Labor force projections are based on expectations of the future size and composition of the population, as well as on the trends in labor force participation rates of different age, gender, race, and ethnic groups, a total of 136 separate categories.
The U.S. Census Bureau prepares projections of the resident population. The size and composition of the population are affected by the interaction of three variables: births, deaths, and net immigration. More information about population projections is available on the Census Bureau web site. BLS converts these population projections to the civilian noninstitutional population concept, a basis for labor force projections. BLS develops participation rate projections using data from the Current Population Survey (CPS) conducted for BLS by the Census Bureau.
For this latest round of projections, the Census Bureau’s 2012 population projections based on Census 2010 population weights were used as the base for the labor force projections. The size and composition of the population affect not only the labor force projections, but the projected composition of GDP and the demand for workers in various industries and occupations.
BLS currently disaggregates the various race and ethnic categories into 5-year age groups by gender. Participation rates for these groups are smoothed, using a robust-resistant nonlinear filter and then transformed into logits. The logits of the participation rates are then extrapolated linearly by regressing against time and then extending the fitted series to or beyond the target year. When the series are transformed back into participation rates, the projected path is nonlinear.
After the labor force participation rates have been projected, they are reviewed from the perspectives of the time path, the cross section in the target year, and cohort patterns of participation. The projected participation rate for each age, gender, race, and ethnicity group is multiplied by the corresponding projection of the civilian noninstitutional population to obtain the labor force projection for that group. The groups are then summed to obtain the total civilian labor force. The labor force outlook plays a critical role in long run macroeconomic trends and is therefore the most important exogenous data within the BLS macroeconomic projections.
BLS’ macroeconomic projections are produced using the MA/US model, licensed from Macroeconomic Advisers, LLC (MA). The 2012—2022 projections are the first to employ the new model, which was introduced in late 2012; previously the Bureau relied on MA’s Washington University Macro Model (WUMM). MA/US has the same foundations as WUMM: consumption follows a life-cycle model and investment is based on a neoclassical model. Foreign sector estimates rely on forecasts from Oxford Economics. However, many improvements were made; most notably, the model is explicitly designed to reach a full-employment solution in the target years. Within MA/US, a sub-model calculates an estimate of potential output from the nonfarm business sector, based upon full-employment estimates of the sector’s hours worked and output per hour. Error correction models are embedded into MA/US to align the model’s solution with the full-employment submodel.
Certain critical variables set the parameters for the nation’s economic growth and determine in large part the trend that GDP will follow. In developing the macroeconomic projections, BLS elects to determine these critical variables through research and modeling, and then supplies them to the MA/US model as exogenous variables. The in-house labor force projections, described above, are of particular importance, as they are the primary constraint on future economic growth. Other fundamental exogenous variables in the model include energy prices and assumptions about fiscal and monetary policy. For the 2022 projections, initial estimates of key economic variables, as well as the underlying exogenous assumptions, were reviewed by a panel of Federal economists. The final solution was evaluated for consistency with the detailed output and employment projections. The specific assumptions and target variables for the 2022 projections are presented in the December 2013 Monthly Labor Review.
Demand is the key determinant in explaining future jobs. Therefore, underlying the projections of employment by industries and occupations, BLS publishes a projected final demand matrix consisting of roughly 120 demand categories by 200 commodity groups. Aggregate gross domestic product (GDP) as well as some underlying subcomponent categories are determined by the MA/US model and serve as constraints to BLS’ more detailed projections of GDP. Solutions supplied by MA/US include the aggregates projections of: personal consumption expenditures (PCE), private investment in equipment and software (PIES), residential and nonresidential construction, change in private inventories (CIPI), exports and imports of goods and services, as well as consumption and investment of federal defense, nondefense, and State & Local government. BLS uses several behavioral models as well as distributional trends and/or assumptions in breaking out these twelve categories of GDP supplied by the macro model to the detailed matrix of final demand data.
Personal consumption expenditures (PCE) and Private Investment in Equipment and Software (PIES) are projected using the Houthakker-Taylor model(1) and the Modified Neoclassical Model of Investment. Projections are made for 76 PCE product categories and 28 PIES asset groups consistent with the national income and product account (NIPA) data published by the Bureau of Economic Analysis (BEA).
These column totals for PCE and PIES are adjusted as necessary to ensure consistency between the aggregate projection from the MA/US solution and the detailed estimates based on the Houthakker-Taylor and Modified Neoclassical models. Column totals for the remaining components of final demand are output directly from the MA/US software. Although net exports is equivalent, adjustments are made to the MA/US projections of trade goods and services to account for re-exports and re-imports, effectively transforming the data from a NIPA based estimate to an I-O framework.
Bridge tables are developed based on the most recent Benchmark and annual Input-Output Accounts published by BEA. For some columns the bridge table is held constant from the last historical year while other components forecast the bridge table based on trends over the historical series. The bridge tables are used to distribute projected column controls out to the roughly 200 commodity groups or rows within the final demand matrix. The only exception to the use of bridge tables occurs in the column of Change in Private Inventories (CIPI). Business inventories by detailed commodities are projected based on a two stage least squares model where inventories are regressed on lagged values of both inventories and commodity output. Results are aggregated and adjusted to conform to the projection of CIPI from the macro model.
As a last step, data are converted from purchaser value to producer value. Margin columns are projected for each component of final demand. Summing across the rows of a particular component (ie. PCE) with its related margin columns (consisting of transportation costs as well as wholesale and retail markups), results in a vector of producer value data by detailed commodity. For example, in buying a sweater, the margin column would subtract the retail markup by the vendor from the textile commodity row and move that value to the retail trade commodity row. In estimating employment, it is helpful to know the producer value of the data as it helps separate employment in the wholesale, retail, and transportation industries from the remaining economy.
Adjustments to the initial estimates of the final demand matrix are made based on research and analysis by industry experts including information pertaining to energy forecasts, existing and expected shares of the domestic output, known changes to trade agreements, expected government political and policy changes, and so forth.
Industry output is derived using a set of projected input-output tables consisting of two basic matrices for each year, a "use" and a "make" table. The use table consists of final demand, from the preceding step, together with intermediate demand and value added. The use table shows the use of commodities by each industry as inputs into its production process. The make table allocates commodity output to the industry in which it is the primary commodity output and to those industries in which it is secondary. In percentage form, the use table provides the direct requirements table and the make table becomes a market share table. These two tables are then used to create total requirements tables which yield the projected levels of industry and commodity output required to satisfy projected final demand.
The next step is to project the industry employment necessary to produce the projected output. To do so, projected output is used in regression analysis to estimate hours worked by industry. The regression model utilizes industry output, industry wage rate relative to industry output price, and time. Additionally, average weekly hours are derived as a time trend for each industry. From these hours’ data, projected wage and salary employment by industry is derived.
For each industry, the share of self-employed and unpaid family workers is extrapolated using historical data. These data are derived from the ratio of self-employed and unpaid family workers to total employment and extrapolated based on time and the unemployment rate. The ratio, along with the projected level of wage and salary employment is then used to derive the projected number of self-employed and unpaid family workers and total employment by industry. Projected average weekly hours and total hours for self-employed and unpaid family workers also are derived from these data.
Implied output per hour (labor productivity) is calculated for each industry for both the total and for wage and salary employees. These data are used to evaluate the projected output and employment.
Factors Affecting Industry Employment
Many assumptions underlie the BLS projections of the aggregate economy and of industry output, productivity, and employment. Often, these assumptions bear specifically on econometric factors, such as the aggregate unemployment rate, the anticipated time path of labor productivity, and expectations regarding the Federal budget surplus or deficit. Other assumptions deal with factors that affect industry-specific measures of economic activity.
Detailed industry employment projections are based largely on econometric models, which, by their very nature, project future economic behavior on the basis of a continuation of economic relationships that held in the past. For the most part, the determinants of industry employment are expressed both in the structure of the models’ equations and as adjustments imposed on the specific equations to ensure that the models are indeed making a smooth transition from actual historical data to projected results. However, one of the most important steps associated with the preparation of the BLS projections is a detailed review of the results by analysts who have studied recent economic trends in specific industries. In some cases, the results of the aggregate and industry models are modified because of the analysts’ judgment that historical relationships need to be redefined in some manner.
Table 2.7 Employment and Output by Industry presents historical and projected information about employment and output for aggregate and detailed industries. Industry sector employment projections prepared in the Division of Industry Employment Projections (DIEP) used comprehensive modeling techniques that estimate output as well as employment.
To allocate projected industry employment to occupations, a set of industry-occupation matrices are developed. These include a base-year employment matrix for 2012 and a projected-year employment matrix for 2022. These matrices, referred to collectively as the National Employment Matrix, constitute a comprehensive employment database. For each occupation, the Matrix provides a detailed breakdown of employment by industry and class of worker. Similarly, for each industry and class of worker, the Matrix provides a detailed breakdown of occupational employment.
Base-year employment data for wage and salary workers, self-employed workers, and unpaid family workers come from a variety of sources, and measure total employment as a count of jobs, not a count of individual workers. This concept is different from that used by another measure familiar to many readers, the Current Population Survey’s total employment as a count of the number of workers. The Matrix’s total employment concept is also different from the BLS Current Employment Statistics (CES) total employment measure. Although the CES measure is also a count of jobs, it covers nonfarm payroll jobs, whereas the Matrix includes all jobs.
The Matrix does not include employment estimates for every industry which employs an occupation, or every occupation employed within an industry. Some data are not released due to confidentiality and/or quality reasons(2). Employment data in the National Employment Matrix are presented in thousands. Detailed data may not sum to totals because of rounding or because some data are not released.
2012 Base-Year Employment
For most industries, the Occupational Employment Statistics (OES) survey provides data for the occupational staffing patterns—the distribution of wage and salary employment by occupation in each industry—and Current Employment Statistics (CES) data provide information on total wage and salary employment in each nonfarm industry. Estimates of occupational employment for each industry are derived by multiplying each occupation’s proportion–or ratio–of employment in each industry, based on OES survey data, by CES industry employment.
BLS staff obtains industry and occupational employment data for workers in all agricultural industries except logging(3), workers in private households, self-employed workers, and unpaid family workers from the Current Population Survey. Data are used for workers’ primary job only. CPS data are coded using the 2010 Census occupation classification system. Although the Census system is based on the Standard Occupational Classification (SOC) system used by OES, it does not provide the same level of detail. CPS employment data were proportionally distributed to detailed SOC occupations using the employment distribution from the OES data.
Total base-year employment for an occupation is the sum of employment across all industries and class-of-worker categories—the combination of wage and salary, self-employed, and unpaid family workers. Occupational employment within each industry, divided by total wage and salary employment in each industry, yields the occupational distribution ratios used to project occupational employment. These ratios are referred to as staffing patterns.
2022 Projected-Year Employment
Projected-year employment data for industries and class-of-worker categories are first developed at a higher level of aggregation, and then distributed to corresponding detailed Matrix industries and by class of worker.
To derive projected-year staffing patterns, BLS economists place base-year staffing patterns under an iterative process of qualitative and quantitative analyses. They examine historical staffing pattern data and conduct research on factors that may affect demand for occupations within given industries during the projection decade. Such factors include shifts in product mix, and changes in technology or business practices. Once these factors are identified, change factors are developed which give the proportional change in an occupation’s share of industry employment over the 10-year projection period. These change factors are applied to the 2012 occupational staffing patterns to derive projected staffing patterns. An occupation’s projected share of an industry may increase, decrease, or remain the same, depending on the change factors and underlying rationales.
For each industry, the projected-year employment is multiplied by the projected-year occupational ratio to yield projected-year wage and salary occupational employment for the industry. Occupational employment data for self-employed and unpaid family workers are projected separately. Total projected-year occupational employment is the sum of the projected employment figures for wage and salary, self-employed, and unpaid family workers.
Factors Affecting Demand for Occupations within Industries
BLS projections of wage and salary employment are developed within the framework of an industry-occupation matrix, which shows the occupational distribution in each industry—the proportion of each industry’s employment which each occupation comprises. Historical data indicate that the occupational distribution within industries shifts over time as the demand for some occupations changes relative to that of other occupations.
Among the various factors that can affect the demand for workers in an occupation in particular industries are technology, business practices, the mix of goods and services produced, the size of business establishments, and offshore outsourcing. BLS staff analyzes each occupation in the matrix to identify the factors that are likely to cause an increase or decrease in demand for that occupation within particular industries. Their analyses incorporate judgments about new trends that may influence occupational demand, such as the use of the Internet and electronic commerce.
Estimating Replacement Needs
Projections of job growth provide valuable insight into future employment opportunities because each new job created is an opening for a worker entering an occupation. However, opportunities also result when workers leave their occupations and need to be replaced. In most occupations, these replacement needs provide more job openings than employment growth does. Further detail is presented in Estimating Occupational Replacement Needs technical documentation.
BLS provides information about education and training requirements for hundreds of occupations. In the BLS education and training system, each of the occupations for which the office publishes projections data is assigned separate categories for education, work experience, and on-the-job training. Occupations can be grouped in order to create estimates of the education and training needs for the labor force as a whole and estimates of the outlook for occupations with various types of education or training needs. In addition, educational attainment data for each occupation are presented to show the level of education achieved by current workers. Further detail is presented in Measures of Education and Training technical documentation.