US Census Bureau


Guide to the
1997 Economic Census


Assembling Time Series Data

One of the preeminent virtues of the Economic Census program is that comparable data have been collected at fixed intervals and with consistent definitions across decades. Nonetheless, so long as reports were designed in the context of limited budgets for printing, census reports typically included very little historical data. Comparative statistics, covering the current and most recent previous census, have generally been included only for the United States and states. Left to the user is the assembly of time series--such as the growth of retailing in a particular area, or trends in a particular manufacturing industry.

Acquiring Reports from Previous Censuses

Printed reports from the 1992 and earlier economic censuses are no longer available for sale. At the same time, all 1992 census publications are available on the web in portable document format (pdf). Reports from the 1967 to 1992 economic census, indeed all Census Bureau reports issued 1968 to 1996, may be purchased from the Census Bureau on microfiche or as paper copies generated from microfiche. Collections of historical census reports, including reports before 1967, are maintained at certain major libraries, and individual reports may be borrowed through interlibrary loan.

CD-ROMs. Many of these data are also available in machine-readable form. CD-ROM coverage of the 1997 Economic Census is complete, and CD-ROMs for 1992 and 1987 excluded only "Miscellaneous Subjects" reports. Further, the 1992 and 1987 CD-ROMs also include selected data covering earlier time periods. Volume 1j of the 1992 CD-ROM series includes a national time series from the Annual Survey of Manufactures from 1958 to 1995, and monthly retail sales from 1967 to 1994. Volume 4 of the1992 CD-ROM series, entitled "Nonemployer Statistics", includes Geographic Area Series files for 1987 for retail trade, wholesale trade, service industries, and manufacturing, as well as 1982 data for manufacturing, in formats that mirror their 1992 counterparts. More comprehensive data for 1987, and a few data sets for 1982 and 1977 are included on the final 1987 Economic Census CD-ROMs (1e and 2).

CD-ROMs with 1992 and 1987 data are still available for sale. Data files are recorded in .dbf format and can be used directly by or imported into many modern database and spreadsheet programs. Included MS-DOS-based EXTRACT software supports data selection with menus and labels data displays.

Tape. Key files from the 1972, 1977, and 1982 economic censuses were issued on computer tape. These files may be obtained from the National Archives and Records Administration.

Industry Comparability

The implementation of NAICS causes major disruptions in the availability of comparable information across time periods. In recent history, the SIC system was updated 3 times (in 1967, 1972, and 1987) and each time a significant number of new industries was introduced into the existing framework. What is different for 1997 is that the whole framework was changed.(1)

While data for nearly half of the SIC's in use in 1992 can be derived from 1997 NAICS industries, a substantial number of industries cannot be much more than approximated under NAICS. That makes the 1997 Economic Census particularly important, because census questionnaires identified industry components finely enough that data could be categorized under either NAICS or SIC; and as a result certain key data could be published according to the old system as well as the new. The Comparative Statistics report presents the number of establishments, sales, employment and payroll for each SIC for the nation and each state, for both 1997 and 1992. Thus, basic SIC-by-state time series can be carried backward from 1997 to 1987, and farther to the extent that particular industries are not affected by SIC changes in 1987, 1972, and 1967.

NAICS time series can go forward from 1997, but they cannot generally go backward to earlier years, because many NAICS categories require information that was not collected in 1992 and earlier censuses. For instance, NAICS 45321, Office Supplies and Stationery Stores, differs from SIC 5943, Stationery Stores, primarily by the addition of certain office supply stores that were previously classified in wholesale trade. Census questionnaires prior to 1997 did not separately differentiate office supply stores from other kinds of office supply wholesalers, so NAICS 45321 cannot be estimated for prior periods.(2)

The Census Bureau has estimated some national-level NAICS time series back to 1992. That project, intended to support time series for monthly surveys, is limited to broad industry categories within manufacturing, retail trade and wholesale trade.

Users have access to correspondence tables between the old and new systems at and in the formal NAICS Manual. These tables show for each NAICS industry the SIC categories or parts thereof that comprise them, and for each SIC industry the NAICS industries or parts thereof to which their establishments are likely to be reclassified. The 1997 Economic Census Bridge between NAICS and SIC report takes that correspondence a significant step farther by showing the number of establishments, sales, employment, and payroll at the national level for each of those intersections between the old and new systems. For example, the Bridge report shows that the office supply stores that were transferred into retail from wholesale trade increased sales of the new retail Office Supply and Stationery Stores category by nearly ten-fold over the SIC industry of the same name.

The Bridge tables are so powerful that one might be tempted to "update" historical figures by SIC to NAICS by applying proportions derivable from the tables. The user should employ such "synthetic estimation" with caution, however, since nationwide proportions may not reflect relationships in particular geographic areas or in prior time periods. Many new industries reflected in NAICS did not exist to a significant extent in prior periods, for example, satellite telecommunications.

At broader levels of classification, the changes between SIC and NAICS are further confounded by the rearrangement of the hierarchy. For example, the service industries division of the SIC was subdivided into five new NAICS sectors and parts of four others. Less noticeable, but perhaps more troublesome, are shifts affecting sectors--like manufacturing, wholesale trade, and retail trade--that retain their status as sector titles in NAICS but are being affected by changes in scope. Retail trade is smaller under NAICS than under SIC just because eating and drinking places, which accounted for roughly 10 percent of SIC-based retail sales, were transferred to the new Accommodation and Food Services sector. Retail losses are offset partially by transfers between retail and wholesale trade, such as the office supply stores mentioned above. Manufacturing also shrinks somewhat under NAICS because significant components have been reclassified elsewhere.

Scope of Economic Census Programs

Prior to 1992, the Economic Census program covered less of the American economy. In 1987 and earlier years, the census did not include Finance, Insurance and Real Estate; and it included only selected transportation industries within the Transportation, Communication, and Utilities sector. The addition of those components boosted census coverage from roughly 76 percent of the gross domestic product in 1987 to about 98 percent in 1992. The coverage of service industries expanded in 1967, 1977, and 1987. Thus, time series available for some industries are relatively short.

Geographic Comparability

Most students of economic trends confine themselves to looking at the nation, states, and counties. County boundary changes are few and far between, while many places, metropolitan areas, and ZIP Codes change boundaries over time. Geographic comparability of sub state areas may be a moot issue between 1992 and 1997, because there are no plans to publish 1997 data for counties, places and metropolitan areas on a basis allowing for comparison with 1992 data (that is, SIC). Questions as seemingly routine as "Did manufacturing employment in my area go up or down?" may remain unanswered for 1997.

1. Paul T. Zeisset & Mark E. Wallace, "How NAICS Will Affect Data Users", published on the Internet at

2. Ibid.

Previous topic Next topic