Industrial production and capacity utilization: the 2002 historical and annual revision.

In late 2002, the Federal Reserve published a revision of its index of industrial production (IP) and the related measures of capacity and capacity utilization. The primary feature of the revision was the reclassification back to 1972 of production and capacity indexes for individual industries from the Standard Industrial Classification, or SIC, to the North American Industry Classification System, or NAICS. Also, the revision, as usual, updated all measures to incorporate newly available and more comprehensive source data for recent years, and it introduced improved methods for measuring the annual real output of communications equipment manufacturing.

Along with the updating and the restatement of the data using NAICS, the base year used for comparison has changed. Now all production and capacity indexes are expressed as percentages of output in 1997; the previous comparison base was 1992. The rebasing affects all series from their start dates, which are 1919 for total IP and manufacturing IP, 1948 for manufacturing capacity, and 1967 for total industrial capacity. The Federal Reserve's accompanying indexes of industrial electric power use, which begin in 1972, have also been restated to accord with NAICS, rebased to use 1997 as a comparison year, and revised to incorporate previously unavailable data.

The new information resulted in an upward revision to the rate of increase in industrial production and capacity from 1997 to 2000 (chart 1). Improved estimates for the production of communications equipment and semiconductor manufacturing accounted for most of the upward revision; revised estimates for the output of newspapers and related publishers also contributed. The upward revision to the rate of increase in production was greater than the upward revision to the pace of capacity expansion. As a result, for the 1997-2000 period, the average rate of industrial capacity utilization--the ratio of production to capacity--is 0.7 percentage point higher than that previously reported. The higher utilization rates were concentrated in the selected high-technology group of industries (semiconductors, computers, and communications equipment); in the motor vehicle, fabricated metal product, and machinery manufacturing industries; and in utilities.

[GRAPHIC OMITTED]

On balance, the picture of the industrial sector in recent years is little changed by the revision (see appendix tables A.1-A.10 for detailed revision results). The most recent business-cycle peak in monthly IP is still June 2000, at 116.2 percent of 1997 output (table A.1), and the drop in the index from then until December 2001 is 6% percent, about the same as the previously reported decline. Accordingly, the second quarter of 2000 remains the peak in the rate of industrial capacity utilization, and the low is the fourth quarter of 2001 (table A.2). The revised utilization rate reaches 83.5 percent in the second quarter of 2000--0.9 percentage point higher than previously reported--before falling 2 percentage points by the end of 2000 and 6 1/2 percentage points further by the fourth quarter of 2001; the cumulative drop in the rate of capacity utilization is 0.5 percentage point steeper than that previously reported.

In January 2002, industrial production rose; as in the earlier data, the January increase was the first monthly increase since September 2000. Monthly gains in industrial production then averaged 0.4 percent per month through July 2002, but from August to October 2002, industrial production retreated, on balance. In the third quarter of 2002, the revised and rebased production and capacity indexes stood at 111.4 and 146.2 percent of 1997 output, respectively. The rate of industrial capacity utilization in the third quarter of 2002, at 76.2 percent, is essentially unchanged from the previously reported data (table A.7). The rate is more than 5 percentage points below its 1972-2001 average and about 3 percentage points below the trough in the 1990-91 recession but 5 percentage points above the trough in the 1982 recession. (1)

The updated measures continue to show that manufacturing IP, after having increased rapidly in 1999 and the first half of 2000, fell sharply in 2001 and rose at a tepid rate, on balance, in the first three quarters of 2002 (table A.3). On the basis of the revised production indexes and the results of the Census Bureau's 2001 Survey of Plant Capacity, capacity utilization in manufacturing still shows a sharp drop in 2001 (table A.4), and the expansion of manufacturing capacity still exhibits a noticeable slowing from the rapid pace posted in the last half of the 1990s. The factory operating rate has increased since its business-cycle low in the fourth quarter of 2001, but as of the third quarter of 2002, its level of 74.3 percent was more than 6 percentage points below its long-term average. (2)

As in the earlier data, the output of selected high-technology industries--computers, semiconductors, and communications equipment--increased at an average rate of more than 40 percent per year from 1994 to 2000 but dropped off sharply in 2001 (chart 2). For 2002, the revised measures show a more modest rate of increase for their production than previously reported. The rate of capacity utilization in these industries has hovered at or below 63 percent for nearly one year, a level more than 17 percentage points below its 1972-2001 average of about 80 percent. Within this group of industries, the output index for computers was revised down for 1999 and 2000, a move reflecting updated results from the Census Bureau on the value of production in those years. In addition, the indexes for semiconductors and communications equipment were revised up, primarily for 1999 and 2000, a move reflecting new and refined estimates of prices.

[GRAPHIC OMITTED]

The revisions to the IP index for recent years were derived principally from the inclusion of information contained in annual reports issued by the Bureau of the Census: the 2000 Annual Survey of Manufactures (ASM) and selected 2001 Current Industrial Reports. Revised annual data from the U.S. Geological Survey (USGS) on minerals (except fuels) for 2000 and some new data for 2001 have also been introduced. Also, the new monthly production estimates for 2001 and 2002 reflect updated seasonal factors and the inclusion of monthly source data that became available, or were revised, after the closing of the regular four-month reporting window.

The capacity indexes and capacity utilization rates incorporate the revised production indexes, results from the Census Bureau's 2001 Survey of Plant Capacity for the fourth quarter of that year, and newly available 2001 data on industrial capacity from the USGS, the Energy Information Agency, and other organizations. In addition, the relationships used to estimate the current change in manufacturing capacity reflect the inclusion of ASM data on capital spending by industry for 2000 and updated indicators of capital spending by manufacturers in 2001 and 2002.

SUMMARY OF THE HISTORICAL REVISION

The annual industry source data used to construct industrial production and capacity utilization are now being reported using NAICS. The Federal Reserve has adopted NAICS for its monthly statistics on the industrial sector; but to facilitate business-cycle analysis, research, and forecasting, it has done so without changing the scope or historical continuity of these statistics. Specifically, the introduction of the new classification system did not affect the coverage of production and capacity utilization for total industry and manufacturing, and the individual industry components of these measures are available on a NAICS basis back at least to 1972.

The consistency of the production and capacity indexes was further improved by the recompilation of the indexes using current methods, when possible, back to 1972. (These changes are detailed in the section "Current Methods Applied to Earlier Data.") Table 1 summarizes the revised rates of change in the basic measures from 1972 to 1997. The application of current methods for benchmarking IP to annual real output measures, estimating changes in capacity, and aggregating individual series resulted in a small downward revision to the average rate of change in industrial production and capacity from 1972 to 1987. All told, however, the average utilization rate, at 81.5 percent of total industrial capacity from 1972 to 1997, was little changed by the revision.

The 2002 revision also introduced refined methods for grouping individual industry IP series into major market groups for analysis of industrial production and for grouping industrial capacity and capacity utilization rates by stage of process. These changes, which are explained in the section "New Market and Stage-of-Process Aggregates," begin with data for 1967; the revised rates of change in IP by major market groups from 1967 to 2002 on are shown in table 2. The revisions shown reflect not only the refined industry composition of the groups but also, as mentioned above, the application of current methods and available source data to estimates for earlier periods.

The changes in monthly IP reflect the updating of seasonal factors for all years using current methods and the inclusion, when possible, of current monthly and quarterly source data. All told, the revised rates of change in monthly IP from 1972 on are highly correlated with the previously reported rates; the simple correlation coefficient between them is 0.91, and the correlation between the revised and earlier quarterly rates of change is 0.97. In addition to revised changes in production, the monthly changes in capacity utilization reflect the application of current methods for interpolating annual changes in capacity.

The months and years of the peaks and troughs in industrial production since 1972 are shown in table 3. The months of the peaks and troughs associated with the recessions that began in 1973, 1981, and 2000 are unchanged with this revision. The months of the peaks in IP before the onset of the episodes leading to the troughs in 1980 and 1991 did change. As in the earlier data, however, industrial production remained within a narrow range for more than a year before both downturns, and the changes did not alter the picture of cyclical activity in either period. The profile of the industrial expansion in the 1990s--rapid increases in IP punctuated by a slowdown in 1995 and again in 1998 in the aftermath of the Asian crisis--also is unchanged.

Although the timing of the business-cycle episodes in industrial production is essentially unchanged by the revision, the 1973-74 recession is now reported to be somewhat shallower than previously reported. As a result, the drop in capacity utilization--about 14 1/2 percentage points from November 1973 to May 1975--is about 1 3/4 percentage points smaller than in the earlier data. Also, the recovery from 1975 to 1979 is now a bit less strong, and the peak reached by capacity utilization is not as high as previously reported. Finally, as already discussed, the revised rate of capacity utilization for total industry is higher, on average, from 1997 to 2000, but then it drops a bit more steeply in 2001 than previously reported.

NEW NAICS INDUSTRY STRUCTURE

The Federal Reserve still defines the industrial sector as manufacturing, mining, and electric and gas utilities. The changes from the SIC system to NAICS, however, altered the industry composition of manufacturing. Specifically, NAICS moved the logging industry and the newspaper, periodical, book, and directory publishing industries from manufacturing to other sectors; logging was placed in agriculture, and the publishing industries were placed in the new information sector. (3) For the statistics reported in the Federal Reserve's monthly G. 17 release, the manufacturing measures will continue to comprise those industries included in the NAICS definition of manufacturing plus the logging and newspaper, periodical, book, and directory publishing industries that have traditionally been considered manufacturing.

An abbreviated version of the new industry structure appears in table 4. The G.17 release shows the aggregate of industries representing the NAICS definition of manufacturing, along with the aggregate of industries representing the traditional definition of manufacturing. For the most part, the two series are similar, in terms of their long-term trends (see memo items in table 1) and their basic cyclical profile. However, the average annual proportion of the traditional manufacturing measure in total industrial production is about 85 percent, whereas the proportion of manufacturing (NAICS) is about 80 percent (see table A. 10).

Conversion of the Data to NAICS

The historical source data needed to compile IP and capacity utilization are not publicly available on a NAICS basis before 1997; hence, the issuance of thirty years of NAICS industry statistics represents a major effort by the Federal Reserve to preserve the historical continuity of the basic measures presented in its G.17 release. As a result, many frequently used industry series whose definition and coverage were altered by NAICS--communications equipment, construction equipment, and chemicals, to name a few--are still available with substantial history.

The restatement of the industrial production and capacity utilization data from 1972 to present on a NAICS basis relies on the results of a research project conducted by the Federal Reserve Board and the Center for Economic Studies of the Bureau of the Census (see "Technical Note on Reclassifying Data in the Census of Manufactures from the SIC to NAICS" for further information). In a nutshell, the project developed NAICS codes for each establishment in the files of seven Censuses of Manufactures (COM)--1963, 1967, 1972, 1977, 1982, 1987, and 1992. The information needed to derive NAICS-based source data for industrial production and capacity utilization was obtained by tabulating the historical COM establishment-level data using the NAICS codes assigned by the research project.

The derivation of NAICS-based source data was an extensive effort. It involved reconstructing many working data sets that underlie the estimation of IP and capacity, including the annual comprehensive estimates of industry value added and value of production and the annual (fourth quarter) survey data for industry utilization rates. All in all, annual figures for most variables reported in the Censuses and Annual Surveys of Manufactures (shipments, value added, cost of materials, inventories, capital spending, production-worker hours, and the like) were derived at the six-digit NAICS level from 1972 on. Utilization rates from the Survey of Plant Capacity were reconstructed beginning in the fourth quarter of 1974, the start date of the survey. The Federal Reserve's data on monthly electric power use were derived at the four-digit NAICS industry level from data in the Annual Survey of Manufactures, which were also restated to accord with NAICS. The 2002 NAICS was used for all restatements and conversions. (4)

Restructuring of Industry Subsectors

NAICS substantially restructured many industries within manufacturing. One significant change was the reorganization of high-technology industries. NAICS created a new subsector for high-technology manufacturing, computer and electronic product manufacturing (NAICS 334), which combined into a sensible aggregate industries that had been scattered across various two-digit SIC industry groups. For some time, the G.17 has reported output, capacity, and capacity utilization for selected high-technology industries: semiconductors, computers, and communications equipment manufacturing. These industries account for most of the new NAICS 334 subsector. The new subsector also contains audio and video equipment and navigational, measuring, electromedical, and control instruments. The output, capacity, and capacity utilization measures for selected high-technology industries will continue to be reported in the monthly G. 17 release along with data for the new NAICS subsector.

Within the transportation equipment manufacturing subsector, NAICS introduced a new industry group for motor vehicle parts manufacturing (NAICS 3363). The group contains eleven new six-digit industries for motor vehicles parts, many of which--metal stamping, vehicular lighting and electronic equipment, motor vehicle seating and interior trim, and motor vehicle air-conditioning manufacturing--were previously parts of a wide range of two-digit SIC major groups, including fabricated metals, furniture, apparel, and electrical and nonelectrical machinery. As a result of these changes, the monthly IP index for motor vehicle manufacturing now comprises nine NAICS-based industry series. Some of these series, such as metal stamping and motor vehicle air-conditioning manufacturing, were separate series in the previous IP index, and the change entailed a simple rearrangement of the data. As in the previous structure, each industry series in motor vehicle parts, except for metal stamping, is further disaggregated into two sub-industry indexes--one for the production of original equipment and the other for the production of replacement parts. All told, the motor vehicle parts industry group is now represented by seventeen individual IP series, and the proportion of the industry group in the overall index is 3.4 percent, about 1 1/4 percentage points larger than it was in the SIC-based IP data.

Another change split the SIC two-digit textiles and products major group (SIC 22) into two NAICS subsectors: textile mills (NAICS 313) and textile product mills (NAICS 314). Within these subsectors, a few industries in SIC 22 were moved to apparel manufacturing (NAICS 315), and a few others previously not in the SIC textile group were newly included (mainly from the SIC two-digit group for apparel). The implementation of the NAICS structure for textiles in the IP index was accompanied by an extensive review of available source data; the result was the introduction of several product series new to the IP system. In particular, the IP physical product measures for NAICS 3131 (fiber, yarn, and thread mills) were broadened relative to the corresponding SIC series to include wool fibers. In addition, IP measures for NAICS 3132 (fabric mills) now use a quarterly production series for cotton and synthetic fabrics. Finally, tire cord production (part of NAICS 3149) is now being compiled as a separate IP series derived from physical product data.

The revised industrial production index contains monthly output indexes for 227 NAICS six-digit (or combination of six-digit) industries; previously, the index represented 207 SIC-based industries. Of course, the industrial production index contains many sub-industry indexes, developed from product data, that are used to compile market groups and, ultimately, the total index. The introduction of NAICS does not change the way in which product data are used to compile monthly IP; taking these product-based sub-industry indexes into account, the revised IP index is now built from 295 individual component series. (5)

The implementation of NAICS for capacity and capacity utilization resulted in the introduction on net of nine new series in the system. The new industry series are mainly in the chemical and machinery manufacturing subsectors (NAICS 325 and 333, respectively); a new capacity series for lime and gypsum product manufacturing (NAICS 3274) was derived using capacity data issued by the Gypsum Association and introduced from 1972 on. All told, the capacity measures now are built from eighty-five industry series, most of which are NAICS four-digit industries (or combinations of them). (6)

The NAICS subsectors that are now being published in the regular monthly release are shown in the bottom half of table A.5 (which reports changes in IP) and in table A.7 (which shows capacity utilization rates); additional industry detail are being published in the supplemental tables available from the Board's web page for the G.17. The annual proportions of the new industry subsectors in total IP from 1994 on are shown on the bottom portion of table A.10.

NEW METHODS IN THE REVISION

In this revision, new or refined methods for three series were introduced as follows: (1) a new benchmark index for the real output of communications equipment manufacturing, (2) a refined structure of the monthly IP index for semiconductors, and (3) improved methods for estimating light vehicle capacity.

Communications Equipment Manufacturing IP

The Federal Reserve improved the methods it uses for compiling the production series for communications equipment manufacturing (NAICS 3342) from 1988 on. In recent years, the Federal Reserve has made numerous improvements in its measures of real output for the high-tech sector. Two years ago, it introduced a new production index for one component of communications equipment, local area network (LAN) equipment. With the current revision, new results for other types of communications equipment--namely, fiber optic equipment, cable modems, public branch exchanges, and cellular communication equipment--have been introduced. (7)

Price indexes for three types of equipment that are used to run fiber optic networks--multiplexers, dense wave division multiplexing equipment, and digital cross connects--are shown in chart 3. As may be seen, prices for fiber optic equipment declined noticeably from 1996 to 2001. The value of the production of these devices also grew rapidly in the late 1990s and reached more than 10 percent of the total value of communications equipment production in 2000. Overall, value added in communications equipment manufacturing was 2.0 percent of total IP from 1994 to 2000.

The new results are annual price measures that more accurately reflect the technical advances and quality change in the equipment produced by the communications equipment industry. An annual price index for communications equipment manufacturing, which incorporates the previously issued statistics on LAN equipment as well as the new results, is shown in table 5. This index was used to construct the annual IP benchmark index for communications equipment manufacturing and affects the annual changes in that index from 1988 on.

Semiconductors and Related Devices IP

The IP series for the manufacture of semiconductors and related devices (NAICS 334413) is now built from five sub-industry indexes--microprocessor units (MPUs), metal oxide semiconductor (MOS) logic devices excluding MPUs, MOS memory chips, other integrated circuits (linear and analog), and optoelectronics and other discrete devices--from 1992 on. The new series are not published separately, but their inclusion in the IP structure improves the accuracy and compilation of the published monthly index for semiconductor and related electronic components (NAICS 334412-9). Value added for this series averaged 3.4 percent of total IP from 1994 to 2000.

The data on the value of production for the new subcategories of semiconductors, which are not available in reports from the Census Bureau, were developed from information issued by trade associations, private research companies, and company reports. The basic data, which are monthly and quarterly and based largely on reports issued by the Semiconductor Industries Association and Dataquest, are adjusted to comprehensive annual measures issued by the Census Bureau. The price measures for each component, which are updated annually and are thus subject to revision each year, are developed from (1) revised data from the same sources, (2) quarterly data on microprocessor prices that are available annually from Micro Design Resources, and (3) producer price indexes issued by the Bureau of Labor Statistics.

Light Vehicle Capacity

The, capacity of automobile and light duty motor vehicle manufacturing (NAICS 33611) is estimated from plant-level data; in the most recent model year, sixty-six light vehicle assembly plants were operating in the United States. For each of these facilities, capacity in units was developed from data on the actual number of shifts, the length of the shifts, and the speed of the assembly line (line speed). Aggregation of the plant capacities using model-specific prices from 1987 on yielded a capacity index consistent with the production index.

The methods for determining plant capacity from shift and line speed data were refined to better reflect current operating practices and technology. (8) With this revision, a plant's line speed at capacity was determined by the peak within the past ten years; previously, the peak line speed was obtained from all available data, which may have covered more than ten years. The revision also introduced a nonstandard shift configuration, with plants able to rotate three …