro
fire engines
steam-powered
steel in 1866). This supplementation
leum in 1859, and Bessemer
basket
of goods with
addi
allows us to update
the antebellum
industries
for the postbellum
tional high-growth
period, yet does
com
criteria or historical
not fundamentally
violate
the selection
of the index.
parability
I selected
the 1850 U. S. Census
For a variety
of reasons,
as the basis for tabulating
War
value-added
weights,
pre-Civil
for the postbellum
value-added
and the 1880 U. S. Census
economic
historians
Most
component
importantly,
weights.
as
the
first
Census
reliable
view
1850
the
adequately
widely
con
decennial
Earlier
industrial
antebellum
surveys
survey.
and
1840
McLane
in 1810,
1832
ducted
(the
1820,
Report),
were
detail.
and lacked
industrial
underenumerated
severely
to its pre
out in comparison
stands
the 1850 Census
Indeed,
Office
the U. S. Census
because
decessors
adopted
primarily
more
methods
that
and
innovative
survey
vastly
sophisticated
returns.
Not only did the federal
industrial
govern
improved
but it
enumerators
for each firm reported,
ment
compensate
taxa
firms fearing
southern
firms
also reassured
(particularly
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1187
and even imposed
of their responses
tion) of the confidentiality
on establishments
to participate.
The
that refused
penalties
of the index's
the midpoint
1850 census
year
approximates
which
the fa
should mitigate
(1790-1915),
125-year
sample
of a Laspeyres
index. Finally,
characteristic
miliar
growth-bias
the
industrial
ized
1966].
through
Gallman
of 1850 and
surveys
revisions
the seminal
reworked
official
1880
have
been
standard
of Gallman
census
schedules
to produce
greater
industry
through
classifications
industrial
standard
two-digit
censuses
that Gallman
decennial
improved
are the most
and
1880 returns
complete
1880
1960,
[1956,
from 1840
across
comparability
(SIC). Of the five
upon, the 1850 and
at the
comparable
level.
industry
II lists
Table
in the
and mining
series
the manufacturing
index
and
their
relative
industrial
maps
importance.
production
series with
its
The first set of columns pairs each physical-volume
1850 value-added
sus and Gallman's
umns
documents
weight
adapted
subsequent
the 1880-base
from the
primarily
The second
revisions.
weights.
Following
1850 Cen
set of col
im
Frickey's
in Table
II
the percentage
principle,
weights
puted-weighting
an industrial
within
have been distributed
group
proportionally
an
series.
individual
series
has
Otherwise,
multiple
containing
been allocated
the entire two-digit
Lumber
industry weight.
ship
are assigned
for instance,
the value added in the lumber
ments,
The motivation
and wood products
for the imputed
industry.
to
is
excessive
top
assigning
weighting
principle
guard against
to those two-digit
line representation
industries
that are better
series
represented
by source data. In cases where
disaggregated
census
lack precise
with
classifications,
correspondence
industry
I have
the Federal
Reserve's
of prorating
adopted
approach
from auxiliary
weights
II. D.
Index
Table
production.6
information.
Construction
III presents
Physical
the new annual U. S. index of industrial
and min
output across the 43 manufacturing
6. On account
of expanded
data coverage
and an improved methodology,
the
III differ somewhat
index data
in Table
from previous
versions
unpublished
(e.g.,
Davis
distributed
via private
Research
[2002] and later variants
correspondence).
ers should
consider
in Table
III as the final and "correct"
the series
series.
QUARTERLY JOURNAL OF ECONOMICS
1188
TABLE II
A List
Major
of
industry
Quantity-based
Index
and
Components
index component
Products
Food & Kindred
Milled wheat flour
Refined sugar consumption
Hog packing
Beef packing
0.10 0.26
Salted mackerel
rice
Cleaned
Products
Textiles
& Textile
Cotton consumption
0.15 0.36
Wool stockings
0.09 0.06
Mixed cloth regalia
Raw silk imports
Lumber
& Wood Products
Lumber shipments
& Publishing
Printing
Relative
1850 weights
groups
Chemical
& Fuel Products
2.39 3.48
coal
Anthracite
Bituminous
coal
Sperm oil refining
Whale
oil refining
0.48 0.28
Salt production
and explosives
Gunpowder
chemicals 0.14 0.13
Dyeing
Whalebone
processing
Crude petroleum ? 1.90
& Accessories
Ordnance
Their
Importance
1880 weights
(%)
Series
Industry
Industry
11.02
6.36
1.24
0.87
0.78
4.78
0.08
0.03
0.41
0.32
0.05
0.02
8.23
1.28
0.24
0.34
Firearms
13.12
10.87
6.86
2.09
0.34
0.24
2.66
0.81
1.20
0.36
0.05
0.09
21.40
21.80
20.03
21.47
0.96
8.88
12.57
12.57
9.04
0.09
8.05
Products
Leather
& Leather
8.95 5.10
Sole leather
4.14
Leather hides
0.03
Boots and shoes, U. S. troops
& Metal
Products
Metals
8.13 7.33
Pig iron production
Gold
mining
Tinsmithing
0.47
Coppersmithing
0.21
Lead smelting
Die-sinking
0.06
Copper mining
? 1.61
steel
Bessemer
and open-hearth
? 0.17
Zinc production
& Machinery
Transport
Equipment
5.40
Merchant
ships
Locomotives
2.80
steel plows
Reaping machinery;
1.15
U. S. Navy vessels
0.13
Hand fire engines
? 0.15
Steam fire engines
Instruments
Musical
& Scientific
organs
Pipe
Telescopes
? 0.30
Pocket watches
8.88
Newspapers
8.04
13.12
8.05
9.04
2.93
0.01
13.07
12.93
1.30
0.85
0.61
2.66
1.72
0.26
0.12
0.44
0.07
13.10 14.02
2.70
3.62
5.88
0.58
0.01
4.71
1.16
0.85
0.77
0.66
0.19
Sources: Davis [2002] and Appendix E in Davis [2004a].
Components ranked by their relative importance in 1849/50 value added.
0.08
(%)
Series
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1189
TABLE III
A New
Year
1810
1811
1812
1813
U.
Index
S.
Index
of
Industrial
Year
Production,
Index
1790-1915
Year
4.291
1790
1832
31.53
4.490
1791
1833
35.15
1875284.2
4.881
1792
1834
33.58
1876294.0
5.441
1793
1835
37.57
1877297.8
6.014
1794
1836
40.25
1878
314.0
6.578
1795
1837
39.68
1879
356.4
6.699
1796
1838
40.70
1880
400.9
6.374
1797
6.213
1798
6.615
1799
1839
1840
1841
46.06
43.88
46.35
1881
478.1
1882
509.5
1883
523.8
1800
7.273
1842
47.66
1884505.3
1801
7.745
1843
53.10
1885
490.7
1802
8.210
1844
59.45
1886
550.5
1874300.7
1803
7.977
1845
65.36
1887604.7
8.077
1804
1846
75.93
1888656.7
8.698
1805
1847
87.37
1889
675.4
9.283
1806
1848
94.89
1890
781.0
9.718
1807
1849
98.33
1891793.1
8.063
1808
9.389
1809
10.258
11.299
11.252
12.063
1850
1851
1852
1853
1854
1855
102.39
106.05
122.84
139.98
142.72
143.44
Index
1892852.0
1893778.2
1894
722.0
1895
846.8
1896
820.9
1897874.5
1814
13.172
1856
149.44
1898
1,030.3
1815
13.584
1857
146.17
1899
1,129.7
1816
12.595
1858
137.54
1900
1,182.2
1817
13.119
1859
156.39
1901
1,277.1
1818
13.838
1860
157.86
1902
1,369.3
14.412
1819
1861
156.47
1903
1,420.3
14.834
1820
1862
167.35
1904
1,353.2
1821
15.792
1863
187.31
1905
1,565.6
1822
17.291
1864
200.61
1906
1,644.9
17.081
1823
1865
190.11
1907
1,713.3
1824
18.559
19.886
1825
1866
1867
193.91
210.05
1908
1909
1,446.5
1,705.5
1826
20.535
1868
220.90
1910
1,783.9
1827
21.073
1869
237.47
1911
1,717.9
1828
21.391
1870
242.97
1912
1,899.1
20.117
1829
1871
255.29
1913
1,975.0
23.801
1830
1872
275.74
1914
1,774.1
1831
28.085
1873
302.17
1915
1,954.2
Sources: See the text. Index base year is census year 1849/1850 = 100.
QUARTERLY JOURNAL OF ECONOMICS
1190
as an index num
and expressed
ing industries has been aggregated
ber from 1790 through 1915 with the base census year 1849/50.7
in several
The index was constructed
steps. First, each com
series ipit was
indexed by expressing
the physical
quan
ponent
of series
i for each year
to the
t, qit, relative
tity of output
for series
i in base year t = 0, qi0, equal to
produced
quantity
census year 1849/50. Next,
the individually
indexed components
were weighted
or value added vi0, in
by their relative
importance,
census year 1849/50
(see Table II) in order to arrive at an annual
a standard
industrial
fixed
production
Laspeyres
IPt, using
index formula:
weighted
(1)
=
IPt
^?=iiPit'vio
-~sn->
^i=lvi0
(Qit\
where
vi0
=
pi0
qi0
and
ipit
\Qi0/
=
[-?),
sum in (1) yielded a fixed-weighted
and where
the weighted
index
census
in
of industrial
based
1849/50.
year
production
on account of data attrition,
When
certain series disappeared
I followed
the index as if the
the standard
of computing
approach
in the missing
in the weighted
series equaled
the growth
growth
in those years. This sequential
of the remaining
series
average
to prevent
discontinuous
ratio splices
served
linkage
through
out of the sample.8 Other
indus
series dropped
jump-offs when
and new
industries
fire engines),
tries died
(e.g., hand-drawn
over the course of
or pocket watches)
(e.g., locomotives
emerged
a good was
a series disappeared
because
the sample period. When
a
not produced,
with
still
entered
the
index
the absent
good
zero.
a
of
obser
Likewise,
by
quantity
positive weight, multiplied
first produced were recorded,
vations
before a good was actually
as zero in the index. The indexing procedure
above
by definition,
census
was
the last
for the 1879/80
then repeated
year using
column
of value-added
weights
in Table
II.
7. Seventh
returns
embrace
Census
year
years. Census
parts of two calendar
to a census
refers
henceforth
1849/50
(two years,
year)
separated
by a slash,
over the twelve-month
June
to production
1, 1849. To
beginning
period
pertains
each calendar-year
value-added
with
1849/50
better
product
weights,
correspond
base using
the
in its final form to a census-year
converted
index was
equivalent
approximation:
geometric-mean
=
IPamno
((/Pnsso)5"2)-!,
L((/Pil849)"12)
was
i in year t. This procedure
of series
the
volume
where
represents
physical
IPit
for 1879/80.
repeated
in 1827, but
the index possesses
8. For instance,
coverage
industry
complete
in Romer
described
the procedure
in 1826. Following
loses three series
[1994] and
Reserve
of Governors
ofthe Federal
the U. S. Board
[1986], I ratio-spliced
System
the terminated
index excluding
in 1827 to a reweighted
index
the full-coverage
back the final index.
in 1826 in order to extend
series
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1191
in 1849/50
two overlapping
and
indexes based
Armed with
were
in
the
two
series
chro
1879/80 value-added
linked
weights,
in the final index of
The percentage
segments.
changes
nological
III reflect
index with
the 1849/50
base
the fixed-weight
Table
and
reflect
with
the
1879/80
other
from 1790 through
the
1850,
1915. For the intercensal
observations
base from 1879 through
of
from 1851 through
average
1878,1 used a linear time-weighted
was
of
two
This
indexes.
last
the annual percentage
step
changes
it more effectively
the emergence
of new
chosen because
captured
the 1850s than would have an arbitrary
industries
during
splic
series at the period's midpoint
centered
ing ofthe two overlapping
the end of the Civil War.9
The resulting
U. S. index covers the 1790-1915
period. The
with
the calendar year in which
the last of the
index commences
the Constitution.
With
the
thirteen
colonies
ratified
original
commerce
transferred
from those colonies
to
power to regulate
the American
the U. S. government,
"na
economy was officially
in 1790. The index has been carried
tional" beginning
through
in light ofthe
1915 (the standard
terminus
ofthe prewar period)
concerns
and comparability
that characterize
earlier
deficiencies
for the post-Civil War years, in particular
indexes constructed
the
index.
Frickey
in 1915 due primarily
index terminates
to data con
My
or degrade
on
straints.
Since several data sources are interrupted
or
onset
account
the
of
the
Great
of World War
I,
Depression,
in
I
and
bodies,
changes
reporting
procedures
data-collecting
could not reliably extend the annual
index past 1919, the year in
which
the Federal Reserve
Board's monthly
industrial
production
index begins. However,
for situations
where
for secu
controlling
lar U. S. economic
trends is necessary,
my index may be linked to
other data to create an annual
industrial
index that
production
runs from 1790 through 2000 and beyond. Specifically,
I suggest
new an
the following
the
First,
two-step
approach.
ratio-splice
nual index to the Miron-Romer
index of industrial
in
production
1916. Second,
in
this
1919
to
series
Federal
Re
the
ratio-splice
serve Board's
index of industrial
That said, I strenu
production.
=
9. Specifically,
let xj849/50
and x,1879/80 =
\n(IPj849/5?)
ln(IP}8\9/50)
the
rates
in the two
represent
ln(/PJ879/80)
growth
\n.{IP}^,S0)
logarithmic
indexes. We can then define
rate xftinal that pertains
the growth
to the
overlapping
=
+
index as a xftinal
industrial
(1 X w;,)jt*849/50
final
production
wtx}919/S0,
=
=
=
where
wt
((t
0W<1850,
wt
1W>1880,
and, to link the indexes,
wt
<
t <
1880.
the final industrial
index as
1850)/30)V1850
production
Thus,
in Table
III reflects
the accumulation
of xftlnal to IP^4.9/50
forward
from
presented
1850.
QUARTERLY JOURNAL OF ECONOMICS
1192
Av:
k&l
*.-
,
"' "'
.
^;*c*,^??
-;~.
Ilgf->-..
-^
;
"
2.0;
-^^
'
-'.'''..-. iwH? a??ss
ito nu/a^im'i?aWJ&*^^
Figure
Comparison
Sources:
with
See
the
Conventional
Postbellum
II
U.
S.
Industrial
Production
Indexes
text.
to base year 1899 and expressed
in logarithms.
series rescaled
For
Conventional
to the logarithm
1.50 has been
of the Miron-Romer
added
throughout
or Nutter,
3.25 to the Frickey
index.
index, and 4.00 to the Frickey-Leong,
clarity,
index,
tests on the business-cycle
that conducting
statistical
ously warn
in
structural
breaks
of this spliced
(i.e.,
properties
volatility)
series before and after 1915 is inappropriate
under any context.
series and data reliabil
This is primarily
because
the constituent
con
of
the
linked
industrial
indexes varies
three
production
ity
over time, and so, as Romer
[1986, 1989] has
siderably
in
evidence
of
breaks
structural
any potential
strated,
in
reflect
data
tended spliced series may
changes
simply
rather than genuine
regime shifts in the U. S. economy's
properties.
II.E.
Comparison
Production
with
Existing
Postbellum
demon
the ex
quality
cyclical
Industrial
Indexes
assess
II
its creditability
To visually
and reliability,
Figure
traces the new index alongside
three previously
post
published
of industrial
Civil War measures
the Frickey manu
production:
version
of the Frickey
index, an augmented
series, and
facturing
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1193
and Romer
the
recent Miron
the more
[1990] index.10 Although
toward raw materials
index is heavily
skewed
and
Miron-Romer
the monthly
series is noteworthy
because
semi-finished
products,
are
II con
its thirteen
unaltered.
components
essentially
Figure
in trend across the various measures.
the close agreement
new
not depart in any significant
index
the
does
way from
Indeed,
our conventional
secular development.
notions
of post-Civil War
veys
of uniformity
stems
in part from the fact that the
degree
common
share several
most notably
cotton
indexes
components,
and pig iron.
textiles
A perhaps
II is the new index's
less obvious feature of Figure
as
defined
the
lower year-to-year
standard
deviation
variation,
by
cor
in logarithmic
rates.
the
growth
strong growth-rate
Despite
relation
between
the new and old indexes,
the new
series
is
This
than its postbellum
less volatile
systematically
counterparts.11
in the new index is approximately
The severity
of cyclical swings
20 percent
lower than the Frickey
series and nearly
10 percent
lower than Nutter's
chain-linked
index. These
hold
margins
or not one includes
whether
the Civil War
Under
the
years.
that
variances
of
the
the
index
the
and
pre
present
hypothesis
constructed
data are identical,
traditional
variance
ratio
viously
are statistically
tests indicate
that the volatility
differences
sig
at the 10 percent
level with respect to the Frickey manu
index
but
level vis-a-vis
the
only at the 30 percent
facturing
Nutter
industrial
index.12
The reduced
in the new data
is noteworthy
and
volatility
over
the
because
older
series
has
been
criticized
for
reassuring
In particular,
Romer
fluctuations.
stating business-cycle
[1986, p.
are "exces
stresses
that
the
data
331]
Frickey manufacturing
on the order of 26 percent when
ex
sively volatile"
artificially
tended and compared with the post-WWII
Federal Reserve
Board
a similar volatility
index. Incidentally,
the Frickey
index disports
new
in
the
in
to
relation
index. In fact,
gap
my
postbellum
period
on account
this variance
differential
is likely understated
of sev
nificant
comes
10. The
index
from Nutter
affixes
augmented
Frickey
[1962], who
to the Frickey
data to better
[1950] index of mineral
Leong's
production
approxi
an industrial
mate
index.
I have
from other postbellum
abstracted
production
on grounds
indexes
of availability
and comparability.
See Davis
[2002] for details.
over the 1866-1914
11. For
rates
the Spearman
rank
log growth
period,
correlation
is 0.79 with
the Frickey
and 0.79 with
coefficient
series
the Nutter
series.
12. Typical
caveats
the Gaussian
ofthe
under
apply regarding
assumptions
lying distribution.
QUARTERLY JOURNAL OF ECONOMICS
1194
TABLE IV
Postbellum
Indexes:
Comparison
New
Value-added
base:
Panel
Final products
Intermediates
Raw
materials
Panel
annual
1849/50
A. Component
Component
index
1879/80
share
35.1
21.5
(%) of index,
Structure
by value
Miron-Romer
added
11.7
23.5
47.6
share
Market
Index: Frickey
1899
1899
34.7
17.8
43.4
B. Component
of
15.4
9.3
64.9
(%) of index,
by number
75.4
of series
Final products
55.3
53.5
25.0
Intermediates
13.2
14.0
15.0
7.7
31.6
32.6
60.0
69.2
Raw materials
23.1
Sources: Author's calculations based on information in Frickey [1947],Miron and Romer [1990], and
Davis [2002, 2004a]. Components classified according to historical Federal Reserve market groups as defined
in U. S. Board of Governors ofthe Federal Reserve System [1986].
eral artificially
in the Frickey
"smoothed"
components
present
and Nutter
series.13
in the new index can be attrib
The reduced volatility
observed
uted to the range of commodities
specified.14 To see why this is true,
the product mix of the new index can be contrasted with those of
a standard
constructed
indexes using
previously
post-Civil War
classification
the product mixes
scheme. Table IV summarizes
for
the new and previously
indexes based upon more con
constructed
temporary Federal Reserve market
groups.15 The critical feature of
Table IV is that the product sample of the new index differs quali
it is
from the conventional
series. Indeed,
tatively
post-Civil War
to conclude
reasonable
less
that the new index is systematically
volatile
the paper's
than the Frickey-Nutter
data precisely
because
database
is not replete with raw materials
and intermediate
prod
a focus on long-span data, a salient feature ofthe new
ucts. Despite
a share of final goods similar to
index is that the sample possesses
the currently released Federal Reserve
Taken
the new quantity-based
together,
industrial production
index.
a major
series represent
13. See Davis
details.
[2002] for further
are highly
in year
14. Unless
variations
differences
correlated,
component
in the number
of components.
could simply
reflect
differences
to-year
volatility
case.
new
and old
is not the
The
closer
reveals
that this
However,
inspection
consist
series from the 1870s
indexes
of approximately
40 annual
(minor attrition
across
alters
the exact count at any time).
samples
are admittedly
15. While
War
II industry
imperfect,
post-World
designations
a convenient
to
the Federal
standard
with which
Reserve
classifications
provide
across
evaluate
of processing
embodied
diverse manufactures.
the degree
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1195
TABLE V
New
Index
with
Consistent
A. Existing
Real
year
antebellum
Early
U.
S. Output
industrial
benchmarks
Census
Real
manufacturing
value added
Benchmarks
B. New
data
Industrial
Point
estimate
Range
Fragmentary
estimate
1810
(7.6-12.8)
9.8
10.52
1820
(9.2-15.6)
12.0
14.59
(33.9-46.0)
39.6
40.7
industrial
value
added
1830
production
index
21.58
1840
45.10
1845
100.0
1850
100.0
60.9
61.84
100.0
100.00
Sources: Davis [2002].
Figures were indexed to 1849/50 census-year equivalents using a geometric mean formula. Census years
1810 and 1840 encompass the calendar years 1809-1812 and 1840-1941, respectively, to accord with when
those surveys were actually conducted.
contribution
to the present
what
study because
they ameliorate
on primary commodities.
otherwise
be an overreliance
More
time-series
data cover many
of the
over, these freshly assembled
most sophisticated
and complex wares made during the nineteenth
cite many
ofthe new component
series
century. Economists
widely
in their case studies of the American
for
system of manufactures
their utilization
of a wide range of machine
tools, interchangeable
would
parts,
and
basic
materials.
III. Index
III.A. Antebellum
Industrial
The new index can
of early secular growth.
for
output benchmarks
ate the reliability
in the
secular trend of the new
Implications
Development
to our conventional
speak directly
At the same time, independently
the pre-Civil War era can serve
notions
derived
to evalu
trend of the index. Table V compares
the
index with
the best set of output bench
marks
that exist for the pre-Civil War period. The benchmarks
constant-dollar
industrial
estimates
from dis
represent
adapted
sources
and
all fig
parate
censuses;
incomplete
manufacturing
ures in Table V have been indexed
to the 1850 census year.
The wide ranges in the earliest output benchmarks
attest to the
considerable
America's
formative
economic
uncertainty
surrounding
The main
of
Table
is that
development.
implication
V, however,
QUARTERLY JOURNAL OF ECONOMICS
1196
on the secular development
conventional
views
of the American
are
newer
with
set.
consistent
the
data
economy
generally
The index also tells a more nuanced
story involving
impor
tant events
in the early American
economy,
including
exactly
a watershed
the antebellum
when
U. S. economy
experienced
in its long-run
rate of output growth. Considerable
acceleration
still surrounds
the approximate
"take
date of America's
debate
off and its emergence
from industrial
Given
that
pre
infancy.
at only
1840 estimates
of real output
per capita are available
of a secular
the timing and relative magnitude
intervals,
ten-year
have been difficult
to ascertain.
Rostow's
acceleration
[1971] take
that U. S. industrial
stipulates
productivity
o/f hypothesis
growth
of the economy during
accelerated
only with the "railroadization"
the 1840s and early 1850s. Conversely,
other scholars
(e.g., David
that industrial
[1994]) contend
[1967], Sokoloff
[1986], and Weiss
a decade or so earlier with the
may have accelerated
productivity
of
the factory system.
increasing
adoption
a crude proxy for
III presents
the new index, Figure
Using
rates in
industrial
growth
growth: average five-year
productivity
An
annual
industrial
per capita.
interesting
implica
production
of both
III is that the new data lend defensibility
tion of Figure
were
most
in
while
the
That
"takeoff
debate.
camps
is,
gains
1840s
and
accentuated
the
1850s, productivity
gains were
during
sense.
in the Rostovian
not unique
to this period
to have
in twin
appears
surged
Productivity
growth
to the new index,
before the Civil War. According
peaked waves
in the 1820s
trend occurred
the first significant
surge above
revised
1830s. The first peak agrees with Weiss's
and early
U. S. GDP. The second,
of decennial
estimates
trend-growth
more
boom
graphi
(point B) coincides
productivity
vigorous
the timing
of Rostow's
although
takeoff hypothesis,
cally with
was
that the magnitude
of its acceleration
indicates
the figure
the
And
not entirely
yet despite
twin-peaked
unprecedented.
advanced
before
booms witnessed
1860, industrial
productivity
War.
Civil
the
but pal
at a more
Slight
rapid pace following
the
War
and
after
Civil
before
differentials
pable growth-rate
of the nineteenth
the progressive
industrialization
validate
U.
century
III.B.
S. macroeconomy.
Antebellum-Postbellum
Comparisons
in Industrial
Volatility
The
answered
new
with
index
also
respect
to be
unresolved
questions
permits
nine
of
the
to the cyclical
properties
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
o.U
(A) Shift in production A
from artisan shops I
/
\
to factories accelerates
0 A (B) Rostow's
.
(1971) "take-off* of 1840s & 1850s
/ A
/ \
Pre-Civil Warmean productivity growth,
J800-1860 (2.2 % per annum)
2Q
1197
\ I
V]
Post-Civil War mean productivity growth,
1866-1910 (2.8 % per annum)
' lMlllllllllllllllllllllllllllltllllllllllll.il.
nlll.llllllllllllMllllllllMllllll.lHlllllllll.ullllllll.il
18001805 18101815182018251830 1835184018451850185518601865 18701875 18801885 18901895 .190019051910
Figure
Waves
in Nineteenth-Century
U.
III
S.
Industrial
Productivity
Plot
centered
averages
represents
running
five-year
moving
in per capita
rates
annualized
industrial
log growth
production.
on
a
is defined
basis
because
annual
labor-force
per
capita
unavailable.
Growth
of five-year
Productivity
are
estimates
U.
S. economy.
did significant
teenth-century
Specifically,
in
occur
before and
aggregate
changes
business-cycle
volatility
after the Civil War? Due primarily
to a lack of reliable
output
antecedent
studies
that have
looked
into the issue of
data,
con
have reached
somewhat
antebellum-postbellum
volatility
conclusions.
James
the cyclical
that
[1993] finds
tradictory
in economic
over the course of
fluctuations
increased
severity
the nineteenth
the annual
century.
Unfortunately,
pre-Civil
War data set on which
are widely
James
bases his conclusions
considered
for business-cycle
In the
inappropriate
analysis.16
other study, Calomiris
and Hanes
differences
[1994] evaluate
an artificial
16. James
series of antebellum
GNP
[1993] constructs
by linking
"consensus"
estimates
to Gallman's
(1790-1833)
(1834
Berry's
extrapolations
1859 census
then
tests
James
whether
annual
in
these
years).
changes
spliced
data differ
revised
from
estimates
and
Balke
systematically
postbellum
(e.g.,
Gordon
the value
ofthe
[1989] and Romer
[1989]). Unfortunately,
spliced
Berry
Gallman
data
is quite
limited
in this context
and hence
casts considerable
doubt
on the reliability
of his conclusions.
1198
QUARTERLY JOURNAL OF ECONOMICS
in the volatility
of U. S. industrial
between
the
production
antebellum
and postbellum
(1840-1859)
(1870-1913)
years.
From a set of six consistently
defined manufacturing
and min
con
Calomiris
and Hanes
ing products,
[p. 420] tentatively
were
clude that "cyclical movements
in industrial
production
no larger, and were probably
in the postbellum
smaller,
period
than in the last two decades
of the antebellum
period."
in
their methodology
direct
however,
By design,
precludes
ference
of antebellum-postbellum
because
volatility
changes
Calomiris
construct
and Hanes
the antebellum
[1994] artificially
data that best replicate
in
Frickey's
postbellum
manufacturing
are imposed upon
dex. Since postbellum
productive
relationships
the antebellum
that it is unclear
[1998] remarks
economy, Dick
or more
whether
inclusive
industrial
structural
data,
changes,
or nullify
would
the outcome.17
validate
Indeed, even Calomiris
and Hanes
the initial
[p. 410] stress that their study represents
toward
"an
annual
series
both
the ante
step
building
covering
com
bellum
and postbellum
consistent
and
years through
1914,
of testing whether
for the primary
purpose
parable
throughout,"
from
the antebellum
to
the amplitude
of business
cycles changed
the postbellum
period.
The new index allows us to examine with more confidence
the
of
of
U.
S.
annual
fluctuations
industrial
production
properties
over the 1790-1915
interest
is a comparison
period. Of particular
of industrial
before and after the Civil War, which we
volatility
can confidently
treat as a known break-point
given that the Civil
a "source of pro
as
War was,
Williamson
[1974, p. 5] describes,
several
and
found
change
along
disequilibrium"
regime
dimensions.
of annual
statistics
Table VI presents
summary
logarithmic
new
a
As
baseline
rates
in
index.
the
case, I have defined
growth
as the standard
in index loga
deviation
business-cycle
volatility
to detrending
business
rithmic growth rates. Contrary
methods,
are
rates
because
movements
measured
growth
precisely
cycle
in
the
level
of
and
contractions
indus
absolute
map
expansions
test statistics
and affiliated
Table VI presents
trial production.
mean
on
of the
variance
the
and
the
null
that
-values
p
hypothesis
the antebellum
and
rates are the same between
index's growth
devia
and Hanes
17. Calomiris
[1994]
regress
trend-adjusted
individually
is
in the Frickey
deviations
in the six series on trend-adjusted
tions
index, which
act as
on 1899 value-added
The subsequent
coefficients
based
regression
weights.
a surrogate
in an artificial
of Frickey's
for value-added
extrapolation
weights
antebellum
index over the twenty-year
sample.
TABLE VI
Antebellum-Postbellum
Index
Volatility
Comparison
Equal
m
hypothe
Antebellum
Index
comparison
1791-1860 vs. 1866-1915
(excludes War of 1812)
Panel
A. Logarithmic
s.d.
mean
6.64
5.18
Panel
1791-1860 vs. 1866-1915
(includes War of 1812)
1800-1849 vs. 1850-1899
(19th century only)
index
Years with all series
Series with all years
Calomiris-Hanes
(Replication)
Calomiris-Hanes
(Extension)
(A)
(B)
(2 variants)
?j
s.d.
mean
s.d.
mean
s.d.
mean
s.d.
mean
growth
B. Alternative
C. Alternative
7.35
5.87
7.06
5.08
14.94
6.25
10.90
T-test
period
6.50
5.15
6.71
5.40
s.d.
mean
s.d.
mean
Panel
Attrition-free
Postbellum
period
benchmark
rates,
sam
0.40
7.39
4.66
sample
^
periods
7.39
0.38
6.59
0.39
4.66
4.88
index
6.70
5.82
7.05
4.99
10.97
10.95
construction
<
0.03
S
0.06
?
(0.08)
6.52 ^
0.00
6.19
Unless otherwise noted, summary statistics represent log first differences of index, expressed in percentages.
CO
1200
QUARTERLY JOURNAL OF ECONOMICS
The Brown-Forsythe
median
postbellum
periods.
equality-of
test is a more robust test than traditional
variance
variance-ratio
tests when
the distribution
is independent
but not identically
in Panel A.
The benchmark
is presented
distributed.18
sample
case includes
I observations
The benchmark
all pre-World War
save for those traversing
of 1812 as
the Civil War
and the War
of major
armed
conflicts
soil) are
periods
(fought on domestic
as atypical
occurrences.
treated
Alternative
(in
sample
periods
Panel B) and index construction
methods
(in Panel C) are consid
ered in order to assess
the robustness
of the results
in Panel A.
to conclude
from Table VI that fluctua
It is quite reasonable
were
not markedly
tions in U. S. industrial
different
production
in the periods
This
outcome
before
and after
the Civil War.
in
to
resilient
tenable
appears
changes
sample periods, volatility
and data coverage. When
antebellum
deviations
measurements,
are compared with
their postbellum
the
variance
counterparts,
to
in
VI
of
tests
Table
fail
the
reject
hypothesis
comparison
to
in
be
Standard
deviations
tend
volatility.
growth-rate
equality
lower in the pre-Civil War years, but antebellum-postbel
slightly
not are statistically
lum differences
significant.19
are robust to how one
As shown in Panel C, these conclusions
or data
index attrition
constructs
modest
the index. Potentially,
could have
implications.
important
business-cycle
adjustments
attrition would be a concern if the series
For instance,
component
back in time behave
that drop out of the index further
quite
over
than
constituents.
the
business
cycle
remaining
differently
index and the early
This
the case for the Frickey
is certainly
index.
industrial
of the Federal
Reserve
segments
production
Panel C presents
To explore this possibility,
variance-equal
to the benchmark
index in Table
ity tests for several alternatives
indexes have been com
two different
III. Specifically,
no-attrition
index considers
The first no-attrition
only those years
piled.
in
unaffected
the
final
index
(1827-1900)
component
by end-point
index draws volatility
The second no-attrition
attrition.
compari
series
sons from an index comprised
of only those disaggregate
commences
from 1800 or earlier. Conse
annual
whose
coverage
quently,
percent
70
series reflects
the latter no-attrition
approximately
in the final
the former
of the value-added
index, while
in the first
of normality
test rejects
the null hypothesis
18. A Shapiro-Wilk
a nonparametrie
level. However,
differenced
data at the 10 percent
significance
data.
in the first-differenced
runs test fails to detect
correlation
serial
of de
are obtained
in terms
is defined
when
19. Similar
results
volatility
a band-pass
filter.
data using
trended
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1201
for an abridged
value-added
reflects
coverage
sample.
complete
core
to the
results are insensitive
that the
Panel C demonstrates
ver
in the benchmark
embedded
minor
and adjustment
attrition
index displays
attrition-free
sion of the new index. Neither
sig
or cyclical volatility
in either trend growth
be
nificant
changes
tween the antebellum
and postbellum
periods.20
the results
of Calomiris
and Hanes
Panel C revisits
Finally,
in
that
industrial
who
conclude
cyclical swings
[1994],
tentatively
I
create a
in
War
first
fell
the
post-Civil
period.
likely
production
near replica of their artificial
index for antebellum
(1840-1859)
five of their six
and postbellum
(1870-1913)
samples by accessing
annual series using fixed 1880 value-added
(labeled Calo
weights
in Table VI).211
miris-Hanes
then link the same
(A) "replication"
in Section
described
series using the methodology
II,
to include
the analysis
the pre-1840
(labeled
period
extending
calculations
Calomiris-Hanes
(B) in Table VI). These
provide
a replication
some important
of the Calomiris
First,
insights.
in post
the only case of significant
reduction
Hanes
study yields
in annual fluctuations
The standard
deviations
bellum volatility.
a reduction
30 percent
after the Civil War,
fall approximately
at
15
the
level.
this
However,
percent
statistically
significant
not
does
of
broader
trends.
appear
representative
replication
one explicitly
allows for structural
When
change or for a longer
five annual
antebellum
representation
vanish.
cyclical moderation
is that the Calomiris-Hanes
of the value added
percent
III.C.
Ranking
the Severity
the indications
of
(as is done here),
A logical explanation
for this revision
accounts
for less than 30
group
new
in
the
index.22
specified
of Prewar
Business
Cycles
new index also allows us to quantify
the relative
severity
business
cycles both before and after the Civil War. This
reasons.
be of interest
for several
For one, most
of our
occurrence
views regarding
the
and magnitude
of prewar
are based on an NBER
recessions
annual peaks
chronology whose
The
of U. S.
should
current
results
obtain
for an "adjustment-free"
Similar
excludes
index, which
on a related
index components
whose
in-filled
annual
coverage
gaps were
for details).
(see the Data Appendix
proxy
21. The
for the Calomiris-Hanes
in Table VI in
group
replication"
"study
cludes data for five manufacturing
and mining
industries
(anthracite
coal, bitu
minous
the sixth series
in
coal, cotton
lead, and pig iron) but omits
consumption,
on account
the original
of reporting
in U. S. trade
(coffee imports)
study
changes
statistics.
Inclusion
not substantially
ofthe
alter the findings
since
imports would
coffee roasting
would
receive
by far the smallest
weight.
22. The sensitivity
of the replication
underscores
and Hanes
why Calomiris
that their results were
emphasized
preliminary.
20.
three
QUARTERLY JOURNAL OF ECONOMICS
1202
on nominal
and troughs rest heavily
commod
data?particularly
assessments
the
of contemporary
ob
ity prices?and
qualitative
servers
an
an
Davis
Since
of
(see
[2004b]).
important
advantage
index of industrial
is
that
it
traces
in
production
only changes
new
new
on
the
shed
index
relative
may
physical
quantities,
light
recession
it could be argued that business-cycle
Second,
severity.
in
did
fact
moderate
after the Civil War, yet just along
severity
a pre-Civil
dimensions
other than annual volatility.
For instance,
War economy
booms
and
characterized
small
busts
could
by
large
have gradually
shifted toward an economy with
and
booms
larger
smaller
busts.
Romer
and direct welfare
ap
[1994], one sensible
Following
a
of
is
the
absolute
business
contraction's
proximation
severity
over the business
in real output realized
decline
cycle. Indeed, we
can
quantify
a recession's
cumulative
output
loss
as
the
summa
industrial
and its
between
log differences
production
the sum taken over the years when
level, with
previous
peak
level. This
is the most
is below
that previous
peak
output
a recession's
to calculate
severity when
approach
straightforward
tion of the
using
annual
data.
in
index losses for contractions
Table VII ranks cumulative
in
descend
the antebellum
and postbellum
respectively,
periods,
that the data in Table VII are calcu
ing order of severity. Given
the sum of the devia
lated in logarithms,
output losses represent
level of industrial
from
the
tions
(in percentage
peak
points)
a
of Table VII
The
bottom
contraction.
given
production
during
mean
for
deviation
and
its
standard
loss
the
output
presents
as well as nonparametric
and postbellum
antebellum
recessions,
of
Under
the null hypothesis
test statistics.
rank-sum
Wilcoxon
for an antebellum
rank of severity
the Wilcoxon
test, the average
rank of severity
of a postbel
recession
should equal the average
from its mean-loss
test excludes
The Wilcoxon
lum recession.
recessions
the three marginal
(indi
pre-Civil War
comparisons
are subjacent
to the minimum
loss
losses
cated by *) whose
era. The motivation
for this so
for the post-Civil War
observed
ensure
that an average
called cutoff loss rule is to adequately
the econ
takes into account
cyclical severity
profile of antebellum
losses
all cumulative
by excluding
omy's secular industrialization
index
loss
the minimum
that do not exceed
(see Ro
postbellum
a
mer
similar approach).
[1994] for
test in Table VII does not reject the null hy
The Wilcoxon
activ
In terms of industrial
commensurate
of
severity.
pothesis
a
increased
of
contraction
the
approximately
generic
severity
ity,
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1203
TABLE VII
of American
Severity
Antebellum
Prewar
Recessions:
Cumulative
Postbellum
recessions
Cumulative
Peak
index
Trough
loss
1807
1796
1815
1856
1828
1839
1833
1802
1808
1798
1816
1858
1829
1840
1834
1803
1836
1837
1822
1823
1.22*
1811
1812
0.41*
Subsample
All Antebellum
Mean
Mean
Output
recessions
Cumulative
Peak
index
Trough
1892
1907
1913
1873
1883
1903
1910
22.12
13.77
11.05
10.50
6.14
4.84
4.57
4.51
Loss
1897
1908
1914
1875
1885
1904
1911
loss
29.97
17.38
11.79
10.83
10.13
4.85
3.77
1.43*
output
losses:
Summary
All
recessions
loss
output
*
loss (no
losses)
Wilcoxon
and
statistics
tests
significance
Postbellum
recessions
7.32
Mean
output
9.69
Mean
loss 12.67
output
statistic
rank-sum
equality-of-mean
loss
p -value
12.67
59.0
0.56
Cumulative loss represents sum of percentage-point shortfalls in the logarithm ofthe index between peak
and subsequent years below the peak. * losses indicate contractions that do not exceed the minimum
cumulative postbellum loss. Two Civil War cycles (1861 and 1865 troughs) are omitted, although their
inclusion would not meaningfully affect calculations, p-values are for the null hypothesis of no mean-loss
distribution change before and after the Civil War.
three percentage
business
points after the Civil War. However,
over the course
contractions
did not become
shallower
of the
nineteenth
century, nor did they become more uniform. Another
severe
feature
of Table VII is that the three most
interesting
losses
for
the
fall
under
the
output
postbellum
period
30-year
then con
(1886-1916)
sample
analyzed
by Romer
[1994], who
are only marginally
I recessions
more
cluded that pre-World War
severe than post-World War
II recessions.
In addition,
the new index may radically
change our views on
individual
American
to the NBER
cycles. According
the
recessions
of
1873-1878
and
business-cycle
chronology,
are the longest
on record and are typi
1839-1843
contractions
as periods
of profound
and defla
cally characterized
depression
tion. Yet some economic historians
Friedman
and
Schwartz
(e.g.,
business
1204
QUARTERLY JOURNAL OF ECONOMICS
that these episodes
[1963] and Temin
[1974]) have long suspected
on nominal
be an artifact
of the NBER's
reliance
may
heavy
in the selection
of the early reference
variables
years.
Table VII illustrates
in either
instance
that the quantity
based production
data display
to their
shallower
losses relative
assessment
the
press
qualitative
portrayed
by
contemporary
and wholesale
for the
price indexes. One plausible
explanation
be that the media
confused
commercial
crises
may
disparity
were
with
financial
because
latter
the
characteristic
of
ones,
and
(see
falling
commodity
security
prices
Kindleberger
Peter
Temin
and others
have
[2000]).
Indeed,
hypothesized
that the duration
and severity
of the cyclical
contractions
that
were
more
the 1839 and
followed
1873 panics
exceedingly
severe
in nominal
than real economic
data because
precipitate
deflation
for final
substituted
for
price
goods
essentially
marked
in production.
The new
declines
data
series
output
here
indicates
that downturns
did occur following
employed
the financial
crises of 1839 and 1873, but they were
shorter
lived than described
accounts.
by contemporary
My assessment
not only by the mild depth of these recessions,
is bolstered
but
set
also their breadth:
All 43 index components
in my data
reach their trough before
the official NBER
the
troughs
during
1839 and 1873 downturns.
IV. Conclusion
This
annual
study was motivated
by the lack of a reliable
measure
of nineteenth-century
American
economic
As
activity.
a significant
the so-called
"statistical
step toward
rectifying
an annual
of industrial
this paper
index
offers
dark
age,"
I U. S. econ
War
that spans
the entire pre-World
production
that of the Federal
Reserve
omy. Index construction
parallels
in assembling
base data on
series
Board's
canonical
post-1919
from official
and pri
and minerals
43 distinct
manufactures
In light of the criticisms
vate sources.
[1986,
lodged by Romer
macroeco
constructed
and others
1989]
against
previously
War
has been
series
for the post-Civil
nomic
years,
primacy
ensure
on long-span
index consis
to reasonably
data
placed
of the index is that its
attribute
the strongest
Indeed,
tency.
are
in
De
entirely
components
physical
quantities.
expressed
is
not
index
the
selection
criteria,
overpopulated
spite rigorous
elsewhere.
with
basic
commodities
conveniently
published
unearthed
from
and
recently
Rather,
ignored
compilations
U. S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1205
than two dozen new phys
have yielded more
materials
a broad
of con
that encompass
series
spectrum
ical-output
sumer goods and industrial
machinery.
of an annual measure
of prewar
indus
This study's provision
should be of keen interest to economic historians
and
trial activity
our
of
Since
American
pro
analysts.
knowledge
business-cycle
lim
is severely
duction before
(and, indeed, after) the Civil War
new
this
index
of
industrial
annual
may provide
ited,
production
economists
and historians
into old questions,
alike new insights
to be answered.
and should even allow new questions
As just an
were
as the
of
events
what
the
such
example,
impacts
profound
on
of
of
the
War
and
the
Civil
War
Ameri
1807,
1812,
Embargo
source
industrialization?
ca's path toward
for our limited under
The new index also has implications
on
and
the
duration
of prewar
magnitude,
frequency,
standing
basis and across eras.
U. S. business
cycles, both on an individual
data have dulled
the effi
Scanty
pre-Civil War macroeconomic
studies
that examine
in
differences
cacy of antecedent
possible
annual
before and after the watershed
events
of the
volatility
of the new index is that antebellum
Civil War. One implication
are not statistically
in volatility
differences
postbellum
signifi
as the sample
cant when
is treated
the Civil War
break. This
se
in the average
that the differences
study also demonstrates
are
the
before and after
Civil War
immate
verity of contractions
that pre-World War I contractions
did not
rial, a strong indication
or more uniform
over time. Going forward,
become shallower
this
new index could be used to create an alternative
set of business
to the NBER
and troughs
How would
cycle peaks
chronology.
these new data alter our comparisons
with
the characteristics
of
recessions
and expansions?
postwar
Appendix
1: Brief
Description
of the
Index Components
This appendix
(in alphabetical
order) the
briefly documents
initial coverage,
units, and source types of the 43 phys
physical
ical-volume
series underlying
the new U. S. industrial
production
index. The reader is referred
to the unpublished
Tech
companion
nical Data Appendix
available
from the author
[Davis 2004a],
for more
details
and a complete
list of
upon request,
copious
the reader can assume
that no
otherwise,
were
to
the
made
time-series
data. In all
significant
adjustments
or
sources
were
data
checked
for
cases,
separate
overlapping
errors.
and
consistency
transcription
citations.
Unless
stated
QUARTERLY JOURNAL OF ECONOMICS
1206
1: Anthracite
Series
coal
Initial Coverage:
1790
Details: Direct measure.
anthracite
coal, in net
Pennsylvania
sources and independent
U. S. government
tons, from published
research.
coverage.
industry
Comprehensive
2: Army
boots
& shoes
Series
Initial Coverage:
1808
Details:
Pairs
of leather
boots and shoes
Direct measure.
made
contractors
those on the putting-out
by private
(including
and at U. S. Quartermaster
for the U. S. Army.
depots
system)
Author's
tabulations
from U. S. government
archives.
Compre
hensive
coverage.
industry
3: Beef
Series
cattle
receipts
Initial Coverage:
1827
Details:
Head of beef cattle received during
Indirect measure.
the calendar year at Brighton market
and at Chicago
stockyards.
Author's
from contemporary
trade jour
tabulations
newspapers,
research.
nals, and published
4: Bituminous
coal
Series
Initial Coverage:
1790
In net tons, from sources identical
to
Details:
Direct measure.
coal. Quantities
have been extended
anthracite
back from 1800
1790 by ratio-splicing
national
states, a fair linkage since
through
states
for less than 3 percent
of
the three absentee
accounted
1800 national
output.
5: Cloth
Series
regalia
Initial Coverage:
1808
Units
of wool and silk regalia made
Details:
Direct measure.
on the putting-out
contractors
those
sys
(including
by private
tab
Author's
and
federal
factories,
tem), private
clothing depots.
indus
from U. S. government
archives.
ulations
Comprehensive
try
coverage.
6: Copper
Series
consumption
1806
Initial Coverage:
Domestic
smelter
Details:
Indirect measure.
output,
plus
from all British
copper exported
ports,
imports of all unwrought
and U. S. govern
from British
in long tons. Author's
tabulations
ment
records.
7: Copper
smelting
in
mined
1790 (Product first commercially
Coverage:
on a large scale in 1845; earlier observations
States
the United
as zero in the index).
are recorded,
by definition,
measure.
recoverable
Smelter
Details:
Direct
production,
Series
Initial
U. S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1207
in short tons, obtained
from U. S. government
publica
content,
tions. Complete
coverage.
industry
8: Cotton
Series
consumption
1790
Initial Coverage:
The production
of cotton textiles
Indirect measure.
Details:
the con
items is quantified
and apparel
conventionally
through
over
raw
twelve-month
cotton
linters
the
and
of
period
sumption
mar
in August
when
the cotton crop was predominantly
ending
are expressed
in equivalent
bales
keted. Quantities
500-pound
as reported
and
Bureau,
by the U. S. Census
(gross weight)
at textile mills
and by households
account
for cotton consumed
cotton consump
contract
under the putting-out
system. Annual
as early as 1790, with
cov
are available
continuous
tion figures
in 1826. Coverage
gaps before 1826 have been
erage commencing
on the domestic
to manufac
cotton supply marketed
interpolated
in the estimated
is high for sev
observations
tures. Confidence
of
the fact that approximately
99 percent
eral reasons,
including
as ginned
in the specified
cotton grown
crop year was returned
rates in the extended
after the marketing
year, and that growth
the establishment
and output
of New
series track closely with
cotton
mills.
See
Davis
for
further
details.
textile
[2004a]
England
9: Crude
tin imports
Series
1815
Initial Coverage:
tin from mines
of the
Details:
Indirect measure.
Unwrought
and
United Kingdom,
British
colonies,
foreign countries,
exported
to the United
from all British
States by all vessels
ports, in long
tons, from the Sessional
Papers.
10: Die-sinking
Series
1793
Initial Coverage:
Direct measure.
Details:
U. S. coin production
of all denomi
in grams
tabulations
nations,
(weight; not face value). Author's
from price guides, based on U. S. government
records and private
research.
Series
11: Dyeing
chemicals
Initial Coverage:
1790 (Product first commercially
produced
are recorded, by
in the United
in 1834; earlier observations
States
as zero in the index).
definition,
Direct measure.
Details:
of potash
Pounds
of prussiate
(po
tassium
made
Carter
&
and Henry
ferrocyanide)
by
Scattergood
Bower Chemical
Author's
tabulations
Manufacturing
Company.
from firm archives.
The Philadelphia
firm of Carter &
chemical
was the first and largest American
manufacturer
of
Scattergood
were
and
red
which
of
industrial
yellow
prussiate
potash,
dyeing
QUARTERLY JOURNAL OF ECONOMICS
1208
in calico printing,
utilized
agents
fabric-making,
blueprinting,
etc. Series possesses
bias.
survivorship
12: Farm machinery
Series
Initial Coverage:
1790 (Product first commercially
produced
are recorded, by
in 1833; earlier observations
in the United
States
as zero in the index).
definition,
ma
of reaping and harvesting
Units
Details: Direct measure.
rakers, mowers,
harvesters,
binders;
chinery,
droppers,
including
and steel plows. Author's
tabulations
from firm archives, published
Series
records the
firm case studies,
and private
correspondence.
manufacturers:
output of four pioneer and primary farm-implement
Harvester
International
and
Obed Hussey, McCormick,
Company,
bias.
John Deere. Series possesses
survivorship
13: Firearms
Series
1790 (Product first commercially
Initial Coverage:
produced
are recorded, by
in 1793; earlier observations
in the United
States
as zero in the index).
definition,
small arms
and commercial
Details:
Direct measure.
Military
and
(all models),
contractors,
by federal and state armories,
and
tabulations
from
firms.
Author's
published
unpub
private
firm archives,
and published
lished U. S. government
records,
firm studies. Gunsmiths
and firearm manufacturers
represented
of
for approximately
one-half
in the component
series account
total U. S. firearm production.
14: Fish
Series
curing
1804
Initial Coverage:
barrels
Direct measure.
Salted mackerel
Details:
inspected
as
in Massachusetts
(until 1877) and New England
(thereafter),
in
in
U.
S.
Nearly
complete
government
publications.
reported
made
dustry
coverage.
Series
Initial
15: Gold
Initial
Coverage:
mining
1804
Coverage:
in
Mined
Direct measure.
Details:
stage,
output at refinery
as reported
Com
in U. S. government
fine ounces,
publications.
coverage.
plete industry
Series
16: Gunpowder
1804
Initial Coverage:
and explo
of gunpowder
Pounds
Direct measure.
Details:
Powder
Nemours
du
Pont
de
of
E.I.
interests
sives produced
by
overstate
Series
the
firm
archives.
from
tabulated
Company,
bias.
and
secular
possess
survivorship
industry
growth
fire engines
17: Hand-operated
Series
1790
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1209
tabula
Units
constructed.
Author's
Direct measure.
Details:
more
builder
from various
than 2,000 fire engines
tions of
lists,
and fire depart
historical
archives,
society records, fire museum
the "death" of the domestic
The series captures
ment
histories.
an
estimated
two-thirds
of nineteenth
in
and
1914,
industry
century domestic
production.
18: Hide
Series
receipts
1827
Initial Coverage:
and curing. Re
Details:
Indirect proxy for leather
tanning
at New York
dried
&
hides
of
domestic
and
green
foreign
ceipts
centers
of the
the
and
leather-tanning
premiere
City
Chicago,
from
Author's
tabulations
nineteenth
contemporary
century.
and trade journals.
newspapers
19: Hog packing
Series
in
1790 (First commercial
made
Initial Coverage:
shipment
are recorded,
as zero in
1809; earlier observations
by definition,
the
index).
Details:
Direct
measure.
cinnati, Chicago,
Indianapolis,
from contemporary
newspapers,
research.
Minor
data
adjustments
in Cin
of hogs packed
Quantities
and Omaha. Author's
tabulations
trade journals,
and published
were
necessary.
20: Lead
Series
smelting
1821
Initial Coverage:
measure.
in
Details:
Direct
smelter
Primary
production,
as
in
short tons until
refined
1885;
output
thereafter,
reported
U. S. government
coverage.
industry
Complete
publications.
21: Locomotives
Series
Initial Coverage:
1790 (Product first commercially
produced
are recorded, by
in the United
in 1825; earlier observations
States
as zero in the index).
definition,
Details:
Direct measure.
tabulations
of more
than
Author's
from various
builder
120,000
lists, rail
engines manufactured
road historical
and published
rail
society records, firm archives,
For a complete
histories.
refer to Ap
road-company
description,
[2004a].
pendix B of Davis
22: Lumber
Series
shipments
Initial Coverage:
1827
Details:
Direct measure.
in feet board measure
Shipments
river booms,
seaside
and whole
(b.f.) from ten distinct
ports,
sale
districts
ber-producing
ulations
from
lished studies.
that
all of the principal
represent
virtually
of
the
nineteenth
regions
century. Author's
trade
and various
contemporary
journals
lum
tab
pub
QUARTERLY JOURNAL OF ECONOMICS
1210
23: Merchant
Series
shipbuilding
Initial Coverage:
1790
Details:
Gross
Direct measure.
(Old Custom House
tonnage
mer
Measurement
constructed
basis) of all types of domestically
chant rigs, including
the four major
classes
specialty
(clippers,
Author's
and whalers).
tabulations
from pub
steamers,
packets,
so
lished and unpublished
U. S. government
records, historical
and published
The new series rests
ciety archives,
ship registries.
on a database
or
of approximately
100,000 merchant
vessels,
more than two-thirds
of the American
merchant
fleet built before
I. For a complete
refer to Appendix
C of
World War
description,
Davis
[2004a].
24: Milled
Series
wheat
flour
Initial Coverage:
1798
or manufactured
in
Details:
Barrels
received
Direct measure.
and Minneapolis.
Author's
tabula
Baltimore,
Buffalo,
Chicago,
and various
tions from contemporary
trade journals
published
studies.
25: Naval
Series
shipbuilding
1790
Initial Coverage:
measure.
from U. S.
tabulations
Details:
Author's
Direct
of
U.
S. Navy
and
every
government
publications
ship registries
in
at private
and government
vessel
constructed
displace
yards,
ment
tonnage.
26: Newspaper
Series
publishing
1790
Initial Coverage:
in
of daily newspapers
Details:
Indirect measure.
Number
from numerous
tabulations
circulation.
Author's
bibliographic
lists and trade journals.
27: Petroleum
Series
refining
in
mined
1790 (Product first commercially
Initial Coverage:
are
observations
in
earlier
States
the United
recorded,
1859;
by
as zero in the index; component
receives
only 1880
definition,
in the index).
value-added
weight
in 42
Crude petroleum
Details:
Direct measure.
produced,
gallon
barrels,
from
series M-318
of Historical
Statistics
coverage.
industry
Complete
28: Pig
iron production
Series
1827
Initial Coverage:
Gross tons produced.
Details:
Direct measure.
and unpublished
from various
lations
published
see
D
of
the companion
Section
details,
complete
Appendix.
[1975].
Author's
tabu
sources.
For
Data
Technical
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
29: Pipe
Series
organs
1790
Initial Coverage:
Author's
Details:
Direct measure.
from various
22,000 units constructed
sources.
Comprehensive
Series
Initial
of more
than
and unpublished
coverage.
industry
30: Pocket
tabulations
published
1211
watches
in
mined
1790 (Product first commercially
Coverage:
are recorded,
in 1851; earlier observations
States
by
as zero in the index; component
receives
definition,
only 1880
in the index).
value-added
weight
Direct measure.
Author's
tabulations
of more
than
Details:
from various
80 percent
of movements
produced
unpublished
historical
studies.
society records and published
silk imports
31: Raw
Series
1814
Initial Coverage:
of silk consumption.
Indirect measure
Details:
Raw, thrown,
British
and waste
silk of the United
and
colonies,
Kingdom,
the United
countries
(including
all
vessels
by
from the Sessional
Papers.
32: Rice milling
Series
foreign
United
States
China
to the
and India), exported
in pounds,
all British
ports,
from
Initial Coverage:
1819
rice equivalent,
Details:
Direct measure.
Cleaned
rough rice
in
from
U.
S.
crop,
government
pounds,
adjusted
publications.
errors were corrected
in the U. S. Department
of Agricul
Several
ture
estimates.
33: Salt production
Series
Initial Coverage:
1797
Details:
Direct measure.
bushels
of pro
Inspected
56-pound
cessed salt (all types), at all New York salt wells and reservations,
and from all Michigan
salt producers.
Author's
tabulations
from
were
state government
records. New York
and Michigan
the
states during
salt-producing
34: Sole
leather
receipts
Initial Coverage:
1827
Details:
Direct measure.
Inspected
preeminent
Series
the nineteenth
century.
of sole leather
receipts
and
oak
sides,
sole,
sole,
including
sole, in New
York
to
Boston
Author's
tabulations
from
(prior
consignment).
re
and
trade
New
York
contemporary
reports
journals.
City's
a
measure
of
domestic
sole
leather
offer
reasonable
ceipts
heavy
hemlock
ofthe
because
output
New
union
of civilian
and other finished
York
the
shoes
City was
largest
leather
leather
market
products
at this
QUARTERLY JOURNAL OF ECONOMICS
1212
time,
and
because
boots
and
shoes.
sole
leather
was
the primary
component
in
35: Sperm
oil refining
1793
Coverage:
of sperm oil returned
to
Barrels
Details:
Indirect measure.
fleet. Author's
tabulations
from vari
port by American
whaling
ous published
sources. Nearly
universal
coverage.
industry
36: Steam
Series
fire engines
Series
Initial
1790 (Product first commercially
Initial Coverage:
produced
are recorded, by
in the United
in 1852; earlier observations
States
as zero in the index; component
receives
definition,
only 1880
in the index).
value-added
weight
in engine
Details:
Units
Direct measure.
delivered,
expressed
I
of
minute.
obtained
construction
and
per
speci
gallons
capacity
on over 4,000 engines
from builder
information
and fire
fication
records.
department
the birth and death
Series
The series
is comprehensive,
of the domestic
industry.
and
captures
37: Steel
production
in
1790 (Product first commercially
mined
Coverage:
are recorded,
in 1866; earlier observations
States
the United
by
as zero in the index; component
receives
definition,
only 1880
in the index).
value-added
weight
Thousands
of net tons produced
Direct measure.
Details:
as reported
in
Bessemer
and
processes,
through
open-hearth
Initial
universal
trade journals. Nearly
industry coverage.
contemporary
38: Sugar
Series
refining
1790
Initial Coverage:
of refined
Domestic
Indirect measure.
Details:
production
to pounds.
Author's
tabulations
converted
sugar consumption,
in
series
Related
S. government
spliced
publications.
1822. Nearly
universal
coverage.
industry
Series
39: Telescopes
1790 (Product first commercially
Initial Coverage:
produced
are recorded, by
in 1830; earlier observations
in the United
States
as zero in the index).
definition,
in inches
Refractors
and reflectors,
Details:
Direct measure.
and unpub
from published
Author's
tabulations
of objective.
from U.
lished
records
of historical
societies
hensive
coverage.
industry
40: Whalebone
Series
1804
Initial Coverage:
measure.
Indirect
Details:
and
Baleen
its members.
whalebone,
Compre
in pounds,
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1213
sources
to sperm oil refining.
identical
universal
Nearly
coverage.
industry
41: Whale
Series
oil refining
1793
Initial Coverage:
Details:
Indirect measure.
of whale
Barrels
oil returned
to
sources
American
from
to
identical
port by
sperm
fleet,
whaling
oil refining. Nearly
universal
coverage.
industry
42: Wool
Series
stockings
1808
Initial Coverage:
Details:
Direct measure.
Pairs of woolen
and half
stockings
ar
made.
Author's
tabulations
from
U.
S.
stockings
government
chives. Fairly
coverage.
comprehensive
industry
43: Zinc
Series
smelting
Initial Coverage:
1790 (Product first commercially
mined
in
are recorded,
the United
in 1858; earlier observations
States
by
as zero in the index; component
receives
definition,
only 1880
from
in the index).
value-added
weight
measure.
Details:
Direct
smelter
in
Primary
production,
short tons until
as
recoverable
content
1906; mine
thereafter,
in U. S. government
Minor
data correc
reported
publications.
tions. Complete
coverage.
industry
Appendix
2: Data
Limitations
its aforementioned
the new U. S. indus
Despite
strengths,
trial production
index possesses
definite
limitations
and potential
a higher
biases. To be sure, the index possesses
of mea
degree
surement
error
base
are not
vis-a-vis
modern-day
series
since
the
antiquated
or attrition.
devoid of adjustments
completely
series
that I have
created
Generally
speaking,
retrospectively
to survivorship
bias because
the base
may be most
susceptible
on extant records. Everything
data rest partially
else equal, ret
canvasses
will be less accurate
than a contemporane
rospective
ous producer
are
more
the latter
to cap
inclined
survey because
or since-defunct
ture smaller-scale
businesses.
the
Substituting
of a sample of firms for the output of an entire indus
production
error. Consequently,
to measurement
sur
try is also susceptible
bias
will
to
tend
understate
data volatility
because
vivorship
business
failures
rise during
recessions.
While
these
biases
on locomotives
should not apply
to the series
or navy vessels
of the raw data, it is unclear
given the completeness
exactly how
data
bias affects the novel data
survivorship
fire
farm machinery,
firearms,
engines,
sets for dyeing chemicals,
and gunpowder.
QUARTERLY JOURNAL OF ECONOMICS
1214
Data
attrition
could also potentially
distort
the historical
time series. In this case, however,
in the new
attrition
component
occurs at a slower and lower rate back in time
index actually
to the Frickey
relative
and Federal
Reserve
two well
indexes,
measure
more
known
industrial
indexes
that
recent
production
errors
economic
in
Occasional
and
data
gaps
activity.
reporting
also necessitated
the adjustment
to several
index com
coverage
arose
In many
from
cases, data adjustments
ponents.
simply
errors
observations.
due to typographical
conflicting
Presumably
or subsequent
source data did not always
revisions,
correspond
when
Most discrepan
cross-referencing
antiquated
publications.
cies were minor.
In the case of contradictions,
and
government
most
recent publications
received
preference.
were generally
While missing
observations
confined to one or
two years
Technical
for
Data Appendix
(again, see the companion
in three of the 43
specifics),
slightly
longer gaps were encountered
the miss
product series. I adopted a fairly strict rule in estimating
an
observations:
alternative
annual
series
available
for
the cov
ing
similar cyclical turning points and volatil
erage gap had to possess
to
in each of the three
the
data.
evidence
component
ity
Prevailing
cases indicated that an available
was
a
suitable
proxy
interpolator.
Investment
The
Vanguard
Counseling
Group,
and
Research
Department
Inc.
References
Na
of Prewar
Gross
J. Gordon,
Nathan
"The Estimation
S., and Robert
Econ
Journal
and New
tional Product:
Evidence,"
of Political
Methodology
38-92.
omy, XCVII
(1989),
GNP Series
since 1789: Revised
and Population
Thomas
Senior, Production
Berry,
VA: by the author,
Dollars
in Constant
1988).
(Richmond,
States
since 1870 (New York:
Trends
in the United
F., Production
Burns,
Arthur,
National
Bureau
of Economic
1934).
Research,
for the
"Consistent
Series
Charles W., and Christopher
Hanes,
Calomiris,
Output
Jour
and Preliminary
Periods:
Issues
Antebellum
and Postbellum
Results,"
LIV (1994),
409-422.
nal of Economic
History,
on a Statistical
Growth
Dark Age: U. S. Real Product
Paul A., "New Light
David,
LVII
294-306.
Economic
before
Review,
(1967),
1840," American
Index of U. S. Industrial
Annual
Production,
Davis,
Joseph H., "A Quantity-Based
Fluctua
of Historical
An Empirical
1790-1915:
Business-Cycle
Appraisal
2002.
Duke
Ph.D.
thesis,
University,
tions," unpublished
1790-1915:
"An Annual
Index of U. S. Industrial
-,
Production,
Companion
2004a.
Technical
Data Appendix,"
mimeo,
since
the 1790's,"
S.
Business
of
U.
"An Improved
Annual
-,
Cycles
Chronology
2004b.
mimeo,
International
since 1820: New
Perspectives
Dick, Trevor T. O., ed., Business
Cycles
MA: Edward
Evidence
1998).
Elgar,
(Northampton,
from Historical
1783
"U.
S.
Economic
E.
Robert
and
Growth,
Gallman,
L.,
Stanley
Engerman,
CT: JAI
8 (Greenwich,
in Research
in Economic
History?Volume
1860,"
Balke,
Press,
1983).
U S. INDUSTRIAL PRODUCTION ANNUAL INDEX
1215
American
Railroads
and the Transformation
Fishlow,
Albert,
of the Ante-Bellum
MA: Harvard
1965).
Economy
(Cambridge,
University
Press,
in the United
Production
1860-1914
MA:
States,
Edwin,
Frickey,
(Cambridge,
Harvard
1947).
Press,
University
and Anna J. Schwartz,
A Monetary
Friedman,
Milton,
States,
History
of the United
NJ: Princeton
1867-1960
1963).
(Princeton,
Press,
University
Robert
and Manufacturing
in
Gallman,
E., "Value Added
by Agriculture,
Mining,
the United
Ph.D.
of
States,
1840-1880,"
thesis,
unpublished
University
1956.
Pennsylvania,
in William
N. Parker,
in the
-,
1839-1899,"
"Commodity
Output,
ed., Trends
in the Nineteenth
American
NJ: Princeton
Uni
Economy
(Princeton,
Century
1960).
Press,
versity
"Gross National
in the United
Product
in Dorothy
-,
S.
States,
1834-1909,"
and Productivity
in the United
States
ed., Output,
Brady,
Employment,
after
1800
(New York: Columbia
1966).
Press,
University
"Trade Disruptions
and America's
Irwin, Douglas
A., and Joseph H. Davis,
Early
NBER Working
2003.
Industrialization,"
Paper No. 9944,
in Economic
John A.,
in 19th-century
James,
"Changes
Instability
America,"
American
LXXXIII
Economic
710-731.
Review,
(1993),
Charles
and Crashes:
A History
P., Manias,
Panics,
Kindleberger,
of Financial
Crises
& Sons,
(New York: John Wiley
2000).
Y.
"Index
of
the
Volume
Production
of
S.,
Leong,
Physical
Minerals,
1880-1948,"
Journal
Statistical
XLIV
15-29.
Association,
of the American
(1950),
and
Gabriel
Perez
Fluctuations
in the
M.,
McConnell,
Margaret
Quiros,
"Output
United
States: What
Has
since
the Early
1980s?" American
Eco
Changed
nomic Review,
XC (2000),
1464-1476.
D. Romer,
"A New Monthly
Index of Industrial
Miron,
A., and Christina
Jeffrey
Journal
L (1990),
321-337.
Production,
1884-1940,"
of Economic
History,
G. Warren,
Growth
Production
in the Soviet
Union
(Prince
Nutter,
of Industrial
1962).
ton, NJ: Princeton
Press,
University
Paul
"Gallman's
Annual
Series
for
the
United
1834
W.,
Rhode,
Output
States,
2002.
1909," NBER Working
Paper No. 8860,
Christina
a Figment
of the Postwar
of
Romer,
D., "Is the Stabilization
Economy
the Data?" American
Economic
LXXVI
314-334.
Review,
(1986),
"The Prewar
Business
New Estimates
of Gross National
-,
Cycle Reconsidered:
Journal
XCVII
1-37.
Product,
1869-1908,"
of Political
(1989),
Economy,
"The Cyclical
Behavior
of Individual
Production
-,
Series,
1889-1984,"
Quar
CVI (1991),
1?31.
of Economics,
terly Journal
Business
Journal
LIV
-,
"Remeasuring
Cycles,"
of Economic
(1994),
History,
573-609.
Walt W., The Stages
A Non-Communist
Growth:
Rostow,
of Economic
Manifesto
1971).
(Cambridge:
Cambridge
Press,
University
Kenneth
in Manufacturing
Growth
In
Sokoloff,
L., "Productivity
during
Early
dustrialization:
Evidence
from the American
in Stan
Northeast,
1820-1860,"
L.
and
Robert
E.
Factors
in Ameri
ley
Engerman
Gallman,
eds., Long-Term
can Economic
Growth
of Chicago
1986).
(Chicago: University
Press,
James
and
Mark
W.
"Has
the
Business
and
Stock,
H.,
Watson,
Cycle
Changed
Macroeconomics
Annual
MA: MIT
Why?" NBER
(Cambridge,
Press,
2002),
159-218.
pp.
"The Anglo-American
Business
Economic
His
Temin,
Peter,
Cycle,
1820-1860,"
XXVII
207-221.
tory Review,
(1974),
United
House
of Commons,
Sessional
(New York:
Parliament,
Kingdom,
Papers
Readex
various
issues).
microprint,
U. S. Board
of Governors
of the Federal
Reserve
Federal
Reserve
Bulletin
System,
DC: various
issues).
(Washington,
Industrial
Production:
1986 Edition,
with a Description
-,
of the Methodology
DC:
1986).
(Washington,
U. S. Department
of Commerce,
Bureau
of Census,
Statistics
Historical
of the
United
Colonial
to 1970
Times
DC: 1975).
States,
(Washington,
"Economic
Growth
1860: Revised
before
in Thomas
Weiss,
Thomas,
Conjectures,"
Weiss
and Donald
Economic
inHistori
Schaefer,
eds., American
Development
cal Perspective
CA: Stanford
(Stanford,
1994).
Press,
University
American
A Gen
Williamson,
G., Late Nineteenth-Century
Jeffrey
Development:
eral Equilibrium
(New York: Cambridge
1974).
History
Press,
University