The National Compensation Survey (NCS) program produces information on wages by occupation for many metropolitan areas and also for the nation as a whole. Wage rates for different work levels within occupational groups also are published. Data are available for private industry, State and local governments, full-time workers, part-time workers, and other workforce categories.

The Bureau of Labor Statistics (BLS) selects a sample of establishments in both the private sector and State and local governments in over 150 metropolitan areas and non-metropolitan counties throughout the United States. Data are collected from a sample of establishments in a specified area over a period of a few months. NCS data are used by both private companies and government agencies to help set pay and to compare their wages with those of other employers. The data are a major component of the pay setting process for Federal Government employees. Economic and academic researchers also are interested in the findings.

The NCS Wage Query System presents mean hourly wage data for all workers in an occupation (private industry and State and local government combined). Currently, data may be obtained for over 85 specific geographical areas. Data are also available for the nation as a whole, and for its 9 broad geographic divisions.

The Wage Query System now has the ability to produce modeled hourly wage estimates for many area, occupation, and work level combinations for which BLS has no officially published data. Published estimates are those tabulated directly from the collected data. All published estimates have been reviewed and meet BLS publication standards.

For some occupations, however, estimates for many work levels do not meet BLS publication criteria. In such cases, BLS provides data generated from a statistical model that predicts hourly wage estimates for those unpublished levels. A regression formula is used to produce coefficients for each category - area, occupation, and work level. A modeled hourly wage estimate results from summing these three coefficients. The coefficients are derived annually from national wage data.

It is important to consider the nature of modeled estimates produced by this application carefully relative to published survey-based estimates. First, the modeled estimates should be considered experimental. Second, modeled estimates are only produced in cases where the sample size was too small to tabulate a published estimate. Finally, there is no accurate measure of error tied to the modeled estimates. Work on the model and on measures of the quality of its estimates is ongoing and should eventually strengthen our ability to model hourly wages.

For more information on this subject see Pierce, Brooks, "Using the National Compensation Survey to Predict Wage Rates," Compensation and Working Conditions, Winter 1999, pp. 8-16 or visit CWC Online at http://www.bls.gov/opub/cwc/1999/Winter/contents.htm.