An official website of the United States government
Data from the Quarterly Census of Employment and Wages (QCEW) are aggregations of various kinds of business establishment data, including geographical, industry, ownership, and establishment size data. QCEW data are not estimates. On occasion, when business establishment data are of poor quality or missing entirely, they may be imputed. (See Design section for more information on imputation in QCEW data.) Proration is used to handle multiple-establishment employers for whom the top employment and wage levels are known but distribution at the establishment level is unknown.
QCEW data are validated at the U.S. Bureau of Labor Statistics (BLS) during a pre-established data review period after the completion of data collection and editing conducted by state workforce agency staff. Edits reduce approximately 10.5 million records to a manageable level. Establishments with no change or a statistically insignificant change to their economic data are ignored. Economic data types include employment, wages, industry, and county. QCEW criteria for reviewing individual establishment records are based on over-the-year changes to employment or wage levels for the establishment relative to the total employment level of the county it is located in. Data are validated to keep the number of establishment records that require review at a reasonable level while ensuring that detailed, county-level aggregations remain accurate. Program analysts review only employment and wage changes that exceed the criteria. Any large-scale changes that the analysts find in economic data need to be verified or corrected by working with state workforce agency staff. One of several methods used to validate significant changes is to seek corroboration with total wage-record counts. Wage records, provided directly by the employer, can be used to confirm both current levels and year-ago levels. The availability of wage record counts varies from state to state. As necessary, state staff contact respondents to obtain corrected data.
As a result of this data review process, BLS and state staff make corrections as needed before publication. These corrections can be made to any quarters for the current year. Additionally, in the first quarter, the previous year’s data can be corrected for the last time. Data that do not meet BLS publication standards may be suppressed.
The aggregated totals of employment and wages for each subdomain (e.g., industry, geography, and size) of the QCEW are simply the sum of the establishments belonging to that subdomain. Averages and other statistics for each subdomain are derived by performing the appropriate arithmetic functions.
As mentioned, the BLS role is to impose quality on the raw data. One of the processes for doing so involves editing the data and conducting validation checks. The basic monthly employment edit consists of a six-step statistical test that includes the use of multiple t-tests for month-to-month changes, over-the-year changes, and a 12-month variation in data; some tests are conducted on levels while others are conducted on rates of change.1
Although BLS receives QCEW files from all 53 entities in a timely manner, the files contain estimates for late and missing respondents. Therefore, one step in the data process is to estimate the number of late respondents, the number of missing respondents (i.e., unit nonresponse) and of the number of missing data elements (i.e., item nonresponse). As shown in table 2 of the Design section, as of June 2019 about 4 percent of establishments failed to respond to the QCEW in a timely manner and thus required imputation; the corresponding percentage for employment in that same month and year was about 2 percent, as shown in table 3 of the Design section. The nonresponse rate for wages was about 4 percent in the first quarter, 2019, as shown in table 4 of the Design section.
The current method of imputation applies, to the missing establishment, the change from a year earlier to the previous month’s employment or quarterly wages in order to estimate the current month’s employment or quarterly wages. That is, the current month’s employment for a missing establishment is equal to the previous month’s employment multiplied by its change from a year earlier; a similar procedure is applied to estimate total quarterly wages. A drawback to this procedure is that it uses the trend from a year earlier rather than the current trend.2
BLS has conducted extensive research on alternative imputation methods for both employment and wages. The findings of the research indicate that current trends exhibited by the reported data from similar cells should be applied to nonrespondents. BLS defines the procedure for doing so as the ratio method. According to this method, the ratio for a particular estimation cell is computed as the sum of a current month’s reported employment divided by the sum of the previous month’s reported employment. To impute the current month’s employment for a nonrespondent, the ratio is then multiplied by the nonrespondent’s previous month’s employment. A similar procedure is applied to impute average quarterly wages. The ratio method of imputation will be implemented in the new QCEW processing system.3
Another data-processing step is to link the QCEW data across quarters for various purposes, including (1) editing and imputation; (2) separating establishments into new establishments (openings or births), continuous establishments (existing businesses), and out-of-business establishments (closings or deaths); and (3) performing longitudinal research.4
While collecting data, analysts may see large increases or declines in employment levels within an establishment. These changes may be verified by identifying a predecessor or successor establishment, respectively. Any such change may be due to a number of factors, including a merger between companies, the acquisition of one company by another, improved reporting by a multiple-establishment employer, and a physical relocation of employees. Often, states contact the employer in question to verify the shifts in the data and get an explanation as to why they occurred.
QCEW program analysts may also seek to validate data by asking questions of the State Workforce Agency staff. The questions usually are about an unexpected or large change in the economic or administrative data of a specific establishment or group of establishments. The agency provides the analysts with edited unemployment insurance (UI) data and often has firsthand knowledge of the changes to the data.
Breaks in published data—sudden shifts in employment or wage levels at the macrolevel—can occur for a number of reasons. One major reason is a change in coding, due to either a physical relocation of an establishment, a change in primary economic activity, a change in industry definition, or the correction of a reporting error. Another reason is a change in the reporting status of an establishment. Some businesses with multiple establishments incorrectly identify themselves as a single unit. Eventually, if they are able to provide a breakout of economic and administrative detail for all of their subunits, it turns out that many of these units are in different counties and may require different industry codes.
Both adjusted over-the-year growth rates for the third month of the quarter and average weekly wages are published in the QCEW County Employment and Wages quarterly news release. These growth rates are not published anywhere else. The over-the-year changes in employment and wages are adjusted to account for most of the administrative corrections made to the underlying establishment reports. Adjustments are made by modifying the previous-year levels used to calculate the over-the-year changes. Over-the-year percent changes are calculated with the use of an adjusted version of the final, unpublished quarterly data of the previous year as the base data. The unpublished previous-year levels do not match the unadjusted data maintained on the BLS website. Over-the-year changes based on data from the website or from data published in previous BLS news releases may differ substantially from the over-the-year changes presented in the QCEW news release.
The adjusted data used to calculate the over-the-year changes presented in the QCEW news release account for most of the administrative changes: those occurring when employers update the industry, location, or ownership information of their establishments. The most common administrative adjustments are the result of updated information about which particular county a given establishment is located in. Included in these adjustments are administrative changes involving the classification of establishments whose county was previously reported as “unknown” or simply “statewide” or whose industry was reported as “unknown”. The classification “statewide” is used primarily for multiple-establishment employers with locations in multiple counties. It appears on the accounts master record that aggregates all establishment data. The classification is sometimes used by establishments whose economic activity has no primary location. The classification “unknown” is used by a state that is unable to identify the physical location of an establishment. Beginning with the first quarter of 2008, adjusted data account for administrative changes caused by multiple-establishment employers that submit reports for each of their establishments rather than reporting as a single entity. Beginning with the second quarter of 2011, adjusted data account for selected large administrative changes in employment and wages. These new adjustments allow the QCEW to include county employment and wage growth rates in the news release that would otherwise not meet publication standards.
The adjusted data used to calculate the over-the-year changes presented in any County Employment and Wages news release are valid for comparisons between the starting and ending points (a 12-month period) used in that particular release. Comparisons may not be valid for any period other than the one featured in a release even if the changes were calculated by using adjusted data.
Finally, in accordance with BLS policy, data reported under a promise of confidentiality are published in a way so as to protect the identifiable information of respondents. BLS withholds the publication of UI-covered employment and wage data for any industry level when necessary to protect the identity of employers. Totals at the industry level for the states and the nation include the undisclosed data suppressed within the detailed tables without revealing those data. QCEW confidentiality concepts and practices are largely based on the "Statistical Policy Working Paper 22" (PDF) developed by the Federal Committee on Statistical Methods.
For more information about Confidentiality concepts, as it relates to QCEW data disclosure, see this document about Confidentiality and Data Disclosure.
1 The wage edit includes the use of an interquartile test developed by David Hoaglin, Boris Iglewicz, and John Tukey, “Performance of some resistant rules for outlier labeling,” Journal of the American Statistical Association, December 1986, pp. 991–999, https://www.jstor.org/stable/2289073?seq=1#metadata_info_tab_contents. The edit conditions and formulas are described in “Appendix-F: Edit conditions and formulas,” QCEW Operating Manual (U.S. Bureau of Labor Statistics, 2007), www.reginfo.gov/public/do/DownloadDocument?objectID=48010401.
2 The imputation formulas used by BLS are described in chapter 8, “Imputation of Missing and Delinquent Data,” and Appendix J of the QCEW Operating Manual (U.S. Bureau of Labor Statistics, 2007), www.reginfo.gov/public/do/DownloadDocument?objectID=48010401.
3 For details of the method, including various exceptions, see ICR documents (Office of Information and Regulatory Affairs, Office of Management and Budget, Executive Office of the President), http://www.reginfo.gov/public/do/PRAViewDocument?ref_nbr=201406-1220-001.
4 Details of the methodology are given in Ivan P. Fellegi and Alan B. Sunter, “A theory for record linkage,” Journal of the American Statistical Association, 1969, vol. 64, no. 328, pp. 1183–1210; and Kenneth Robertson, Larry Huff, Gordon Mikkelson, Timothy Pivetz, and Alice Winkler, “Improvement in record linkage processes for the Bureau of Labor Statistics’ Business Establishment List,” pp. 212-221 in Wendy Alvey and Bettye Jamerson, eds., Record linkage techniques—1997, https://www.nap.edu/read/6491/chapter/10#212: Proceedings of an International Workshop and Exposition, March 20–21, 1997, Arlington, VA. Washington, DC: Federal Committee on Statistical Methodology, Office of Management and Budget, 1997. For more information on establishment linkage, births, and deaths, see “Business Employment Dynamics,” Handbook of Methods (U.S. Bureau of Labor Statistics), https://www.bls.gov/opub/hom/bdm/home.htm.