Unique ID: 2015106

Division: Office of Information and Regulatory Affairs
Issue Date: February 13th 2019
Last modified: February 22nd 2019

Autocoding: using payroll systems data for labor market statistics

Using payroll systems data for labor market statistics

Autocoding job titles to allow receipt of electronic files with job titles and wage levels for each employee that can be converted to the US Standard Occupational Classification system. The goal is for firms to provide data in their format from their payroll / HR records and significantly reduce response burden

Project Objective:

Pilot intended to go to production to supplement existing data, Pilot intended to go to production to improve timeliness, Pilot intended to go to production to replace existing data

Project Outcomes:

Improve response rates, increase the quality of input data on wages (replace wage ranges with point estimates), and provide a basis for converting the sampling design from supporting representative estimates based on 3 years of data to an annual time series.

Publications Comments:

We are working closely with another U.S. Bureau of Labor Statistics office that currently does autocoding

Statistical Area

Criminal justice

Project Sources
Project Sources
Type Of Institution: National statistical office
Big Data Source: Other
Region: North America
Country Area: United States
Id Country Regional: country
Other Partners: Other
Partnership Comments: Data are collected in partnership with State Labor Market Information offices.
Accessing Data
Accessing Data
Data Access Rights: Broader access rights
Data Coverage
Data Coverage
Data Coverage: All available data
Coverage Geo Pop: Whole country / high % of market
Cost Comments: These are survey costs that are being paid for by reducing paper form printing costs.
Coverage Period: On going, starting in 2016
Data Quality
Data Quality
Quality Framework: Quality of output statistics
Quality Aspects Evaluated: Privacy and Security, Accessibility, Relevance, Institutional/Business Environment, Validity, Accuracy, including selectivity, Coherence, including linkability to other sources, Other
Validation Comments: We use standard edit reconciliation processes. Autocoding occupations will use large data bases of prior job title / standard occupational classification matches to determine the efficacy of matches from current data.
Quality Framework Comments: All quality aspects apply. We have a variety of statistical measures on quality for the program, inputs, process and outputs.
Data Quality Concerns Comments: We are always concerned and we plan to examine the quality of our job title matches very closely
Quality Assessment Comments: In addition to these traditional measures, we also have a number of process measures and measures related to the reduction of total survey error.
Methods Used: Machine learning (Random forest, etc.)
Technologies: Other
Technologies Comments: We are developing the code for autocoding job titles in house
Income Level: High-income
Iso: US
Timeframe To Produce Indicator: NA
Write Your Own Review
You're reviewing:Autocoding: using payroll systems data for labor market statistics
Your Rating