Reconstruction of CES time series: implementing the NAICS 2022 redefinitions

The Current Employment Statistics (CES) program is a federal–state cooperative program between the U.S. Bureau of Labor Statistics (BLS) and state workforce agencies. The CES State and Area program produces data on employment, hours, and earnings for the 50 states, the District of Columbia, Puerto Rico, the U.S. Virgin Islands, and more than 450 metropolitan areas. The program publishes timely economic indicators each month—usually available 3 to 5 weeks after the reference period—by surveying approximately 122,000 businesses and government agencies that represent about 666,000 individual worksites.¹

This article explains how the reclassification of industries from NAICS 2017 to NAICS 2022 has caused CES to revise many of its series. Many CES series are left unchanged, but a select few experienced larger changes. Starting with the definition of NAICS codes and CES series, this article will cover which industries and codes had the most changes as well as the methods and procedures that CES used to make those changes, with examples and case studies throughout. Finally, the article ends with how the reclassifications from NAICS 2017 to NAICS 2022 affect seasonal adjustment, the CES business birth-death model, non-all employee measures, and the continuity of CES series, in the context of historical comparisons.

NAICS industry definitions

The North American Industry Classification System (NAICS) classifies establishments (individual business locations) by their type of economic activity.² Mexico’s Instituto Nacional de Estadística y Geografía, Statistics Canada, and the U.S. Office of Management and Budget jointly developed NAICS in 1997 and continue to collaborate on NAICS to make the industry statistics produced by the three countries comparable. The three countries review NAICS definitions every 5 years to determine what, if any, changes are required to best capture economic activity and to ensure the accuracy, timeliness, and relevance of industry classifications.

NAICS codes are six digits, with each additional nonzero digit indicating a more detailed subindustry. The first two digits indicate the broad sector of the economy, the third the subsector, the fourth the industry group, and the fifth the NAICS industry. The sixth digit is the national industry digit specific to the United States, Mexico, or Canada. In NAICS, establishments that use comparable capital equipment, labor, and raw material inputs are classified together. BLS collects this information in a supplement to the quarterly unemployment insurance tax reports filed by employers through the Quarterly Census of Employment and Wages (QCEW).³ Establishments operate at a single location and generally engage predominantly in one type of economic activity. If a single location has two or more distinct activities that could be classified under different industry codes and if those activities have separate payroll records (such as a mall or ghost kitchen), then each activity would be treated as a separate establishment despite sharing a location. These NAICS codes form the foundation of the CES industry codes.

CES sometimes combines one or more NAICS industries into a single CES industry code.⁴ This is done primarily when CES does not have enough responses for a given six-digit NAICS code to produce reliable estimates or to prevent disclosure of payroll data for individual businesses.⁵ For example, in some states and areas, CES only publishes wholesale trade overall (CES series 41-000000) instead of publishing series for all six-digit NAICS codes within wholesale trade.⁶ The hierarchy of CES publishable series is referred to as the CES series structure.

Reclassification of industries in NAICS 2022

The NAICS 2022 update resulted in revisions reflecting content and coding changes for seven industries within the scope of CES: mining and logging, manufacturing, wholesale trade, retail trade, information, financial activities, and other services.⁷ The most significant change—which had major impacts on the retail trade and information sectors—with NAICS 2022 was the switch in emphasis from delivery method to product delivered.⁸ For example, in NAICS 2017, brick-and-mortar retail trade establishments had different industry codes than online retail trade establishments because their method of delivering products to customers was different. With the rise of e-commerce in the past few decades, the distinction between brick-and-mortar retailers and nonstore retailers has become less clear. A single retail establishment now often sells merchandise both in physical stores and online. During the most recent NAICS review and update, the distinction between store and nonstore retailers was removed; instead, industries are classified on the basis of the products they retail, not their methods of sale.⁹ For example, under NAICS 2017, a shoe retailer with a physical location would be classified under shoe stores (4482), while one with an online-only business would be classified under electronic shopping and mail-order houses (454110). With NAICS 2022, both fall under shoe retailers (4582).

This principle of reclassification by product instead of sales method also extends to information services, in which establishments in internet publishing are now broken out by publication type. For example, NAICS 2017 code 515 included establishments in broadcasting, except for those using the internet as a primary delivery method. In contrast, NAICS 2022 code 516, the successor to 515, no longer includes the exception for internet broadcasters.

NAICS 2022 impact on CES series structure

Most of the NAICS 2022 changes did not impact CES data series. For example, bituminous coal and lignite surface mining (NAICS 2017 code 212111) was assigned the equivalent surface coal mining (NAICS 2022 code 212114). Because CES does not publish industries at this level of detail for any state or area, there was no impact on the CES series structure.

Other NAICS changes, however, did impact CES series structure.¹⁰ For example, publishing industries except internet (NAICS 2017 code 511) changed to publishing industries (NAICS 2022 code 513), which now includes portions of what was internet publishing and broadcasting and web search portals (NAICS 2017 code 519130). As a result, CES series 50-511000 had to be redefined because it was made up entirely of publishing industries except internet (NAICS 2017 code 511) and was published in multiple states and areas.

Because of the changes to retail establishment classification, in addition to other smaller scale NAICS changes, the CES State and Area series structure was affected in several ways:

• A few series experienced only changes to the official CES title of the series.
• Many series saw compositional changes with relatively small employment differences caused by the switch to NAICS 2022. For retail establishments, this largely resulted in code or title changes with little to no change in published employment.
• Some series were added to or dropped from CES publication. These additions or deletions were mainly caused by either collapses or breakouts of NAICS 2017 codes.

CES prioritized preserving series continuity as much as possible by converting CES series based on NAICS 2017 definitions to their equivalent NAICS 2022 definitions, rather than adding or dropping published industries (see table 1). For some series, this was challenging. For example, the series furniture, home furnishings, electronics, and appliance retailers (NAICS 2022 code 449) is made up of furniture and home furnishings stores (NAICS 2017 code 442) and electronics and appliance stores (NAICS 2017 code 443). Some states and areas were publishing CES series 42-442000 and 42-443000 that were based on the NAICS 2017 codes. If a state or area published both 42-442000 and 42-443000, these series were updated into the equivalent CES series that are based on the NAICS 2022 codes. Therefore, CES code 42-442000 became 42-449100 and 42-443000 became 42-449200 in the 2022 CES structure. The CES series 42-449000 (furniture, home furnishings, electronics, and appliance retailers) was then added as an aggregate of the 2022 series 42-449100 and 42-449200. If a state or area published only one of 42-442000 or 42-443000, the existing series was deleted and replaced with series 42-449000 (as shown in table 2).

Time series reconstruction methodology

CES offers time-series data so that data users can perform comparable longitudinal analyses across the economy. Ideally, CES series histories should be minimally affected by administrative changes such as NAICS reclassifications so that any movements in employment can be correctly attributed to true underlying economic activity. To ensure this directive was met with the implementation of NAICS 2022, time series histories were reconstructed to reflect the updated NAICS definitions.

Because of existing publication detail within CES and differences in publication structure across states and metropolitan areas, CES used a variety of methods to create series histories. For example, in the case in which a retired NAICS 2017 series did not have a clear NAICS 2022 successor series, CES constructed complete histories from establishment level microdata. In the case in which there was a clear successor series, CES modified existing histories using ratios or other methods to bring them in line with NAICS 2022 definitions. Because of the nature of changes and heterogeneity of series availability across CES areas, CES often used multiple methods to create series histories.

CES used its existing series histories as much as possible in the reconstruction of the new NAICS 2022-based time series. Series with simple one-to-one changes, such as code or title changes, had their histories carried forward entirely. For example, wired telecommunications carriers (NAICS 2022-based series 50-517311) was derived entirely from wired telecommunications carriers (NAICS 2017-based series 50-517111). If data at the detailed industries level were needed to reconstruct the history of a higher leveled series, then, CES used QCEW data in a process that has previously been used to reconstruct series.¹¹That process, with examples, is discussed next.

The CES program tallied the sum of QCEW employment for each industry, county, township, and ownership level. While the vast majority of QCEW records have all this identifying information, there are some missing values. If there are employment data without a county or town code, CES distributes those employment records proportionally to the amount of employment in each county and town, for every NAICS and ownership code. If there were employment data that had a county or town code but did not have a corresponding NAICS codes, CES distributed them proportionally to other industries within the corresponding county or town. Records that lacked both NAICS and county or town codes were distributed to counties and towns on the basis of the location’s proportion of total CES-assigned employment within the state and then distributed proportionally to all industries.¹²

For more complex changes, a combination of two or more of the following elements was used to construct the new series’ history:

• The existing history for a component series in part or in whole
• A generated history for a previously unpublished series
• The ratio of employment moving between two given NAICS codes. These redefinition ratios were calculated at the state and area level as a proportion of summed employment moving from a redefined NAICS 2017 code to another NAICS code (whether new in 2022 or preexisting)

For example, the NAICS 2017 definition of other information services (519) included four component industries: news syndicates (519110), libraries and archives (519120), internet publishing and broadcasting and web search portals (519130), and all other information services (519190). Under the NAICS 2022 definitions, the name of the broader industry (other information services) was changed to web search portals, libraries, archives, and other information services. Also, instead of having four component industries, it now had only two: libraries and archives (519210) and web search portals and all other information services (519290). Furthermore, the NAICS 2017 code for internet publishing and broadcasting and web search portals (519130) was distributed into eight NAICS 2022 codes, only one of which remained under 519 (web search portals and all other information services, 519290). Because the CES California statewide series for other information services was published (50-519000), historical data for the series needed to be reconstructed. Table 3 illustrates the redefinitions under 519.

To reconstruct series 50-519000 on a NAICS 2022 basis, the CES program used QCEW microdata and created a history for NAICS 2017 news syndicates (519110), which did not have any definitional changes between 2017 and 2022. The CES program also created a history for the NAICS 2017-based internet publishing and broadcasting and web search portals (519130). Because this industry had definitional changes between 2017 and 2022, the CES program calculated the ratio of the created internet publishing and broadcasting and web search portals (50-519130) history that did not redefine into web search portals and all other information services (519290). The new 50-519000 history was calculated by taking the original history and subtracting the created news-syndicates history and the portion of the internet publishing and broadcasting and web search portals history that was excluded from NAICS 2022 code 519290. The result of this change is evident in chart 1, which shows a downward shift in California statewide employment in series 50-519000.

CES series 50-519000 was not alone in having to change because of technological advancements over the last decade. The shift in focus from the method of product delivery to type of product delivered resulted in extensive breakouts of electronic shopping and mail-order houses (454110 in NAICS 2017) and other direct-selling establishments (454390 in NAICS 2017) into other retail trade industries. CES did not publish either industry under the previous series structure for states or areas. A key concern with this breakout was to maintain the relationship between the series within retail trade. Where nonstore retailers (CES series 42-454000) was published, the program created histories for its four NAICS 2017 components: electronic shopping and mail-order houses (454110), vending machine operators (454210), fuel dealers (454310), and other direct selling establishments (454390). These histories were proportionally distributed between other retail trade NAICS industries according to redefinition ratios calculated using the method described above. Vending machine operators (454210 in NAICS 2017) and fuel dealers (454310 in NAICS 2017) were substantially one-to-one reconstructions into vending machine operators (445132 in NAICS 2022) and fuel dealers (457210 in NAICS 2022), respectively. Electronic shopping and mail-order houses (454110 in NAICS 2017) and other direct selling establishments (454390 in NAICS 2017) breakouts led to small adjustments to nearly all other retail series. An example of the magnitude of this type of change is shown in chart 2.

Other impacts on CES

While most of the impact of the NAICS 2022 implementation was on the re-creation of series histories, other aspects of the CES program were affected by the reclassification.

Seasonal adjustment

The impact on seasonally adjusted data was limited since the scope of the redefinitions were confined to series that CES does not seasonally adjust at the state and area levels.¹³

Business birth–death model

Birth-death processing required changes at the three-digit NAICS level where X-13ARIMA SEATS modeling occurs.¹⁴ The CES program produced new historical birth-death data frames for the redefined NAICS codes to use in the annual forecasting process. The forecasted factors produced with the updated frames were comparable with those produced in previous years under NAICS 2017 definitions.

Impact on non-all employee data types

In addition to all-employee (AE) estimates of nonfarm workers, CES also produces estimates of several non-AE data types, including production workers, average hourly earnings, average weekly hours, and average weekly earnings. Hours and earnings data for AE and production workers are also produced at differing levels of industry detail for different states and areas. In areas in which non-AE data were published, CES reconstructed non-AE data series at a level of detail such that the effects from the NAICS 2022 redefinitions were incorporated. A weighted-link-and-taper estimator was used to create these time series. This estimator accounts for the over-the-month change in the sampled units but also includes a tapering feature to keep the estimates close to the overall sample average over time.¹⁵ The estimates were simulated using existing matched sample data from October 2010 forward.¹⁶ Because sample data prior to October 2010 are less reliable, ratios of the previously published series were carried forward for the reconstructed series from the series’ start year through September 2010.

Conclusion

The NAICS 2022 redefinitions were wider in scope than the 2017 redefinitions and required careful consideration with the reconstruction process to preserve the integrity of CES historical data.

Preserving the unchanged histories of total nonfarm data was a top priority for CES during the reconstruction process. Ideally, a given area with multiple changes to its series structure would show zero net effect on aggregate total nonfarm employment level because NAICS redefinitions simply reclassified already existing employment across industries.

A combination of existing histories, ratios of documented employment movement, and QCEW data were used to map historical data in redefined industries while maintaining aggregate and residual relationships.