Note to Reviewer (1220-0011) Forms Test

October 6, 2010 (revised February 18, 2011)

MEMORANDUM FOR : Reviewer of 1220-0010

FROM : Kenneth W. Robertson, Chief

Division of Current Employment Statistics

Bureau of Labor Statistics

SUBJECT : Feasibility Test of New Collection Form for the Current

Employment Statistics Survey

The Bureau of Labor Statistics (BLS) will begin testing new enrollment, collection, and multi-unit fax forms for the Current Employment Statistics (CES) survey. The forms do not ask for new data items, but are intended to provide improved usability for respondents and increased interviewer efficiency. The new forms were developed under contract with Dr. Don Dillman, who worked with BLS managers and interviewers at the Kansas City Data Collection Center.

Four of our Data Collection Centers (Atlanta, Dallas, Kansas City, and Niceville, FL) will be participating in the test. Six interviewers at each location will participate. Three interviewers at each location will be asked to use the new forms for ten months of enrollment/collection work, while three others will serve as a control group. We expect that this process will yield approximately 4,800 test reports for analysis and the test should last 10 months.

Copies of the new enrollment and collection forms are attached to this memorandum as well as fax versions of the forms. To make the form easier to understand for the respondent, there are minor variations of the instructions for several major industry groupings and between the paper and fax versions of the forms. This is consistent with the current collection forms. Respondents will receive ongoing collection forms and be considered for conversion to an automatic collection method consistent with current practice. We plan to monitor self-reporters from amongst the test and control contingents for a period of at least 4 months to evaluate longer-term reporting, to ensure that there are no negative long-term impacts associated with the new forms design.

The design for the test includes, as mentioned above, a test group and a control group. A systematic stratified random sample design will be used to assign cases to each interviewer, ensuring that paired test/control interviewers get similar panels. The strata will include broad industry groupings, employment size of business, the single/multi-worksite status of the company, and four geographic region designations. Interviewers will be paired by experience level; there will be both a test and control group for senior interviewers, for mid-level interviewers, and for less-experienced interviewers. Forty cases, a normal monthly enrollment caseload, will be assigned to each interviewer in the test and control groups using the design as described. All other reports will be collected using the current CES form for the duration of this test.

At the conclusion of the collection period we will compare the test group paradata to the control group paradata. Response rates and duration of interview (controlled by collected and not-collected status) will be the primary statistics compared across interviewer levels, single/multi status, establishment employment size, and geography. Also, we will evaluate the effectiveness of the new form in obtaining item-level response. Finally, we will be reviewing respondent and interviewer comments about the new form.

The test design is expected to provide a reliable indicator of response rate differences as small as 1.454% at the 90% confidence level for the total test group of 4,800 cases. We plan to examine differences across subgroups as small as 500 cases (in each of the test and control groups). For these smaller-sample cases a difference in response rate of 4.505% will be required to identify a difference as significant at the 90% confidence level.

The test design is expected to provide a reliable indicator of differences in the duration of interview measure as small as 29.750 seconds at the 90% confidence level for the total test group. We also plan to examine differences across subgroups as small as 500 cases (in each of the test and control groups) for this statistic. For these smaller-sample cases a difference in duration of interview of 92.176 seconds will be required to identify a difference as significant at the 90% confidence level.

If you have any questions about this request, please contact Ken Robertson by telephone at 202-691-5440 or by e-mail at Robertson.Ken@bls.gov.

2