In this section we outline the elements of a plan for a comprehensive survey of employers.
An overview of the survey's implementation plan is provided as Exhibit 5.3.
5.3.1 Sample Design
The sample design for the comprehensive survey plan reflects one possible approach to identifying and selecting businesses for participation in the employer survey. The target population includes all business establishments and public agencies that have at least one employee.(24) The business establishments would be identified using the InfoUSA business directory.(25) This directory is recommended based on its national coverage, file characteristics (including flags for establishments and contact information for human resources personnel), easy access, and cost. The InfoUSA sample frame could be augmented with data from the Longitudinal Employer-Household Dynamics (LEHD) data source, which includes additional information on employer and employee characteristics for employers in more than 20 states.
The plan uses a multi-staged sampling strategy in which establishments are initially stratified in terms of three size categories: small (fewer than 20 employees), medium (20-99 employees), and large (100 or more employees). The number of establishments selected in each of these strata would be proportional to the total number of workers employed by all companies in that group. Within each establishment size stratum, four regional strata corresponding to the four U.S. Census regions would be established. Here, the sampling fraction would be proportional to the number of establishments for that region.
Alternatively, employers in selected sectors (e.g., service sectors) could be oversampled. Oversampling particular groups of establishments would result in more reliable specific estimates for such groups. However, it could also result in a larger sample and increased survey costs.
In his MTCES and four-city employer surveys, Holzer used these same establishment-size categories. However, these surveys did not stratify by regions. Instead, separate samples were drawn for specific cities.
Based on these design specifications, a sample of 3,000 survey completions is advisable. Because the number of TANF recipients in the general population is relatively small, a general survey of employers might find a relatively small percentage of employers quite possibly less than 25 percent aware that they have recruited or hired current and/or former TANF recipients. Nonetheless, a sample that includes 3,000 survey completions would permit reliable estimates of the prevalence of various employer practices.(26)
This sample would also permit extensive subgroup analysis. For example, it should be possible to make reliable estimates for business establishments in the service and manufacturing sectors as well as for establishments in the public sector. It should also be possible to make separate estimates for different regions of the country.(27)
5.3.2 Survey Method
The implementation plan for the comprehensive survey of employers is based on a four-step design, which is intended to produce an overall response rate of at least 70 percent. As summarized in Exhibit 5.3, the plan has several important features:
1. Screening Telephone Survey. An initial screening survey would be conducted with sampled establishments to verify important characteristics such as whether they are still in operation, the number of full- and part-time employees, and the name and contact information for the individual responsible for entry-level hiring. A separate screening survey improves the overall efficiency of the full telephone survey by ensuring that the interviewers target the correct person within the organization and that only eligible firms are included in the study. The screening survey would be conducted using computer-assisted telephone interviewing (CATI) and, on average, last no longer than five minutes.
2. Pre-Survey Announcement Letter. Using the contact information gathered during the initial screening survey, a pre-survey announcement letter would be sent to individuals in sampled establishments who are responsible for entry-level hiring decisions. This letter would describe the survey, its sponsors, and when and how the individual would be contacted for a telephone survey interview. Contacts also would be given the option of calling a toll-free hotline to complete a survey at their convenience, schedule an interview appointment, or ask questions about the survey. This letter would be sent approximately two weeks prior to initiating the study's telephone survey component.
3. Telephone Survey Administration. Identified respondents from the sampled establishments would be contacted by telephone to complete a survey on their practices for hiring and employing current and former TANF recipients. The length of the survey's administration would not exceed 20 minutes. CATI would be used to administer the survey and enter data, which greatly improves survey efficiency and overall data quality.
The survey questionnaire would open with a few additional screener questions to ensure that the correct contact person within the establishment has been identified. The survey also would be designed and programmed to allow for multiple respondents for cases where this is appropriate. Initially, no more than 15 attempts would be made to reach the contact person. At this juncture, survey supervisors should review the situation to determine if an alternative contact or method should be used to reach the selected establishment. In addition, approximately three weeks after beginning the telephone survey, reminder letters should be sent to all non-respondents. This letter would remind the contacts about the survey, its importance, and the options available for completing the survey.
Interviewers should be trained to work with respondents in ways that minimize survey refusals. This includes flexible interview scheduling procedures such as the option to complete the survey over the course of more than one call, offering alternative options for completing the survey (e.g., via internet) and addressing respondents' concerns about survey questions, content, and procedures. Some respondents can be expected to refuse the survey. In all cases, individuals who refuse the survey must be treated with respect; however, we recommend one attempt to convert the refusal.(28) Individuals who refuse would receive one conversion refusal contact from a supervisor. At this time, the supervisor would probe to determine whether it is the "establishment" that is refusing or the individual. If it is the individual, another contact within the organization would be sought. In situations where respondents indicate that they cannot participate because of company policy, survey supervisors would probe to find out whether the policy allows for mail surveys and, depending on the overall response rate, this individual might be asked to complete a mail survey at some later date.
5.3.3 Survey Pre-Testing
The core survey instrument would be pre-tested with a limited sample of establishments. To the extent possible, the pre-test would reflect the actual survey circumstances, including administering the survey using CATI. The administration of this pre-test should be carefully monitored and, at the survey's conclusion, the pre-test survey respondents should be asked about their experience taking the survey. It is particularly important to identify questions that are difficult to answer and to obtain reactions to the flow and general tone of the questionnaire.
"report.pdf" (pdf, 212.99Kb)