1 Beginning in 1982, inventories are stated at current cost and are notcomparable to the book value estimates for prior years.
2 Ratio based on December seasonally adjusted inventory data, and themonthly average of the unadjusted annual shipments.

Source: U.S. Census Bureau, Current Industrial Reports, series M3,Manufacturers' Shipments, Inventories, and Orders: 1987-1997;earlier reports; (accessed March 20 2000).

http://www.census.gov/indicator/www/m3/index.html

*

MANUFACTURERS' SHIPMENTS, INVENTORIES, AND ORDERS (M3) SURVEY

DESCRIPTION OF SURVEY

The Manufacturers' Shipments, Inventories, andOrders (M3) survey provides broad-based, monthlystatistical data on economic conditions in thedomestic manufacturing sector. The survey measurescurrent industrial activity and provides an indicationof future business trends. Data are used by theExecutive Branch of the Government for developingeconomic, fiscal, and monetary policy; the Bureau ofEconomic Analysis as components of the grossdomestic product estimates; and trade associations,corporate economists, and other members of thebusiness community as an analytical tool forassessing the current and future economic conditionof the country.

COMPOSITION OF INDUSTRY CATEGORIES

There are 80 separately tabulated industrycategories in the M3 survey. These categories aregroupings of the 459 manufacturing industries asdefined in the 1987 Standard IndustrialClassification (SIC) Manual. Because companiesprovide data on a voluntary basis, reporting in manyof these 80 categories is not sufficient to warrantseparate publication of the data. Consolidatedreporting by some large companies also limits thequality of the data in some categories. A companywhich reports on a consolidated basis (entirecompany or a division) is included in the M3category in which the reporting unit as a whole isclassified and has no influence on the movement ofthe other industries which may be included in itscombined report. As a result, it has been necessaryto combine the 80 industry categories into 45publication levels for shipments and total inventories.For new and unfilled orders and inventories by stageof fabrication, it has been necessary to introducefurther combinations because of the lower responserates for those items.

COMPOSITION OF THE SURVEY PANEL

The monthly M3 estimates are based on informationobtained from most manufacturing companies with$500 million or more in annual shipments. In orderto strengthen the sample coverage in individualindustry categories, the survey includes selectedsmaller companies. The sources of the companiesfor the survey are the quinquennial censuses ofmanufactures and the annual surveys ofmanufactures in the interim years. For firms thatoperate in a single M3 industry category, thereporting unit typically comprises all operations ofthe company. Most large, diversified companies,however, file separate reports for divisions withsignificant activity in different industrial areas. Thecomposition of each company or reporting unit of acompany in the survey usually includes more thanone plant or establishment and frequently includesindustry activities outside the M3 category in whichit is classified. The survey methodology describedlater in this chapter assumes that themonth-to-month changes of the total operations ofthe reporting units classified in each industrycategory effectively represent the month-to-monthmovements of all establishments which make up thecategory. In 1962, the initially selected sample forthis survey included all companies with more than1,000 employees and smaller companies selectedwith probabilities proportional to their employmentsize within each industry category. As there wassome deterioration in response rates for companieswith between 100 and 1,000 employees, in January1975, the staff selected a supplemental sample ofapproximately 1,000 companies from the universe ofcompanies in this size range in order to strengthenthe estimates. Although the response rate was onlyabout 60 percent for this group, beginning in January1978 these data were added to the panel andincluded in the calculations of the monthly estimates.Because of poor response rates, the survey nolonger included companies with less than 100employees; instead, data for these companies wereestimated by using overall industry month-to-monthmovements based on data reported by the largercompanies. In recent years, the size of the value ofshipments of the company or reporting unit ratherthan the employment size has influenced the selectionof companies to increase response rates. Using thiscriterion, census staff visit or otherwise contact largecompanies that did not report in the survey torequest their participation or reconsideration of aprevious decision not to participate. Also, staffrequest data for large diversified reportingcompanies to provide additional industry categoriesnot previously provided. Another method used forimproving response has been contactingnonreporting companies by letter. Staff send letterson an ongoing basis to companies in industrycategories with low response rates. In 1990, staffselected a probability sample and mailed requests toabout 400 midsize companies in the plasticsindustry. The purpose was to test the viability ofprobability sampling, especially in industriescomprised primarily of smaller, less diversifiedcompanies. As a result of these efforts, we increasedresponse by adding about 45 to 55 percent of thecompanies contacted to the panel. However,respondent dropouts frequently offset theseincreases. The current coverage levels in the surveyshow that reported data represent approximately 55percent of the shipments estimates at the totalmanufacturing level, while the individual coveragerates for the 20 two-digit major SIC industries varyfrom about 20 to 99 percent.

LIMITATIONS OF THE DATA

The monthly data presented in this publication aresubject to both sampling and nonsampling errors.Sampling errors occur because reports are receivedfrom a sample rather than from the completeuniverse of manufacturing companies. Because thepresent composition of the panel is not based on aprobability sample, the amount of sampling errorcannot be quantified. Nonsampling errors, on theother hand, are attributable to many sources. Theuse of company or divisional reports to estimate themonthly change for establishments is one source ofnonsampling error. The use primarily of largecompanies to represent the month-to-monthmovement of all companies is another potentialsource. In addition, response and processing errorsmay be present, although computer edits andanalytical review of the data detect the mostsignificant errors of this kind prior to tabulation.

MONTHLY ESTIMATION PROCEDURE

A link relative procedure derives the monthlyuniverse estimates of shipments, unfilled orders, andtotal inventories for each industry category. Theuniverse estimate for the previous month is multipliedby the monthly ratio of change tabulated forreporting companies in the current month to arrive ata universe estimate for current month. When anindividual company reports unusually large changesfrom the previous month, or when a particularcompany has unique data patterns differingsubstantially from the movement shown by the restof the reporting panel in a particular industrycategory, the company is excluded from the ratio ofchange calculation but included in the universeestimate of level. The effect of this procedure is torestrict the estimation for nonrespondents and firmsnot in the survey panel to the general trend of theindustry. The universe estimate of new orders isderived from the monthly estimate of shipments plusthe change in unfilled orders between the current andprior period. The estimate includes orders that arereceived and filled in the same month as well asorders that have not yet been filled. It also includesthe effects of cancellations and modifications ofpreviously reported orders. The standard ratioestimate procedure is not followed for new ordersbecause not all companies report new orders, andsome that do report this item limit their reporting tospecific products for which long lead times arerequired in the production cycle. These companies,in effect, exclude new orders received for productsthat are shipped from inventory. A modifiedprocedure also is used to estimate the stage offabrication inventory data. In this procedure, thetotal inventory data estimated for each tabulatedindustry category are retabulated to the appropriatetwo-digit SIC major group levels and serve ascontrol totals for the stage of fabrication data. Initialestimates are made for each of the stages offabrication at the two-digit SIC level using the ratioestimation procedure. The differences between thesum of the stage of fabrication detail and totalinventories at the two-digit SIC level are thenallocated proportionally to the stage of fabricationfigures to form the estimates. The reasoning behindthis procedure is that a significant number ofcompanies report total inventories but cannot reportinventories by stage of fabrication.

TRADING DAY ADJUSTMENT

Variations in the rate of manufacturing activityresulting from different numbers of trading days inthe same month for different years and variations inthe length of months can be an important cause ofmonth-to-month fluctuations in the shipments data.For many industries, these irregularities can beidentified approximately and removed so that theunderlying trend cycle stands out clearly. Recentsoftware and data processing improvements havefacilitated research and implementation of uniquetrading day adjustment factors for each series. Theseimprovements include refinements to the trading dayanalysis in the X-11 seasonal adjustment program,additions to diagnostic software that identifiesprogressively unstable data patterns, and redesign ofthe monthly survey data processing system. Most ofthe shipments series have trading day factorscalculated in the trading day regression analysisroutine of the X-11 software. The adjustment from
1987 forward used these new factors. Table FN1of M3-1(90) shows the factors used before 1987.However, eight of the shipment series did not haveidentifiable trading day patterns and are only beingadjusted for length of month variations.

SEASONAL ADJUSTMENT METHODOLOGY

The monthly data are adjusted for seasonality at themost detailed level tabulated in the survey, using theX-11 ARIMA version of the Census Bureau'sseasonal adjustment program. The seasonallyadjusted estimates for shipments, unfilled orders,and total inventories for M3 industry categories arecalculated by dividing the unadjusted estimates byseasonal adjustment factors computed by the X-11ARIMA seasonal adjustment program. Seasonallyadjusted new orders are computed by adding thechanges between current and prior period seasonallyadjusted unfilled orders to the current month'sseasonally adjusted shipments. The inventory bystage of fabrication data are seasonally adjusted atthe two-digit SIC major group level for each stage.If the sum of the adjusted stage of fabrication doesnot equal the major group totals resulting fromsumming the seasonally adjusted total inventories forthe individual industries, the difference isproportionally allocated to the stage of fabricationdetail. Staff calculate seasonal factors concurrentlyand include the current period observation in thecalculation of the seasonal factor for that month. Thefactors reflected in this report are based on usingdata from January 1982 through December 1996for shipments, unfilled orders, and inventories. Forinformation on specific measures used in theseasonal adjustment analysis, selection of optionswithin the X-11 program for the individual industryseries, and tests for the presence of seasonality,contact the Manufacturing and ConstructionDivision, U.S. Census Bureau, Washington, DC20233, or call 301-457-4749.

BENCHMARK PROCEDURE

The M3 survey data are benchmarked to reduceboth sampling and nonsampling errors. The relativelysmall monthly sample size as well as the differencesthat result from collecting the monthly data on adivisional basis as compared to the benchmark dataon an establishment basis account for most of therevision. Also, some monthly reports received toolate to be included in the monthly publications areadded to improve the revised estimates of change ofthe historical monthly data. The benchmarkingprocess for this report included comparing monthlydata reported by large companies with their annualdata submitted in the 1994 and 1995 Annual Surveyof Manufactures (ASM). Estimates of annualshipments and end-of-year total and stage offabrication inventories calculated in the monthlysurvey were benchmarked directly to the datacompiled in the quinquennial censuses ofmanufactures and the interim ASM's. The unfilledorders levels were adjusted based on changes in theratio of unfilled orders to shipments of companiesreporting in the monthly survey and the ratio of theASM shipments estimate to the summed monthlysurvey data. Since new orders are defined as thechange in unfilled orders between the current andprior periods plus the current period shipments,implicit benchmark estimates for new orders werederived. *DEFINITION OF TERMS

VALUE OF SHIPMENTS

The value of shipments data in the M3 surveyrepresents net selling values, f.o.b. plant, afterdiscounts and allowances and excluding freightcharges and excise taxes. The ASM uses the samedefinition. However, the data reported in the twosurveys frequently are not equal because ofdifferences, discussed earlier, in the reporting unit.For multiestablishment companies, the M3 reportstypically are company- or divisional-level reportsthat encompass groups of plants or products. Thedata reported are usually net sales and receipts fromcustomers and do not include the value of interplanttransfers. The reported sales are used to calculatemonth-to-month changes which bring forward theuniverse estimates developed from the ASM. Thevalue of products made elsewhere under contractfrom materials owned by the plant is also included inshipments as well as receipts for contract workperformed for others, resales, miscellaneousactivities such as the sale of scrap and refuse,installation and repair work performed byemployees of the plant, and the receipts for researchand development performed at the plant. In theshipbuilding industry, the value of work done in agiven year varies considerably from the value ofshipments because of the long lead time between thestart and completion of a large vessel. In both theannual survey and the M3, the value of work doneduring the year, rather than the value of shipments, isrequested. The same procedure is followed foraircraft produced on a cost-plus contract. Aircraftproduced on a fixed-price contract are reported inthe usual manner when the products are shipped.Where value of work done is reported in place ofshipments, the work-in-process inventories arenormally charged to the customer as costs areincurred and are not accumulated as finished goodsinventories by the shipyard or the aircraft producer.The value of shipments figures developed from theASM contain duplication at the M3 industrycategory and higher level aggregates, since theproducts of some four-digit SIC industries are usedas materials by other industries within the same M3industry aggregate. The significance of theduplication within the specific M3 industrycategories varies depending on their four-digitindustry composition. It is most pronounced in a fewhighly integrated industry areas, such as primarymetals and motor vehicles and parts.

INVENTORIES

Inventories in the M3 survey are collected on acurrent cost or pre-LIFO (last in, first out) basis.Because different inventory valuation methods arereflected in the reported data, the estimates differslightly from replacement cost estimates. Companiesusing the LIFO method for valuing inventories reporttheir pre-LIFO value; that is, the adjustment to theirbase-period prices is excluded. In the ASM,inventories are collected according to this samedefinition. However, there are discontinuities in thehistorical data in both surveys. Until 1982,respondents were asked in the ASM to report theirinventories at book values; that is, according towhatever method they used for tax purposes (LIFO,FIFO, and so forth.) Because of this, the value ofaggregate inventories for an industry was notprecise. Effective with the 1982 Census ofManufactures, instructions for reporting inventorieschanged. LIFO users were asked to reportinventories prior to the LIFO adjustment, as well asthe LIFO reserve and the LIFO value afteradjustment for the reserve. Because of this change,the 1982 census data for inventories were publishedon both bases. Inventory data in subsequent ASM'sare available only on a pre-LIFO basis, with nobook value estimates comparable to the earlier data.This change in instructions for reporting current costinventories was carried to the monthly surveybeginning in January 1987. The data for 1982 to1987 were previously redefined (but notrecollected) on a pre-LIFO or current cost basis.As described in the section on benchmarkingmethodology, the procedure used in adjusting thesemonthly estimates from a book-value basis to thepre-LIFO levels used the pre-LIFO end-of-yearlevels from the ASM as constraints and the samegeneral procedure for minimizing revisions to themonth-to-month change in the monthly book-valuedata . Inventory data are requested fromrespondents by stage of fabrication; that is, finishedgoods, work in process, and raw materials andsupplies. There are several limitations to the qualityof these data for two reasons. First, response to thestage of fabrication inquiries is lower than for totalinventories because all companies do not keep theirdata monthly at this level of detail; those companieswhich do maintain monthly detailed recordsfrequently do not have data for all three stages offabrication or have quantity (physical volume) dataonly. Second, a product considered to be a finishedgood in one industry, such as steel mills shapes, maybe reported as a raw material in another industry,such as stamping plants. Therefore, within thetwo-digit SIC major groups the same type ofinventory may be included under different stage offabrication categories. Like total inventories, stage offabrication inventories are also benchmarked to theASM pre-LIFO data, but the stage of fabricationdata are benchmarked at the two-digit major grouplevel.

NEW ORDERS RECEIVED AND UNFILLED ORDERS

New orders, as reported in the monthly survey, arenet of order cancellations and include ordersreceived and filled during the month as well asorders received for future delivery. They also includethe value of contract changes which increase ordecrease the value of the unfilled orders to whichthey relate. Orders are defined to include thosesupported by binding legal documents such assigned contracts, letters of award, or letters of intent,although in some industries this definition may not bestrictly applicable. In the case of letters of intent, thefull amount of the sales value is included if the partiesare in substantial agreement on the amount;otherwise, only the funds specifically authorized tobe expended are included. Unfilled orders includeorders (as defined above) that have not beenreflected as shipments. Generally, unfilled orders atthe end of the reporting period are equal to unfilledorders at the beginning of the period plus net neworders received less net shipments.

*