Statistical Policy
S.I.C. PURSUITS: THE CONSEQUENCES AND
PROBLEMS OF CLASSIFYING ESTABLISHMENTS FOR GOVERNMENT
STATISTIC
Rolf R. Schmitt and Michael Rossetti
Representatives of the U.S. Department of Transportation
to the Technical Committee on Industrial Classification,
which was convened by the Office of Management and Budget to
revise the SIC system.
(This paper was published in the Proceedings of the Urban
and Regional Information Systems Association: 1987, Volume
IV, pp. 15-24.)
ABSTRACT. The Standard Industrial Classification
(SIC) system provides a uniform method of classifying
establishments for the collection of Federal data. SIC codes
affect the extent and detail of data collection activities
and of tabulations, which in turn have subtle yet pervasive
affects on the way in which we perceive the national and
regional economies and related public policy issues.
The 1987 revisions to the SIC system are the first major
classification changes to be made in 15 years. The revision
process involved a delicate balance of Federal, State,
local, and industry interests while bringing the system more
closely in line with the economy.
This paper provides a brief overview and critique of the
SIC system to encourage better provision and use of economic
data. The paper illustrates how institutional and
statistical considerations can be far more important than
data processing to the quality and quantity of information
for public decision making.
Introduction
The Standard Industrial Classification (SIC) system
provides a detailed, comprehensive approach to classifying
economic activity, covering activities from wheat farms and
automobile assembly plants to airport runway vacuuming and
phrenologists. While the SIC Manual that describes the
system in excruciating detail is dry and tedious for all
except classification experts and connoisseurs of Trivial
Pursuit, it documents one of the most widely used, least
understood standards for government statistics. The
philosophy and details of SIC system must be thoroughly
understood, however, if data on economic establishments are
to be collected and used effectively.
This paper provides a brief overview of several issues
raised during the recent revision of the SIC system so that
providers and users of economic data can better understand
the strengths and weaknesses of the information they
generate for public decision making. This brief critique
assumes that the reader will consult the SIC Manual for a
more thorough understanding of SIC concepts and definitions
(1).
Basic SIC Concepts
The SIC system is designed to provide a comprehensive,
detailed framework for the collection and tabulation of
establishment-based statistics by the Federal Government.
Establishments can be a one location company or an
individual facility of a nationwide corporation.
Establishments are classified into Industries by the
products made or the services rendered. Industries are
identified by a four-digit number, and are aggregated into
three-digit Industry Groups, two-digit Major Groups, and
Divisions that are identified by a single letter. The
definition of each Industry is typically accompanied by one
or more examples, called index items, that are used for
clarification and illustrative purposes. For example,
manufacturers of PCV valves are listed under Industry 3714
(Motor Vehicle Parts and Accessories), which is part of
Industry Group 371 (Motor Vehicles and Motor Vehicle
Equipment), which is part of Major Group 37 (Transportation
Equipment), which is part of Division D (Manufacturing).
Some agencies have an additional level of detail for
kinds of business or specific products manufactured by the
Industry. The fifth digit is not part of the SIC standard
(except in the case of Auxiliaries discussed later), and
often misleads data users to think that the SIC system is
based on commodities rather than on economic activity. Other
governmentwide classification systems exist for commodities.
An additional point of confusion is the distinction
between establishments and enterprises. Enterprises are
companies and are defined by ownership; establishments are
essentially facilities and are defined by economic activity.
While most enterprises consist of one establishment, the
larger enterprises own many establishments that are
classified in several different Industries. As a
consequence, enterprise statistics such as those based on
tax returns are collected and tabulated under a separate
classification system.
Industries are officially recognized in the SIC system
when establishments that perform similar economic activity
collectively meet size, specialization, and other criteria.
Some Industries are defined to include the miscellaneous
establishments that do not fit elsewhere in the Industry
Group or Major Group, so that every establishment can be
assigned to an Industry. These residual Industries often
contain a stunning diversity of activities. For example,
Industry 4789 (Transportation Services, Not Elsewhere
Classified) includes among other activities operators of
horse-drawn for-hire carriages and private establishments
primarily engaged in the operation of space flights.
Importance of the SIC System
SIC definitions of Industries and establishments are used
by agencies throughout the Federal Government to identify
the scope of economic censuses and surveys and to tabulate
the results. The definitions can affect the data bases that
are used to calculate the Consumer Price Index and other
measures that are used by all levels of government. Since
the statistics based on the SIC system are widely reported
by the press and used by public decision makers, this
detailed classification system has subtle yet pervasive
affects on our perception of the Nation's economic structure
and health.
The importance of the SIC has grown in recent years for several reasons:
- SIC-based statistical measures designed for policy
research and planning are being used increasingly for
other purposes that have direct impacts on the public. An
establishment's classification can determine its
eligibility for certain Federal benefits, its tax
liability, and its liability for health and other
regulations.
- Many agencies are turning to SIC-based surveys to
replace data previously collected by regulatory agencies
using other classification schemes.
- Agencies at all levels of government responsible for
economic, social, transportation, and other programs are
performing more frequent and sophisticated economic
analyses of their activities, requiring greater amounts
of SIC-based economic data. Trade organizations and
others in the private sector are also making greater
demands on SIC-based data to support the advocacy of
their causes before public officials.
- The popular press now regularly reports statistics
and findings that are based either directly or indirectly
on the SIC system.
In short, the SIC system is providing a framework for
much of the popular perception of economic issues, the
ensuing political debate, and the subsequent government
actions.
The SIC Revision
Until 1987 SIC revisions were implemented, the growing
importance of the SIC system was in stark contrast to the
system's declining representativeness of the Nation's
economy. The SIC system had not been revised significantly
since 1972, when the regulatory environment, technology,
domestic and international markets, corporate structure, and
other aspects of industry organization were radically
different from today. Revisions to the SIC system in 1977
were minor, and an effort to overhaul the system in 1982
collapsed due to budget pressures.
To correct the growing disparity of the SIC system and
the economic structure that it was supposed to reflect, the
Office of Management and Budget (OMB) convened the Technical
Committee on Industrial Classification (TCIC). The TCIC
included representatives from the major data collection
agencies of the Federal Government, as well as from a
healthy cross section of data consuming agencies. The TCIC
met on a weekly or more frequent basis from 1984 to 1986 to
solicit and review public comments. Over 1,100 changes to
the SIC Manual were proposed by government agencies, trade
associations, businesses, and individuals in response to a
Federal Register notice on February 22, 1984 (49 FR 6982).
The TCIC published its recommendations in the Federal
Register on February 14, 1986 (51 FR 5640), and final
decisions by OMB were published on the following October 1
(51 FR 35170). Innumerable editorial details kept the TCIC
busy until the new SIC Manual was printed in Summer, 1987.
Types of revisions to the SIC system include: the
creation of several new Industries; the consolidation of old
Industries that no longer meet size criteria or that can no
longer be distinguished in the current deregulated
environment; and, the shifting of some Industry and Industry
Group boundaries to reflect shifts in economic activity.
Changes are most extensive in high technology areas such as
computers and in services where deregulation has resulted in
substantial changes in the way business is transacted
(particularly in transportation, communications, finance,
and banking).
The revised SIC reflects a long series of compromises
among TCIC members. Representatives of data collection
agencies usually advocated consolidation of existing
Industries to resolve classification problems, and resisted
other changes to minimize costs of redesigning their data
collection programs. Data consumers generally advocated the
creation of more Industries to get more detailed
information. Agencies that had used the SIC system
extensively for years advocated changes that would minimize
disruptions to their time series, while major changes to the
SIC were occasionally proposed by agencies that were more
recent users of SIC-based data or that have experienced
substantial changes in their Industries of concern.
The revisions involve a carefully constructed balance
among:
- the need for extensive changes to bring the SIC
system into closer alignment with the Nation's dynamic
economy, especially with respect to the SIC's
overemphasis on the declining manufacturing sector and
underemphasis on the growing services sector of the
economy;
- the need for radical changes to several principles of
classification that underlie the SIC system and that are
woefully inadequate for representing a dynamic and
increasingly integrated economy;
- the need to minimize changes to maintain
comparability of data for time-series analysis; and
- the need to minimize changes to minimize costs of
converting files of data users and sample frames of data
providers.
Many of the compromises involved budgetary, data
processing, and institutional needs of the TCIC's active
participants, which tended to work against the largest
changes that were both highly needed and very costly because
of the size of the Industries involved. Another force
against change was the classic Catch-22 that data were
needed to determine whether an emerging activity was big
enough to become a new Industry, and the data was not
available because the activity had not been previously
identified as an Industry. TCIC decisions had to be made
from the gut level in some of those cases.
Even though the TCIC's decisions were not always
technically pure, the sometimes heuristic and often
political process resulted in significant improvements to
the SIC system that is now in effect. The general problems
that remain require an overhaul that may be beyond
institutional, budgetary, and data processing resources of
the decentralized Federal statistical community for the
foreseeable future. The data user must be aware of these
problems for now to make more intelligent uses of economic
data collected for 1987 and beyond.
General Problems with the Revised (1987) SIC
System
The TCIC deliberations revealed that traditional SIC
concepts are being eroded by the growing number and
importance of multi-activity establishments, particularly
between industries that are converging in the development of
new products and services. This is illustrated by explosive
growth of intermodal transportation services, such as the
growing number establishments that transcend the traditional
boundaries between motor carriers and airlines for express
letter and parcel delivery service. The emergence of
multi-function establishments creates or exacerbates four
issues that are not addressed in the 1987 SIC Manual. These
issues include: definition of the establishment concept in
an increasingly integrated economy; better identification of
auxiliary establishments; inconsistencies between the
4-digit Industry, 3-digit Industry Group, and 2-digit Major
Group levels of the SIC hierarchy; and overlaps across
boundaries of Divisions.
Definition of Establishment
The concept of establishments is easy to implement when
individual facilities have distinct and stable locations.
Increasingly integrated manufacturing processes, nearly
ubiquitous availability of freeways and air transportation,
and a revolution in the economics and technology of
computers and communications are bringing establishments
closer together functionally while allowing them to become
geographically dispersed and footloose. Establishments are
becoming mobile nodes on a dynamic network.
The SIC Manual recognizes that many establishments in
transportation, communications, and public utilities in
Division E include geographically dispersed activities. This
recognition that establishments may be part of a
transportation system rather than just a single facility is
used only to delineate the activities to be included with a
specific establishment. For example, the maintenance
facilities, control towers, and other parts of a railroad
are lumped together with the home base of those railroad
employees since they are traveling frequently between the
individual facilities.
This recognition that a facility is really part of a more
distant establishment needs to be extended to classifying
the establishment's Industry. This is done to some extent in
transportation, where local pick up and delivery service
that is part of an intercity carrier is classified in the
Industry for intercity or long-distance service by that mode
of transportation, and local service that is not an integral
part of the intercity service is classified in the local
Industry for that mode.
Formal recognition that the function of the network
defines the function of the establishment on that network is
difficult because the SIC system is supposedly based on the
activity of the establishment rather than the enterprise
(company or network) that it serves. Until a broader
definition of establishment is adopted, the basis of
classifying establishments will vary between what goes on
the the establishment without regard to other units and what
the establishment does as part of a network of other units.
Identification of Auxiliaries
The SIC Manual recognizes two kinds of establishments:
operating and auxiliary. An operating establishment sells
most of its goods or services to establishments owned by
other enterprises, and is classified by what it makes or
does. An auxiliary establishment sells most of its goods and
services to other establishments of the enterprise that also
owns it, and is classified by the activities of the
establishments that it serves. For example, a local motor
carrier that serves a grocery chain is classified as an
operating establishment with local trucking in Industry
Group 421. A local motor carrier subsidiary of the grocery
chain is classified as an auxiliary establishment in
Industry 5411 (Grocery Stores).
Under the 1987 SIC Manual, a fifth digit has been added
to indicate what the auxiliary establishment does for the
Industry that it serves. The trucking example would be
classified as a warehousing auxiliary if the trucking arm of
the grocery store was based with the warehouse, or as an
auxiliary not elsewhere classified. Most auxiliary
establishments are either central administrative offices,
and a large percentage are in-house research and development
laboratories.
The new fifth digit does not provide enough detail in the
identification of auxiliary establishments by their function
because of policy analysis needs and a potential problem
with time-series statistics. The need for a more detailed
identification of auxiliary activities is particularly
particularly acute in the trucking example. Less than half
of all trucking activity in the United States is performed
by motor carriers. The majority is performed by private
trucking, such as the trucking auxiliary of the grocery
chain already mentioned. Many government policies involve
establishments that are engaged in trucking, whether private
or for hire, and statistics are needed on both segments of
the industry. Data on operating establishments engaged in
trucking are easily found in Industry Group 421, but
trucking auxiliaries are not identified even with the newly
ordained fifth digit. Data on the auxiliary establishments
are also needed if a major shift in revenues, employment,
and other statistics is measured for the trucking industry,
since the shift may reflect a change in ownership rather
than true change in economic activity. In some States, a
grocery chain that spins off its trucking auxiliary to
become a motor carrier can cause significant changes in the
Industry statistics even though the trucks may be providing
the same service to the same stores under independent
ownership.
These problems are conceptually easy to resolve by
expanding the auxiliary function code to four digits. Rather
than use the four categories of auxiliary activity in the
new fifth digit, the auxiliary activity could be identified
by the code of the Industry into which the establishment
would have been classified if it had been an operating
establishment. The trucking auxiliary of the grocery chain
would thus be classified by the number 5411.4214. The TCIC
selected the single digit code over this option for the 1987
revision on the grounds that implementation would be too
difficult at this time.
Inconsistent Hierarchical Structure
The continued growth of multifunction and intermodal
establishments highlights two fundamental weaknesses of the
SIC system's hierarchy of Industries, Industry Groups, and
Major Groups: the lack of a rigorously defined, consistent
hierarchy across the four-, three-, and two-digit levels;
and, the lack of places to classify interindustry
activities.
A rigorously defined, consistent hierarchy above the
Industry level is important in the SIC system because:
- substantial data are collected or published only at
the two- and three-digit level;
- the hierarchy specifies relationships among
Industries, providing direction to coding and to
aggregation of statistics; and
- the TCIC is very reluctant to move activities across
Industry Group and Major Group boundaries.
The importance of this hierarchy is in contrast to the
lack of rigor in its definition. Only Industries are defined
by rigorous, detailed, explicit rules. Most TCIC
recommendations for reorganizing Industry Groups and Major
Groups were based on intuitive reasonableness or on the
limits of a base 10 numbering system. (Industries were
reallocated among Industry Groups when more than nine were
together.) This lack of rigor has resulted in
inconsistencies and caprice, such as the grouping of public
warehousing with trucking in Major Group 42 even though the
warehousing industries are more logically placed with
transportation services (Major Group 47).
Even if the hierarchy were rigorously defined, there are
few places for interindustry activities to be classified.
Activities which are the outgrowth of an interaction between
Industries, such as piggyback service between trucks in
Major Group 42 and railroads in Major Group 40, are
typically buried in one of the affected Industries or under
"Not Elsewhere Classified." Proposals for new Industries are
useful but piecemeal patches to the problem, especially when
the affected Industries reside in different Industry Groups.
The inconsistent hierarchy and other problems cannot be
resolved completely without a draconian overhaul of the SIC
system. The SIC system might have to be restructured from a
top-down approach that starts with Divisions, rather than
from the TCIC's bottom-up approach that starts with
Industries. The requisite magnitude of change may never be
possible given the substantial costs of changing data
collection programs and converting long-term series of data
to a new SIC system.
Boundaries
The problems with the hierarchical structure of the SIC
system extends to the Division level, although the issue is
primarily one of boundaries rather than of how many and what
type of categories should be defined. The boundary problems
are most notable between the Divisions for manufacturing and
services, and between Public Administration and the other
Divisions. Boundary problems also exist between the
Divisions for manufacturing, wholesale trade, and retail
trade.
Problems with the boundary between Manufacturing
(Division D) and Services (Division I) involve a growing
number of borderline industries that are packaging services
into goods and turning goods into services. The shift from
services to goods is illustrated by the new Industry 7372
(Prepackaged Software), the location of which was
extensively debated because computer programs were
traditionally produced as a service until microcomputers
created a mass market for software. When is a computer
program a good or service? Is it a question of whether the
program is sold in shrink-wrap plastic? The shift goods to
services is illustrated by Publishing (Major Group 27),
where many printed goods are becoming more popular in
electronic form and potentially causing establishments to
move from their manufacturing roots in Division D to less
readily apparent categories in Division I (such as Industry
7375 for Information Retrieval Services).
The borderline Industries are particularly important
given the pervasive references to the manufacturing and
service sectors of the economy using a plethora of explicit
and implied definitions of those sectors. The often reported
shift of the Nation's economy from manufacturing to services
is due in part to the changing character of goods and
services just noted and to the increasing tendency of
manufacturers to contract with other firms for services that
were previously performed in-house. Data users must be
keenly aware of such definitional rather than functional
moves of establishments and activity when analyzing
purported shifts among sectors of the economy.
Another dynamic and inconsistent boundary exists between
the public and private sectors of the economy. Prior to
1972, all government and government-owned establishments
were classified in Division J. The Division was limited in
1972 to public administration and to other government
activities that do not have counterparts in the private
sector. This principle has not been implemented uniformly.
For example, public transit and fire protection are both
activities that have private sector counterparts and should
therefore not appear in Division J. In the 1987 SIC Manual,
public and private transit are properly together in Division
E (Transportation), but fire protection shows up in
Divisions A (Agriculture), I (Services), and J (Public
Administration).
The inconsistent classification of public sector
activities with their private sector counterparts is
exacerbated in the case of central administrative offices.
In concept, activities of the public sector that do not have
counterparts in the private sector are classified in Public
Administration. In practice, most industries in Public
Administration are central administrative offices of
establishments that may or may not have private sector
counterparts. The SIC system thus classifies most central
administrative offices of public agencies as operating
establishments, and all central administrative offices in
the private sector as mere auxiliaries that are classified
with the activity being administered. The inclusion or
exclusion of auxiliaries in data must be carefully noted in
analyses of public and private sector office growth that has
become so important to regional and local economic
development.
The boundary problems of the Public Administration
Division are symptomatic of an ambivalence toward the way to
characterize the public sector in the SIC system that has
been made worse by recent efforts to "privatize" many
government activities. The 1972 SIC revision changed the
Division from Government (which included all public sector
activities) to Public Administration (which is a mixed bag
of administrative activities and those operational
activities that are commonly unique to the government).
Other approaches need consideration in future revisions,
such as moving all operational activities of government to
Divisions A through I and changing Division J to
Administration for central administrative office of both
public and private sectors.
There are similar boundary problems among Wholesale
Trade, Retail Trade, and other Divisions. Part of the
problem is the difference between popular perception and the
SIC definition of wholesale trade. While the public thinks
of wholesale trade in terms of large volume discount
outlets, the SIC definition is based on sales to other
businesses. Retail trade is sales to households. The
distinction between wholesale and retail is getting harder
to make as the trade responds to growing purchases by
households of products originally designed for business
(such as microcomputers).
Conclusion
For its shortcomings, the SIC system remains the most
detailed, thorough framework for collecting and analyzing
economic data that is in widespread use. The SIC system is
based on statistical principles of classification, on both
rigorous and anecdotal evidence of change in the economy, on
budgetary and institutional constraints, and on the personal
interactions of TCIC members through years of debate and
decisions. Providers and users of economic data for public
decision making must be aware of the SIC's principals,
nuances, strengths, and weaknesses when developing and
interpreting establishment-based data; and those concerned
with quality and quantity of future economic data should be
prepared to provide comments when the TCIC reconvenes to
consider further changes to the SIC system for 1992.
Bibliography
(1) U.S. Executive Office of the President, Office of
Management and Budget, Standard Industrial Classification
Manual, 1987. Washington, D.C.: U.S. Government Printing
Office, 1987.
|