b. The 1953 Annual Survey Sample:

The universe from which the 1953 Annual Survey sample was drawn was the most current sampling frame that could be assembled, and included some groups that had been inadequately covered by the earlier sample plans.

The new universe consisted of:

(1) All manufacturers that were included in the first quarter 1951 County Business Patterns (CBP) tabulations, except those which had ceased operations before 1953, according to BOASI records.

(2) All manufacturers that received Federal Social Security Employer Identification Numbers (E. I. Numbers) between the first quarter of 1951 and the third quarter of 1953, including "successors" to manufacturing companies that were active during the first quarter 1951.

(3) All manufacturers omitted from the first quarter 1951 CBP list because they were inactive at that time, but who reported to BOASI for the first quarter of 1953 because they had subsequently resumed operations.

(4) All establishments that BOASI reclassified from nonmanufacturing (wholesale, construction, "unknown," etc.) into manufacturing industries, between the first quarter of 1951 and the third quarter of 1953.

(5) Apparel and leather jobbers (regarded as nonmanufacturing by BOASI, in accordance with the S. I. C. System), whose names were obtained from trade association and other special lists.

(6) Large printing and publishing and railroad equipment manufacturing establishments, not covered by the Social Security system, whose names were obtained from special lists.

Census records were used in conjunction with the BOASI lists to define the establishments of "multiunit" companies (i. e. companies with more than one manufacturing establishment). E. I. numbers of the various subsidiary companies controlled by a single parent company, and the individual establishments that Census records had identified with the parent company, were all associated in assembling the company lists for sampling purposes. Special care was taken to obtain complete coverage of government-owned manufacturing establishments that were operated by private firms. Some of these establishments had been omitted from earlier annual surveys because the operating companies failed to report them, having mistakenly considered those activities at government facilities as ''out of

Each establishment in the universe was

assigned a sampling rate of 1/1, 1/2, 1/3, 1/5, 1/10, 1/20, 1/40, 1/80, or 1/160, depending upon the establishment's industry classification and employment size. Within industry, the sampling rates assigned decreased systematically with decreasing employment size. Seven different employment size classes (250 or more, 100-249, 50-99, 20-49, 10-19, 5-9, and 1-4) were used and hence a maximum of seven different sampling rates were assigned to any industry.

The specific sampling pattern for each industry was determined from an analysis of its size distribution, its product class concentration pattern, and the Annual Survey publication specifications for the industry. In general, the controlling survey specifications were desired reliabilities (sampling errors) for the primary product classes of each industry. These specifications increased in stringency with increasing importance of the product class. The sampling plan as applied approximated the optimum allocation for specified reliabilities for the individual product classes of each industry.

For some industries the sampling rates, as determined on the basis of the reliability specifications, were modified to meet other survey specifications. The most important examples were: 1. All establishments with 250 or more employees were included with certainty regardless of their industrial classification. 2. For the ordnance, chemical, petroleum and coal, and the metalworking industries (major industry groups 19, 28, 29, and 33 through 38) all establishments with 100 or more employees were included with certainty because of their importance in connection with national defense activities.

Although initial probabilities of selection (sampling rates) were assigned to each establishment, companies rather than individual establishments were defined as the sampling units. Multiunit companies that had any 250 or more employee plant (100 or more for plants in the defense industries) were included in the sample with certainty. For other multi-unit companies, a composite probability of selection was determined by summing the probabilities that had been assigned to each establishment of the company, and dividing the result by the square root of the number of units. In this way, a single probability of selection was assigned to each multi-unit company. Any company that was selected was required to report for all its manufacturing establishments.

The sampling procedure was modified for the apparel industries. Employment, which is a good general size index for most industries, is not a satisfactory size measure for the apparel industries. Apparel jobbers who frequently have substantial product shipments, usually have very few employees, because their production labor costs are covered by their cost of contract work per

Likewise, many apparel establishments are both jobbers and direct producers, and therefore have a typically high shipments relative to employment. Accordingly, the apparel establishments were assigned sampling rates on the basis of their annual shipments in preference to their employment size.

March employment differs substantially from the average annual employment for some seasonal industries. In those cases, the available measures of size tended to be systematically too high or too low. Accordingly, adjustments for such industries were introduced SO that they were sampled on the basis of their equivalent mean annual employment.

Stratified, random systematic sampling was used throughout; i. e. after each unit had been assigned its probability of being selected, all units from the same sub-universe that had been assigned the same probability were grouped together and sequenced by industry and by State within industry; then beginning with a random start, every "n'th unit was selected depending upon the specified sampling rate.

(NOTE: For a detailed description of the sampling procedures used for the 1952 and earlier annual surveys, see Appendix A in the 1952 Annual Survey volume.)

4. Canvassing Method Used in the Annual Survey

The report forms for the annual surveys were distributed and, for the most part, returned by mail. An important exception to this procedure was made in the case of the small, highly mobile Eastern sawmills which were canvassed by field enumeration.

The collection, the screening for completeness and consistency, and the follow-up of unreturned reports for all single-unit and small multi-unit companies were handled by the various Field Offices of the Bureau for the 1949-1952 surveys. For the large multi-unit companies, these operations were conducted directly by the Census Bureau in Washington. To facilitate the screening of reports, the field offices were provided with a series of editing checks patterned after those used by respondents on the report form. (See Form MA-100 Item 13.) Those manufacturers who failed to send in reports and those whose reports were revealed by these checks to be incomplete or otherwise inadequate, were contacted by letter, telephone, or, in a relatively few instances, by personal visit. For the 1953 Annual Survey, however, the entire collection and processing operation, with few exceptions, were handled by the Census Bureau directly from Washington.

5. The Annual Survey Establishment Report Form a. Comparison with the 1947 Census Form:

A single standard report form, MA-100, is sent to all establishments included in the annual surveys (instead of almost 200 different forms used in the census). This report form represents, for the most part, an abridgement of the standard

schedule used in the Census of Manufactures. The basic statistical measures of manufacturing activity, such as employment, payrolls, value of shipments, cost of materials, etc., were defined in essentially the same way as in the 1947 Census. Consequently, historical series that are published in this report may be considered to be comparable, except for those few cases noted in the tables. Three 1947 Census inquiries--analysis of payroll for the period ended nearest October 15, electric energy generated and sold, and highway-type motor vehicles in use--were dropped entirely, and one new inquiry--products bought and resold without further manufacture--was added to the annual survey form. Other important changes that were made in the report form were: (1) the provision for reporting value of products shipped in terms of approximately 1,000 product classes instead of some 6,500 individual products, as was the case in the 1947 Census; (2) the use of product class reference lists, on the basis of which respondents entered on the report forms (in addition to the value of shipments figure), brief descriptions and code numbers of the product classes (this reference list replaced the preprinted product descriptions and codes appearing on the census forms); and (3) the addition of a check list of questions at the end of the form to enable respondents to examine their figures for completeness and accuracy before submitting their completed reports.

b. Subsequent Revisions in the Annual Survey Report Form:

The 1953 establishment report form differed in the following details from those used in one or more earlier annual surveys: (1) The Force account construction workers" item, which was included as a separate functional category of employment on the 1952 and 1951 report forms, was combined with "all other" employees in 1953; (2) merchandising activity (i. e. cost and value of resales) was treated as a separate item on the 1949 and 1950 establishment report forms, but the item was "'integrated" with the materials cost and products shipped sections of the 1953, 1952, and 1951 report forms; (3) data on the quantity of electric energy used (i.e., purchased plus generated, minus sold) was requested on the 1953 report, while the earlier forms requested data only on the cost of electric energy purchased; (4) inventories data for 1953 were obtained for three categories ("finished products," "work-in-process, and "materials, supplies, fuel, etc.") while in previous years work-in-process and materials were combined as a single item; (5) industrial water intake was included on the 1953 report form for the first time

(respondents were requested to report the quantity of water used and the principal source of their fresh water supply); (6) the metals consumed inquiry, was included in the 1953 and 1950 report forms but was omitted in the 1951 and 1952 surveys; (7) in the products shipped inquiry, an additional column was provided on the 1952 and 1953 report forms, in which the respondent was requested to report the value of shipments for the preceeding, as well as for the current, year.

c. "Self-Editing" Checks for Completeness and Consistency:

A check-list of five "logical'' questions was included in the 1952 and earlier annual survey report forms to assist the respondent in editing his own figures for completeness and consistency. For 1953, this technique was extended to the point of requesting the respondent to calculate four key "editing" ratios: average hourly wage rate; hours worked during the year per production worker; labor cost per dollar of sales; and materials cost per dollar of sales. This would enable him to determine whether his report was consistent and accurate in its essential elements. Although no comprehensive statistical study has been made of the effectiveness of this technique, there is a definite indication that it resulted in improvement in the quality of reporting. There was an exceedingly high rate of response to this item.

d. Product Classes and the Product Class Reference List:

The limited funds alloted for conducting the Annual Survey required that a single products shipped inquiry be substituted for the more than 200 different product forms used in the 1947 Census, each containing a preprinted list of products. This drastic simplification of the products shipped inquiry made it necessary to use a "reference list,"containing the approximately 1,000 principal classes of products made in manufacturing industries, in order to assist respondents in reporting their shipments of products. The product class reference list also was used to facilitate "selfcoding' of product information by respondents. They were asked to assign the appropriate 5-digit class codes shown on the reference list to their products and to report the value of shipments for such products.

The Annual Survey of Manufactures product classes are generally comparable to those published for the 1947 Census. In 1947, the approximately 6,500 individual products for which statistics were collected, were classified into about 1,000 product classes. Each class comprises a relatively homogeneous grouping of individual products associated primarily with a given industry.

The annual survey product classes are similarly grouped according to the industry primarily responsible for their output. Thus, the

This group of products is said to be the "primary" products of the establishment; all other products made by the establishment classified in the particular industry are referred to as "secondary" products.

While some establishments produce only the primary products of the industry in which they are classified, this is rarely true of all of the establishments in an industry. The general statistics (employment, payrolls, value added by manufacture, etc.) shown for an industry, therefore, reflect not only the primary activities of the establishments in that industry, but also their activities in the manufacture of secondary products and, in some instances, their auxiliary activities as well. For a description of each of the manufacturing industries and (1) the extent to which establishments in each manufacturing industry specialized in producing only the primary products of the industry and (2) the proportion of the U. S. total value of output of these primary products accounted for by the establishments of the industry, see Appendix C, "Industry Descriptions,'' 1947 Census of Manufactures, Volume I, II, or III.

In some of the borderline areas between manufacturing and other major divisions of the Standard Industrial Classification system, such as retail trade, wholesale trade, or construction, manufactured products are often made by establishments primarily engaged in these nonmanufacturing activities. Examples of such products are: venetian blinds, awnings, millwork, dried fruit, ice, prepared feed, poultry dressing, housefurnishings, canvas products, and dairy products. Since no reports were obtained in the Annual Survey from establishments primarily engaged in nonmanufacturing activities, national totals for these products as published are understated to the extent of such nonmanufacturing production.

b. Historical Shifts in Plant Classification:

The classification of establishments reporting a change in major activity from the previous year was modified in the 1953 Annual Survey from earlier annual survey procedures. In the annual surveys covering the years 1949-1952, fluctuations in industry classification of plants from one year to another were minimized by a modification of the rule for classifying establishments on the basis of the principal products reported for the current year. Small establishments (generally those with fewer than 100 employees) were retained in their previous (historic) classification; changes in industry classifications were generally limited to the larger plants (defined as having 100 or more employees) which reported a significant change in major activity. Significant changes were defined by a quantitative rule that tested the effect changing the industry code of an establishment on both historical comparability and current level of activity. (See "Historical Shifts in Classification'' in the General Explanations section of 1952 Annual Survey volume.)

of

For the 1953 Annual Survey of Manufactures, all establishments were classified on the basis of their principal activity during 1953. This change in procedure was found to be desirable because of the fact that (1) the change in sample design introduced new reporting plants, many of which had not been assigned industry classifications covering their activities in recent years; (2) the procedure followed in the 1949-1952 Annual Surveys of applying the historic-current code test generally had a negligible effect on the industry totals published. It should be noted that this change in procedure does not affect the shipments data published for classes of products, which are tabulated on a "wherever made' basis (that is, without regard to the industry classification of the establishments reporting such shipments of the product class).

c. Combinations or Regroupings of SIC industries:

In 1947 and in the succeeding annual surveys of manufactures covering the years 1949-1953, some combinations and re-groupings of SIC industries were instituted. These re-groups of SIC industries were made when (1) industries were found to be of relatively little economic significance (2) to avoid disclosing data for individual companies (3) to avoid publishing industry totals where "overlapping" production of products primary to two industries in the typical plant was particularly significant and a breakdown of such plant data into two separate establishment reports is not usually feasible.

A list of the combinations or re-groupings of SIC industries between 1947 and 1953 are shown in Appendix A.

7. Coordination of Census-BOASI Employment Data

a. Background of coordination:

The Bureau of the Census and Bureau of Old-Age and Survivors Insurance both publish statistics on employment and payrolls prepared from different employer reporting forms. Since OASI forms are primarily for administrative purposes and Census forms are for statistical purposes, each source has unique characteristics of scope, information reported, and period covered by the reports, but both contain pay-period employment data.

With the publication of 1947 Census of Manufactures figures on number of manufacturing establishments and employment and the corresponding 1947 OASI data, it became apparent that there were significant differences between the two series. Generally, OASI data were found to show more manufacturing employment and more establishments. Much of the difference was accounted for

by differences in coverage. Differences also appeared because of variations in industrial classification and definition of establishments and because the BOASI had estimated employment not reported for some manufacturing establishments. In addition, some of the differences could be ascribed to the fact that OASI employment figures covered the mid-March payroll, while Census employment represented an average of the 12 monthly figures for 1947. There were, however, apparent discrepancies between the two series at the State and county levels, as well as by industry classification, which could not be accounted for by these differences of scope or reporting period. a result, both Bureaus agreed to coordinate the two series in order to provide the Government and the public with comparable and consistent industry statistics.

As

The program began with coordination of 1947 Census of Manufactures data with 1947 OASI data for manufacturing industries, and has continued for subsequent years with coordination of OASI and Annual Survey of Manufactures data. The goal has been to minimize differences between the two series by using the same industry codes for identical manufacturing units and to develop a comparable reporting unit structure. This program of coordination of the records of the two Bureaus eliminated many of the past differences and has improved both sets of data, a matter of importance for both Bureaus, for the Government in general, and for the public. The extensive coordination which has been achieved is an important step toward the goal of an improved and more unified Federal statistical program.

has

b. Coordination of Data for Years Before 1953

The first coordination efforts by the two Bureaus were directed primarily at the industry coding problem for the mass of single-unit establishments which were mostly small units. Attention was next focused on establishments of multi-unit concerns, because comparison of the statistical data showed that the two Bureaus were often receiving conflicting establishment reports from some of the largest employers. Because of such inconsistencies there were important differences in the industry codes assigned to these establishments.

Effective coordination in manufacturing required a case-by-case comparison of the total number of employees reported and of the industry and geographic codes. In order to carry out this coordination, regular employees of the BOASI were sworn in as Census special agents, and thus became subject to the laws protecting the confidentiality of data reported to the Bureau of the Census. Similarly, to retain the confidentiality of the OASI records, selected Census employees were authorized to work on the OASI file. Such a project was first undertaken for about 30,000 of the largest establishments in the 1949 Annual Survey of Manu

The 1950 coordination procedure was designed to take advantage of the results of the 1949 case-by-case coordination. For some companies that had been coordinated in the 1949 comparison, the Census employment data were used as a basis for adjustment. There was a careful review of all large companies and of all companies which had reported inconsistently to the two agencies for 1949. In some instances, the problems were resolved by correspondence with the employer. An additional comparison of all establishments with over 100 employees (accounting for over 70 percent of manufacturing employment) was made between Census and OASI 1950 and 1949 data. Intensive checks were made of differences detected in this comparison. By this means, significant processing and respondent errors for large establishments were virtually elminated from the 1950 data.

The 1951 coordination again included a case-by-case check of all manufacturing establishments with over 100 employees in the Annual Survey, as well as smaller establishments of multi-unit employers. Intensive study was made of differences detected in this comparison. The objectives and methods of adjustment followed the principles established in previous years, but much of the work of identifying and classifying problems for study was performed by machine. Again, many problems were resolved by communicating with the employers.

c. Coordination of 1953 Data

Returns from the 1953 Annual Survey of Manufactures were similarly coordinated with 1953 OASI records. A comprehensive case-by-case check of 1953 data was made for all large establishments. Discrepancies were resolved by review of reported information, and, in some cases, by contacting employers.

The net difference between the Annual Survey of Manufactures and County Business Patterns manufacturing employment total for 1953 is trivial (only 3,000 or about one-fiftieth of one percent). This net difference obscures a larger gross difference. This gross difference, however, is also very small and within the sampling error of the Annual Survey of Manufactures estimates. The percentage net and gross differences are somewhat larger for individual industry groups and States, but are still not significant.

For a comparison of Census - BOASI employment totals, by State and by major industry group, see the introductory text of 1953 County Business Patterns, Section "Comparability With Other Data, Part II, "Manufacturing Employment Data in the Census Bureau Annual Survey of