How can we improve the reliability of employee compensation official statistics?

Challenge

The U.S. economic indicators are important because they provide a comprehensive picture of the nation’s financial health and influence decisions made by government officials, businesses, and individuals. For some, key components are granular statistics on employee compensation.

The Bureau of Labor Statistics’ Office of Compensation and Working Conditions asked Westat to develop methodology to address the production of reliable employee compensation estimates for large numbers of detailed categories using sparse data collected across multiple studies.

Solutions

Westat researchers developed, evaluated, and tested a novel way to integrate employee compensation data from two surveys. Their multilevel model reconciled the survey estimates for common variables and used the relationship between variables to enable estimation for all the subgroups represented in at least one of the surveys. The precision of each estimate was improved by “borrowing strength” across all the information available.

For this, Westat performed the following activities:

• Conducted a thorough literature review of dozens of small area estimation and statistical data integration articles
• Identified and prepared survey data and specifications to serve as inputs into statistical models
• Developed novel modeling approaches to learn about the predictive power of the data and methods to construct compensation estimates for domains with sparse data
• Adopted existing state-of-the-art approaches for modeling and validation
• Programmed complex data investigation and statistical modeling to implement, evaluate, and test the methodology developed for over 500,000 domains
• Partnered statistics and data science experts with IT colleagues to develop a secure data enclave environment designed to support parallel processing and computationally intensive modeling

This work bridges the statistical fields of data integration and small area estimation. While this work was motivated by a statistical data integration problem, it builds upon small area estimation approaches developed to yield precise estimates for population subgroups with small sample sizes. The proposed model also brings a novel contribution to the small area estimation field, using unmatched sets of population subgroups in the model specification.

Results

We developed a novel methodological approach for producing a massive number of reconciled, reliable employee compensation statistics for domains defined by a combination of geography, occupation, work level, and characteristics (time/incentive, full-time/part-time, union/nonunion). Representing an innovative synthesis of methods, this work has the potential to improve employee compensation official statistics. The model predictions may serve as key data in producing economic indicators, such as the quarterly Employment Cost Index and the Employer Costs for Employee Compensation, of interest to government agencies and institutions.

This work can impact people’s lives by providing more accurate and detailed insights into wage trends, helping workers understand fair compensation for their roles and locations. Policymakers and businesses can use these improved statistics to shape fair labor policies and ensure competitive pay. Ultimately, this can lead to better wage transparency and economic decisions that benefit employees across various industries.

ARTICLES IN PEER-REVIEWED JOURNALS

Erciulescu A.L., Opsomer J.D., and Schneider, B.J. (2022). Statistical data integration using multilevel models. Canadian Journal of Statistics, 51(1), 312-326. https://doi.org/10.1002/cjs.11688.

Erciulescu A.L., and Opsomer J.D. (2022). A model-based approach to predict employee compensation components. Journal of the Royal Statistical Society, Series C, 71(5), 1503-1520. Available at https://doi.org/10.1111/rssc.12587.