Supporting_Statement_B_(1220-0032)_2017

Annual Refiling Survey

1220-0032

August 2017

Supporting Statement to Request for OMB Approval

Annual Refiling Survey

B. Collection Information Employing Statistical Methods

1a. Universe

The universe for the ARS survey includes all private sector employers subject to State UI laws in all 50 states; Washington, D.C.; Puerto Rico; and the Virgin Islands. It also includes Federal agencies whose employees are covered by UCFE. This will result in a universe of approximately 10.1 million private sector business establishments. The most current Quarterly Census of Employment and Wages (QCEW) files for each State serve as the ARS sampling frame.

b. Sample Size

The survey was designed so that roughly one-third of the approximately 10.1 million private sector reporting units would be sampled each year. However, due to the Consolidated Appropriations Act 2014, the BLS, in coordination with the Department of Labor (DOL) and OMB, curtailed the QCEW program through a permanent budget reduction of $4.8 million. This curtailment was accomplished, in part, by reducing the scope of and changing the approach to the ARS. To develop a reduction strategy with minimal impact on QCEW data quality, an analysis considered how frequently establishments changed their industry, location, and ownership. The analysis looked at these changes over three ARS cycles (nine years) and showed that there were industries that experienced low rates of change. Thus, certain NAICS industries exist that traditionally have low volatility based on factors such as infrastructure, intellectual property, or advanced degrees that make them unlikely candidates for transitioning from one industry to another. Rather than surveying one-third of businesses in these low-change industries every year, the QCEW program will survey one-sixth of these slow-changing businesses each year, putting them on a six-year cycle. The six-year cycle preserves the ability to identify trends in these industries and to maintain the accuracy of industry and geographic codes while reducing costs and workload.

2a. Sample Allocation and Selection Procedures

The ARS sampling frame used by BLS will be the most current QCEW file. A schedule was developed to balance the workload as evenly as possible so that around one-third of the businesses are surveyed each year. Certain NAICS industries that traditionally have low volatility will be put on a six-year cycle as described previously.

The selection criterion is based on the Federal Employer Identification Number (EIN) of the respondent, which provides a more random sample across all industries. The seventh and eighth positions of the nine-digit EIN are examined with ranges set for respondents to be surveyed. If the EIN is blank and unavailable, a range of digits within the UI account number will be used.

b. Estimation Procedure

These data are being collected to ensure that the longitudinal QCEW data and the QCEW Report have the appropriate industrial and geographical codes. After data are edited and reconciled for apparent inconsistencies and completeness, these data will be the micro data on the longitudinal QCEW data. Estimates of totals at higher tabulation levels for the QCEW Report will be the sum of the appropriate micro level data. Since this is a census, the estimates are not subject to any sampling error. Therefore, no standard errors will be calculated.

c. Accuracy

Because a census is being conducted, the estimates are not subjected to any sampling error.

d. Problems

There are no unusual problems requiring specialized sampling procedures.

e. Frequency

This survey will consist of a three-year cycle, with approximately one-third of all business establishments sampled each year. Establishments in certain NAICS industries that traditionally have low volatility will be put on a six-year cycle. Establishments assigned a NAICS code 999999 (Unclassified Establishments) are surveyed annually.

3a. Response

The States are required to attain a 70% response rate in establishments or 80% in employment.

If necessary, two non-response follow-up mailings of the ARS letters as well as telephone contact of key non-respondents are used to increase the response rates.

During FY 2016, all States were able to attain either a 70% response rate in establishments or 80% in employment. The vast majority of States attained both goals. The average total response rate was 80.2% in establishments and 82.1% in employment.

b. Non-response Adjustment

Establishments that fail to respond after extensive follow-up efforts will be assigned NAICS codes by staff that have carefully reviewed establishment information gathered from company websites, business directories, and similar reputable sources.

c. Reliability

Because this survey is a census, no sampling errors will be calculated.

To control non-sampling errors, quality control procedures are incorporated into the survey's design. These procedures include follow-up of all non-respondents and validation of all edit failures. The States and the BLS regional offices also receive industrial coding training and procedural assistance in conducting this survey. Regional offices conduct yearly quality assurance reviews in each State.

Automated ARS processing systems are used for several tasks, such as selecting the employers to be surveyed each year, preparing mailings, and editing the data. These systems further reduce the incidence and impact of non-sampling errors.

4. Tests

Work is proceeding on a text analysis tool called the NAICS autocoding tool. The tool will utilize inputs such as Legal and Trade Names, employer-provided information submitted via ARSweb written descriptions, and other pertinent items to determine the likelihood of a particular NAICS code appropriate for an establishment. Additional supplementary data like employment, wages, and firm age could also be used to refine this NAICS autocoding tool. Potential uses for the tool would be for our State partners to assign NAICS codes to unclassified units, aid in decreasing nonresponse, and to identify units that could be misclassified. Development efforts are in the early stages but BLS has hopes to pilot the autocoding tool in the near future. This effort is consistent with BLS continually pursuing options to reduce employer burden and costs and to take advantage of technological innovations. BLS plans to put the tool into production upon successful completion of the development and testing phases.

5. Statistical Responsibility

Mr. Larry L. Huff, Division Chief, Statistical Methods Staff, Office of Employment and Unemployment Statistics is responsible for the statistical aspects of this survey.

3