A number of industries are stratified by size of establishment the and/or by region, and stratified production- or nonsupervisory-worker data are used to weight the hours and earnings into broader industry groupings. Accordingly, the basic estimating cell for an employment, hours, or earnings series, as the term is used in the summary of computational methods, may be a whole industry or a size stratum, a region stratum, or a size stratum of a region within an industry.

Benchmark adjustments

Employment estimates are compared periodically with comprehensive counts of employment which provide "benchmarks" for the various nonagricultural industries, and appropriate adjustments are made as indicated. The industry estimates are currently projected from March 1971 levels. Normally, benchmark adjustments are made annually.

The primary sources of benchmark information are employment data, by industry, compiled quarterly by State agencies from reports of establishments covered under State unemployment insurance laws. These tabulations, cover nearly ninetenths of the total nonagricultural employment in the United States. Benchmarks data for the residual are obtained from the records of the Social Security Administration, the Interstate Commerce Commission, and a number of other agencies in private industry or government.

The estimates relating to the benchmark month are compared with new benchmark levels, industry by industry. If revisions are necessary, the monthly series of estimates are adjusted between the new benchmark and the preceding one, and the new benchmark for each industry is then carried forward progressively to the current month by use of the sample trends. Thus, under this procedure, the benchmark is used to establish the level of employment; the sample is used to measure the month-to-month changes in the level. A comparison of the actual amounts of revisions made in the last 3 benchmark years is shown in table G.

Design

The sampling plan used in the current employment statistics program is known as "sampling proportionate to average size of establishment." This design is an optimum allocation design among strata since the sampling variance is proportional to the average size of establishments. The universe of establishments is stratified first by industry and then within each industry by size of establishment in terms of employment. For each industry, the number of sample units is distributed among the size class cells on the basis of average employment per establishment in each cell. In practice, this is equivalent to distributing the predetermined total number of establishments required in the sample among the cells on the basis of the ratio of employment in each cell to total employment in the industry. Within each noncertainty stratum the sample members are selected at random.

Under this type of design, large establishments fall into the sample with certainty. The size of the sample for the various industries is determined empirically on the basis of experience and of cost considerations. In a manufacturing industry in which a high proportion of total employment is concentrated in relatively few establishments, a large percentage of total employment is included in the sample. Consequently, the sample design for such industries provides for a complete census of the large establishments with only a few chosen from among the smaller establishments or none at all if the concentration of employment is great enough. On the other hand, in an industry in which a large proportion of total employment is in small establishments, the sample design calls for inclusion of all large establishments and also for a substantial number of the small ones. Many industries in the trade and services divisions fall into this category. To keep the sample to a size which can be handled by available resources, it is necessary to accept samples in these divisions with a smaller proportion of universe employment than is the case for most manufacturing industries. Since individual establishments in these nonmanufacturing divisions generally show less fluctuation from regular cyclical or seasonal patterns than establishments in manufacturing industries, these smaller samples (in terms of employment) generally produce reliable estimates.

In the context of the BLS employment and labor turnover statistics programs, with their emphasis on producing timely data at minimum cost, a sample must be obtained which will provide coverage of a sufficiently large segment of the universe to provide reasonably reliable estimates that can be published promptly and regularly. The present sample meets these specifications for most industries. With its use, the BLS is able to produce preliminary estimates each month for many industries and for many geographic levels within a few weeks after reports are mailed by respondents, and at a somewhat later date, statistics in considerably greater industrial detail.

Coverage

The BLS sample of establishment employment and payrolls is the largest monthly sampling operation in the field of social statistics. Table H shows the approximate proportion of total employment in each industry division covered by the group of establishments furnishing monthly employment data. The coverage for individual industries within the division may vary

annually to new benchmarks. In addition to taking account of sampling and response errors, the benchmark revision adjusts the estimates for changes in the industrial classification of individual establishments (resulting from changes in their product which are not reflected in the levels of estimates until the data are adjusted to new benchmarks). In fact, at the more detailed industry levels, particularly within manufacturing, changes in classification are the major cause of benchmark adjustments. Another cause of differences, generally minor, arises from improvements in the quality of the benchmark data. Table J presents the average percent revisions of the six most recent benchmarks for major industry divisions. Detailed descriptions of individual benchmark revisions are available from the Bureau upon request.

The hours and earnings estimates for cells are not subject to benchmark revisions, although the broader groupings may be affected slightly by changes in employment weights. The hours and earnings estimates, however, are subject to sampling errors which may be expressed as relative errors of the estimates. (A relative error is a standard error expressed as a percent of the estimate.) Relative errors for major industries are presented in table J and for individual industries with the specified number of employees in table K. The chances are about 2 out of 3 that the hours and earnings estimates from the sample would differ by a smaller percentage than the relative error from the averages that would have been obtained from a complete census.

Table J. Average benchmark percent revision in employment estimates and relative errors1 for average weekly hours and average hourly earnings by industry division

Relative errors (in percent)

Telephone

Telegraph

Reliability of the employment estimates

Although the relatively large size of the BLS establishment sample assures a high degree of accuracy, the estimates derived from it may differ from the figures that would be obtained if it were possible to take a complete census using the same schedules and procedures. As discussed under the previous section, a link relative technique is used to estimate employment. This requires the use of the previous month's estimate as the base in computing the current month's estimate. Thus, small sampling and response errors may cumulate over several months. To remove this accumulated error, the estimates are adjusted

For the two most recent months, estimates of employment, hours, and earnings are preliminary and are so footnoted in the tables. These figures are based on less than the total sample and are revised when all the reports in the sample have been received. Table L presents root-mean-square errors of the amounts of revisions that may be expected between the preliminary and final levels of employment and preliminary and final month-tomonth changes. Revisions of preliminary hours and earnings estimates are normally not greater than .1 of an hour for weekly hours and 1 cent for hourly earnings.

STATISTICS FOR STATES AND AREAS

State and area employment, hours, earnings, and labor turnover data are collected and prepared by State agencies in cooperation with BLS. The area statistics relate to metropolitan areas. Definitions for all areas are published each year in the issue of Employment and Earnings that contains State and area annual averages (usually the May issue). Changes in definitions are noted as they occur. Additional industry detail may be obtained from the State agencies listed on the inside back cover of each issue. These statistics are based on the same establishment reports used by BLS for preparing national estimates. For employment, the sum of the State figures may differ slightly from the equivalent official U.S. totals on a national basis, because some States have more recent benchmarks than others and because of the effects of differing industrial and geographic stratification.

For the States and the areas shown in the B and C sections of this periodical, all the annual average data for the detailed industry statistics currently published by each cooperating State agency are presented (from the earliest data of availability of each series) in a summary volume published annually by the BLS.

Insured unemployment represents the number of persons reporting a week of unemployment under an unemployment insurance program. It includes some persons who are working part time who would be counted as employed in the payroll and household surveys. Excluded are persons who have exhausted their benefit rights and workers who have not earned rights to unemployment insurance. In general, excluded from coverage are those persons engaged in agriculture, domestic service, unpaid

family work, selected nonprofit organizations, some State and local government and self-employment. Also excluded from the insured unemployment count, but included as employed in the household survey, are those persons who earned no wages during the payroll period because they were temporarily absent from their jobs due to taking time off, illness and industrial dispute as well as unpaid vacations. The rate of insured unemployment is the number of insured unemployed expressed as a percent of

average covered employment in a 12-month period ending 6 to 8 months prior to the week of reference. Initial claims are notices filed by those losing jobs covered by an unemployment insurance program that they are starting a period of unemployment. A claimant who continued to be unemployed a full week is then counted in the insured unemployment figure.

Because of differences in State laws and procedures unde

which unemployment insurance programs are operated, State unemployment rates generally indicate, but do not precisely measure, differences among the individual States. Persons wishing to receive a detailed description of the nature, sources, inclusions and exclusions, and limitation of unemployment insurance data should address their inquiries to Manpower Administration, Washington, D.C. 20210.

Many economic statistics reflect a regularly recurring seasonal movement which can be estimated on the basis of past experience. By eliminating that part of the change which can be ascribed to usual seasonal variation, it is possible to observe the cyclical and other nonseasonal movements in the series. However, in evaluating deviations from the seasonal patternthat is, changes in a seasonally adjusted series-it is important to note that seasonal adjustment is merely an approximation based on past experience. Seasonally adjusted estimates have a broader margin of possible error than the original data on which they are based, since they are subject not only to sampling and other errors but, in addition, are affected by the uncertainties of the seasonal adjustment process itself. Seasonally adjusted series for selected labor force and establishment data are published regularly in Employment and Earnings.

The seasonal adjustment methods used for these series are an adaptation of the standard ratio-to-moving average method, with a provision for moving "adjustment factors" to take account of changing seasonal patterns. A detailed description of the method is given in the two publications, BLS Seasonal Factor Method (1966) and X-11 Variant of the Census Method II Seasonal Adjustment Program, Technical Paper No. 15, Bureau of the Census (1967).

Data for the household series are seasonally adjusted by the Census X-11 Method. For each of the three major labor force components-agricultural employment, nonagricultural employment, and unemployment-data for four age-sex groups (male and female workers, 16-19 years and 20 years and over) are separately adjusted for seasonal variation and then added to give seasonally adjusted total figures. In order to produce seasonally adjusted total employment and civilian labor force data, the appropriate series are aggregated. The seasonally adjusted rate of unemployment for all civilian workers is derived by dividing the figure for total unemployment (the sum of four seasonally adjusted age-sex components) by the figure for the civilian labor force (the sum of 12 seasonally adjusted age-sex components). Other series, such as unemployment by duration or employment by major occupational groups, are independently adjusted.

The seasonal adjustment factors applying to current data are based on a pattern shown by past experience. Once each year (in January), these factors are revised in the light of the

previous years' experience. Revised seasonally adjusted series for major components of the labor force based on data through December 1973, plus a short description of the methodology, are published in the February 1974 Employment and Earnings. Many additional series, which are either components or aggregates of the series presented, are available from the BLS upon request.

For establishment data, the seasonally adjusted series on hours, hourly earnings, and labor turnover rates for industry groupings are computed by applying factors directly to the corresponding unadjusted series, utilizing the BLS Seasonal Factor Method. However, seasonally adjusted employment totals for all employees and production workers by industry division are obtained by summing seasonally adjusted data for the component industries. Seasonally adjusted average weekly earnings are the product of seasonally adjusted average hourly earnings and average weekly hours. Average weekly earnings in constant dollars, seasonally adjusted, are obtained by dividing seasonally adjusted average weekly earnings by the seasonally adjusted Consumer Price Index. Indexes of aggregate weekly man-hours, seasonally adjusted, are obtained by multiplying aver age weekly hours, seasonally adjusted, by production or nonsupervisory workers, seasonally adjusted, and dividing by the 1967 base. For total private, total goods producing, total private service producing, trade, manufacturing, and durable and nondurable goods the indexes of aggregate weekly man-hours, seasonally adjusted, are obtained by summing the aggregate weekly man-hours, seasonally adjusted, for the appropriate component industries and dividing by the 1967 base.

The seasonally adjusted establishment data for Federal Government are based on a series which excludes the Christmas temporary help employed by the Postal Service in December. The employment of these workers constitutes the only significant seasonal change in Federal Government employment during the winter months. Furthermore, the volume of such employment may change substantially from year to year because of administrative decisions by the Postal Service. Hence, it was considered desirable to exclude this group from the data upon which the seasonally adjusted series is based.

The revised seasonally adjusted series for the establishment data reflect experience through March 1973. Seasonal factors to be used for current adjustment appear in the June 1973 Employment and Earnings.