Introducing Modeled Wage Estimates by grouped work levels (Monthly Labor Review)
Joana Allamani, Michael Hudak, Adam Issan
The Modeled Wage Estimates (MWE) provide annual estimates of average hourly wages for occupations by job characteristics and within a given geographical location. These estimates are produced by borrowing from the strength and breadth of the Occupational Employment and Wage Statistics (OEWS) and National Compensation Survey (NCS) programs to provide more details on occupational wages than either program provides individually. Job characteristics refer to the attributes of workers within an occupation and include worker bargaining status (union and nonunion), work status (part-time and full-time), basis of pay (incentive-based or time-based), and work level (levels 1–15). In this article, we present experimental estimates calculated by grouping work-level data. Grouped level estimates may help researchers, human resources professionals, jobseekers, and other data users to get a better understanding of how pay varies for entry, intermediate, and experienced work levels.
The U.S. Bureau of Labor Statistics (BLS) first introduced Modeled Wage Estimates (MWE) in the 2013 article, “Wage estimates by job characteristic: NCS and OES program data,” and the first set of estimates were produced in 2016. Since then, MWE data have been published annually and provide average hourly wage estimates for occupations by job characteristics and geographical locations. These estimates are produced by combining data from two BLS programs—the Occupational Employment and Wage Statistics (OEWS) and the National Compensation Survey (NCS).
The occupational and geographic wage data from the OEWS are combined with the job characteristics: bargaining status (union and nonunion), work status (part-time and full-time), basis of pay (incentive-based and time-based), and work level wage data from the NCS. By drawing on the strengths of each of the surveys, the MWE provide more information on occupational wages in geographic areas by job characteristics than each survey can provide individually. This article introduces grouped level estimates by providing the methodology used. The full set of experimental grouped work-level estimates including the relative standard errors and a listing of grouped work levels for each occupation and associated break level are available for download.
The “Guide for Evaluating Your Firm's Jobs and Pay” provides an explanation of how the BLS determines work levels in the NCS. Work levels provide insight into differences in compensation due, in part, to four job factors—the knowledge necessary, job controls and complexity, nature and purpose of contacts, and the physical environment for or of the job. Each factor consists of several levels, with an associated description and assigned points. The total points from the four factors are used to determine the work level for occupations. With the exception of the knowledge factor, a common scale is used for all occupations. . . .
Why is it important to group levels?: As wages tend to increase along with the progression in work level, stakeholders have expressed interest in understanding the differences in pay for entry, intermediate, and experienced work levels. In the labor market, wage progression resulting from more experience, credentials, or complexity of tasks differs by occupation. Using the individual work levels does not allow for comparison of entry level, intermediate level, and experienced level across occupations. But grouping the work levels and allowing them to vary by occupation allows for these broad comparisons. Even though it is possible for data to be available for 15 levels for a given occupation, it generally does not occur as illustrated by the knowledge point differences by occupational family. That is, work levels may be clustered; for example, leveling data for food preparation workers are concentrated within the 1–5 level range. For electro-mechanical technicians, an occupation that typically requires an associate degree or a postsecondary certificate, leveling data are available in the 4–9 range, and for chief executives the 11–15 range. Given the difference in work-level ranges, the NCS program evaluated grouping work levels to allow for additional evaluation of compensation differences.
The Jenks optimization method is an algorithm that identifies the optimal number of groups from a domain (in this case occupations). To do this, the sum of squared deviations from the average (mean) wages is calculated for grouped work levels based on the cluster that minimize the sum of within-group deviations. To simplify the evaluation of work-level clusters, two (to represent entry and experienced levels) and three groups (to represent entry, intermediate and experienced levels) were generated for each detailed occupation (six-digit 2010 Standard Occupational Classification code). In addition to the Jenks natural breaks, the NCS program evaluated the difference in log wages between groups and share of employment. These additional criteria identified occupations that should not be leveled, as meaningful distinctions between groups could not be determined. It also identified occupations that should be estimated using two or three work-level groups. Occupations like food preparation workers, waiters and waitresses, or nursing assistants could only be grouped into the entry and experienced levels, as additional meaningful wage differences were not found beyond these two groupings. For occupations such as loan officers, meaningful wage distinctions were found when using three groups. In 2019, work levels were available for loan officers from level 6 to level 12. As shown in table 1, the wages for levels 6–8 were very similar as were wages for levels 9 and 10. By grouping the work levels using the Jenks optimization method, users can understand the compensation progression of occupations. The associated relative standard errors provide an indication about the reliability of the estimates. Although comparing the individual work levels would provide more granular information, this is not always possible, especially in more detailed areas. The NCS sample is not sufficiently large to provide wage estimates for each work level. By grouping the work levels, more data are available to produce the estimates. . . .
Conclusion: BLS provides a variety of wage data through programs such as the OEWS, Current Population Survey (CPS), the Current Employment Statistics survey (CES), and the NCS. The average hourly wages by work levels provided by the MWE further allows users to examine the compensation factors of workers in the U.S. economy. The addition of grouped work levels helps compare average hourly wages by facilitating the identification of occupations with similar factors. The NCS program continues to evaluate the usefulness and availability of the MWE. These grouped work levels represent an initial attempt to provide grouped work levels based on the Jenks optimization method.
Looking ahead, the NCS program will reevaluate the grouped work levels for each occupation, in particular because of the Modeled-Based Estimation Methodology (referred to as MB3) and 2018 SOC implementation by the OEWS and MWE. In addition, another set of experimental estimates will be published through a factsheet on the MWE website before adding them into the annual production process. This will allow users to provide feedback on the grouped work levels and allow the NCS to consider approaches to publish more data.
MLR article: https://www.bls.gov/opub/mlr/2022/article/introducing-modeled-wage-estimates-by-grouped-work-levels.htm
MWE website: https://www.bls.gov/mwe
National Compensation Survey: Guide for Evaluating Your Firm's Jobs and Pay,” May 2013 https://www.bls.gov/ncs/ocs/sp/ncbr0004.pdf
For further information and to provide feedback, contact:
Joana Allamani firstname.lastname@example.org
Michael Hudak email@example.com
Adam Issan firstname.lastname@example.org