Construction of average weekly hours in detailed industries

In measuring output per hour for detailed industries, the Division of Industry Productivity Studies (DIPS) in the Bureau of Labor Statistics (BLS) constructs annual total hours series for each industry. To construct these series, the Bureau relies primarily on data collected in the BLS monthly Current Employment Statistics (CES) survey of U.S. establishments. The CES provides monthly and annual data on total employment, as well as data on employment and average weekly hours of production and nonsupervisory workers, by detailed industry. However, the CES does not collect data on the hours of nonproduction and supervisory workers, and therefore the hours must be estimated. This report describes the construction of these estimates for detailed industries.^[1]

Average weekly hours of supervisory and nonproduction workers in each industry are derived using data from the BLS-sponsored Current Population Survey (CPS), a monthly household survey, and data from the CES establishment survey. Annual ratios of supervisory worker average weekly hours to nonsupervisory worker average weekly hours for each service-providing industry are calculated from the CPS data. Similarly, ratios of nonproduction worker average weekly hours to production worker average weekly hours for each manufacturing and mining industry are also calculated. These ratios are applied to the CES nonsupervisory or production worker average weekly hours, respectively, for the industry to estimate the nonproduction or supervisory worker average weekly hours by industry.

In this effort, the Division of Industry Productivity Studies (DIPS) uses data sources and methods similar to those used by the Bureau’s Division of Major Sector Productivity (DMSP) and described in “Alternative measures of supervisory employee hours and productivity growth" in the October 2004 Monthly Labor Review. While the methods and data are similar, some differences occur in constructing industry-specific estimates. This notice covers those differences.

The Methodology

Productivity and cost measures for detailed industries are maintained back to 1987. The adjusted hours series discussed here were developed for the period 1987 through 2002. They have since been extended to 2003 and will be updated to current years in future updates of the industry productivity measures.

Defining nonproduction and supervisory workers in the establishment survey

Estimates of nonproduction or supervisory worker employment in each industry are derived from the CES establishment survey by subtracting the estimate of production workers, construction workers, or nonsupervisory workers from the total employees in that industry. In the CES, production workers in manufacturing and natural resources and mining refer to employees who engage directly in the production of the establishment’s products; production workers include working supervisors and group leaders who may be in charge of a group of employees, but whose supervisory function is only incidental to their regular work. In construction industries CES data refer to construction workers, while in utilities industries and in service-providing industries, CES data refer to nonsupervisory workers.^[2] Nonsupervisory workers are defined to include most employees except those whose major responsibility is to supervise, plan, or direct the work of others, such as top executive and managerial positions, officers of corporations, department heads, and superintendents.

Defining nonproduction and supervisory workers in the household survey

In the CPS, respondents are not asked to classify themselves as production, nonproduction, nonsupervisory or supervisory workers. They are asked, however, to report on their occupation and employment status. Using this information, we constructed employment counts and worker categories in a manner that is highly consistent with the way employers report on workers for the CES.

The CPS microdata that were used cover all private wage and salary workers who were employed and at work beginning with data for January 1987. We sorted these data into categories of nonproduction (supervisory) or production (nonsupervisory) workers based on industry and occupation codes. For data prior to 2000, we define nonproduction workers in goods-producing industries as those in the following occupational groups:

• Executive, administrative, and managerial occupations
• Professional specialty occupations
• Technical, sales and administrative support occupations

In nongoods-producing industries, supervisory workers are defined as persons in:

• Executive, administrative, and managerial occupations

Beginning with data for 2000, the CPS occupation codes are consistent with the 2000 population census. We define nonproduction workers in goods-producing industries to include the following:

• Management occupations
• Business and financial operations occupations
• Computer and mathematical occupations
• Architecture and engineering occupations
• Life, physical, and social science occupations
• Community and social service occupations
• Legal occupations
• Education, training, and library occupations
• Arts, design, entertainment, sports, and media occupations
• Healthcare practitioners and technical occupations
• Sales and related occupations
• Office and administrative support occupations

In nongoods-producing industries, supervisory workers include:

• Management occupations
• Business and financial operations occupations

Using this information, employment and hours of relevant occupational categories in each industry were grouped to arrive at estimates for production, nonproduction, supervisory, and nonsupervisory workers consistent with the CES data.

Calculating Average Weekly Hours of Supervisory and Nonproduction Workers

Annual estimates of employment and average weekly hours from the CPS household survey were constructed as annual averages of the estimates derived from the monthly survey data. CPS estimates of annual average weekly hours for each industry were constructed by dividing the sum of average weekly hours for each month in a given year by the sum of monthly employment in that industry during the year.

Hours data for production and nonsupervisory workers in detailed industries from the establishment survey are hours paid. At the major sector level, data on the relationship between hours worked and hours paid from other establishment surveys are available to convert hours paid to hours worked. This information is not available for detailed industries. The CPS, however, collects information on hours worked. Because the estimates of hours obtained from respondents in the CPS reflect hours worked rather than hours paid, the measure of the employed used in estimating employment and hours for individual industries from the CPS data was constructed to include those employed and at work during the survey period, rather than all employees paid. The estimates of average weekly hours for detailed industries constructed from the CPS data thus use the ratio of hours worked to employees at work as a proxy for the ratio of hours paid to employees paid.

Adjustments to Historical CPS Data

Multiple Jobholding

A major redesign of the CPS survey occurred in 1994, causing inconsistencies between CPS data collected before and after 1994. In particular, prior to 1994 the monthly CPS household survey data classified individuals only according to the industry and occupation of his or her primary job. In contrast, from 1994 forward the CPS data also provide information on the industry and occupation of secondary jobs, and information on hours worked at both the primary and the secondary job. Beginning in 1994, the CPS data can be combined to be more consistent with the data from the CES establishment survey, which measures jobs, counting a person who is employed by two or more establishments at each place of employment. BLS adjusted the CPS data prior to 1994 for multiple jobholding to be consistent with the data for 1994 forward. Adjustment ratios were constructed from the 1994 CPS data, where employment and hours of multiple jobholders in all twelve months were classified according to both the old and the new method. (Additional information was also available from the May 1985, 1989, and 1991 supplements to the CPS, but too few observations were available at the industry level from these monthly supplements to construct reliable adjustment ratios.) The industry-specific adjustment ratios based on 1994 employment and hours data were applied to the pre-1994 data to make them consistent with the later data.

Industry Coding

Industry employment and hours data from the CES establishment survey are identified with codes pertaining to the North American Industry Classification System (NAICS). In contrast, the Current Population Survey (CPS) uses the Census Bureau’s Industry Classification System (ICS) for coding industries. Depending on the year, an ICS code may represent a detailed industry as classified in the Standard Industrial Classification (SIC) system or the NAICS system, or a combination of several industries. CPS data classified at the most detailed ICS industry level were used wherever possible.

From 1987 to 2002, each ICS code corresponds roughly to a 3-digit SIC code, although in some instances it corresponds to a 2-, 3-, or 4-digit SIC code or combination of several SIC codes. The ICS industry codes are updated in every decennial Census of Population to take into account SIC changes that introduce new industries and eliminate obsolete ones. The period studied, 1987-2002, includes two different SIC-based classification systems. For the years 1987-1991, the ICS is based on the 1980 Census, which uses the 1972 SIC classification, as modified in 1977. The ICS for the years 1992-2002 is based on the 1990 Census which uses the 1987 SIC classification. As a result of the changes made in the SIC and the ICS, some changes to coding were necessary to keep industries consistent throughout the period of the study. The recoding was done as the datasets were created.

In recent years NAICS has replaced the SIC system for reporting U.S. data by industry. In 2003 the CES data were published on a NAICS basis for the first time, and historical series were reconstructed to reflect the new industry structure. The NAICS system was also officially introduced into the CPS data in 2003, and CPS data for 2000-02 are available on both an SIC and a NAICS basis. In order to have consistent historical NAICS CPS series back to 1987 for estimating supervisory and nonproduction worker average weekly hours by industry, the Industry Productivity Studies staff used a multi-step process to convert the historical CPS data from SIC-based ICS codes to NAICS-based codes. Using the dual-coded CPS data for 2000-02, SIC to NAICS conversion ratios were calculated to reflect the percentage of SIC industry employment accounted for by the employment in each NAICS industry. Adjustments to the initial conversion ratios were later made based on a comparison of the NAICS industry employment levels for 2000-02 generated by applying this ratio with the employment estimates from the CPS data provided on a NAICS basis. NAICS final employment and hours estimates for 1987-2002 were derived by applying the adjusted conversion ratios to the historical SIC-based employment and hours series.

Construction of Adjusted Hours Series

Employment, total hours, and average weekly hours series were constructed separately for supervisory (nonproduction) workers and for nonsupervisory (production) workers for each ICS industry for the years 1987-2002 using CPS data. Using these series, the ratios of supervisory worker average weekly hours to nonsupervisory worker average weekly hours and nonproduction worker average weekly hours to production worker average weekly hours were calculated for each of 301 ICS industries at the greatest level of detail available, representing industries at the 3-, 4-, 5-, or 6-digit NAICS level, as well as combinations of industries. Included were 52 industries that were created by combining data for detailed ICS industries to more closely represent a 3-digit NAICS industry.

In using CPS microdata classified by detailed industry, the number of responses collected for each industry in the household survey was an important factor in assessing the reliability of the estimates. The employment size of the industry and the volatility of the constructed annual ratios were also considered. For some industries, the number of observations was insufficient to produce reliable estimates of employment and hours. The level of industry detail to be used in calculating the average weekly hours ratios was based on the number of responses for that industry in the CPS survey. Data for the most detailed industries classified in the CPS were used wherever possible, as long as the average number of responses for the industry was 492 or greater per year. For detailed industries where there were insufficient responses, the average weekly hours ratios were calculated using data for a more highly aggregated industry.

Once the level of industry detail to be used was determined, ratios of supervisory (nonproduction) worker average weekly hours to nonsupervisory (production) worker average weekly hours were calculated from the CPS data. The ratios were then multiplied by the number of nonsupervisory (production) worker average weekly hours from the establishment survey to create supervisory (nonproduction) worker average weekly hours for each industry. The supervisory (nonproduction) average weekly hours are then multiplied by the number of supervisory (nonproduction) workers from the establishment survey to create total supervisory (nonproduction) worker hours for each industry. These hours were later combined with the hours of nonsupervisory (production) workers from the CES and self-employed and unpaid family workers from the CPS to create revised total hours for each industry.

Results

Previous estimates of supervisory and nonproduction worker average weekly hours used in most industry productivity measures published prior to the completion of this research were held constant at levels derived from the discontinued BLS Employer Expenditures for Employee Compensation (EEEC) survey of 1977 (for manufacturing) or from the 1980 Census of Population (for non-manufacturing industries).

Effect on Levels of Supervisory and Nonproduction Worker Average Weekly Hours by Industry

Using the new method for estimating supervisory and nonproduction worker hours changed the average weekly hours levels of these workers in many industries.^[3] In general, the new levels of nonproduction worker average weekly hours tend to be higher than the old levels for goods-producing industries. The new levels of supervisory worker average weekly hours tend to be lower than the old levels for service-providing industries, particularly for industries in retail trade and accommodation and food services. Many of the large revisions in supervisory worker average weekly hours also occurred in service-providing industries, in retail trade in particular. Of the 477 industries, the mean supervisory (nonproduction) worker average weekly hours were revised by 10 hours or more in 27 industries. Of those, 25 (93 percent) were in the service-providing industries and 20 industries (74 percent) were in retail trade.

Effect on Trends in Supervisory and Nonproduction Worker Hours by Industry

The revised supervisory and nonproduction worker average weekly hours also affected the trends in total hours of supervisory and nonproduction workers for many industries.^[4] Of the 477 industries in this study for which labor productivity measures are maintained, the impact of the new measures on the trend in the hours of these workers was minor (a difference of plus or minus 0.1 percent or less in the average annual change between 1987 and 2002) for 148 industries (31 percent). For 195 other industries (41 percent), the average annual change in supervisory and nonproduction worker hours was revised from (±) 0.2 to 0.4 percentage points per year. In 134 industries (28 percent) the trend in hours of supervisory and nonproduction workers was revised by (±) 0.5 percentage points per year or more. In 28 of these industries, the average annual percent change in hours of those workers was revised by one percentage point per year or more. Five of the 28 most affected industries were in the retail trade sector (NAICS 44-45), while five were in the information sector (NAICS 51) and another 11 were in the manufacturing sector (NAICS 31-33).

Concentration of Supervisory and Nonproduction Workers by Industry

The extent to which changes in supervisory and nonproduction worker hours affect total labor hours depends in part on the concentration of supervisory and nonproduction workers in the industry. In general, the concentration of nonproduction workers in manufacturing industries tends to be higher than the concentration of supervisory workers in service industries. Of the 22 industries where nonproduction or supervisory workers exceeded 50 percent of all workers, twenty-one were in manufacturing.

Effect on Trends in Total Labor Hours by Industry

Because supervisory and nonproduction workers are only a portion of all workers in each industry, trends in total industry hours were less affected than the trends in supervisory or nonproduction worker hours. The average annual percent change in total labor hours for 1987-2002 was revised up or down by 0.2 percentage points or more in 95 of the 477 industries examined. However, changes of 0.5 percentage points per year or more occurred in only 8 industries. The largest revision in total hours growth from 1987 to 2002 was in other electrical equipment and components, NAICS 3359, where average annual hours growth increased by almost 0.9 percentage points. Of the 8 industries where trends in total hours were most impacted, 5 are in the manufacturing sector.

Conclusions

With the construction of new supervisory and nonproduction worker average weekly hours estimates for detailed industries, both the levels and the 1987-2002 trends of these series changed for many industries. These changes affected the levels and trends in total supervisory and nonproduction worker hours, which in turn induced changes over the period in total hours of all workers. Change was expected since the previous estimates were based on the assumption that supervisory and nonproduction worker average weekly hours were constant over time.

The trend in hours of supervisory and nonproduction workers was revised by more than half a percentage point per year in about a third of all the industries examined. However, for many industries the new measures did not have a major effect on the trends in total hours. The 1987-2002 average annual percent change in total labor hours was revised up or down by more than 0.5 percentage points in only 8 industries. The impact of the revisions to supervisory or nonproduction worker average weekly hours on the trends in total hours for an industry depended on a variety of factors, including the magnitude of the revisions to supervisory (nonproduction) worker average weekly hours levels, the concentration of supervisory or nonproduction workers in the industry, and the trends in the CPS-based adjustment ratios.

Notes