Non response

SOCIAL SECURITY STATEMENT SURVEY

OMB No. 0960-NEW

ADDENDUM: Non-Response Analysis

The survey procedures for sample release, contact, callback and refusal conversion efforts are designed to maximize response rate. The target response rate for this survey is 80% as specified in the OMB guidance document. However, since respondent recall of the SSA Statement is vital to the amount and quality of information about their satisfaction with the current document, it is essential to interview respondents as close as possible to their receipt of the Statement. Hence, we will be trying to interview most respondents within one to two weeks of their receipt of the Statement. In order to improve the quality of the data collected by a short recall period, we will have to restrict the number of calling attempts and the elapsed time for refusal conversion as a result of a short field period. If as a result the response rate falls below 85%, we will conduct a non-response analysis as required by OMB protocol.

There are three possible sources of respondent characteristics from which the completed cases can be compared to the non-completed cases for an analysis of potential non-response bias. First, we can use information available on the sampling frame for comparisons between completed and non-completed interviews. Second, we can use auxiliary data to categorize the cases based on geographic location for a comparison of completes and non-completed cases. Third, we can try to obtain information during the interview from all contacts, including refusals and callbacks, which can be used to compare them to completed cases. It should be noted in conducting a non-response analysis in a telephone survey, it is important to distinguish between four general types of outcomes: completed interviews, refused or terminated interviews, contacts that did not yield a complete or refused interview, and non-contacts.

The sampling frame for the baseline SSA Statement Survey includes designated respondent age (date of birth), gender and earnings type. Hence, we can analyze the survey outcome by age, gender and earnings type. If there is a significant difference between completes and non-completes in the sample, we can use these three variables for post-stratification weighting.

Since we have address and telephone number, we can also classify respondents by Census region, Census district, and type of place (central city, SMSA remainder and non-SMSA). We can compare respondents to non-respondents on district or region and type of place. If there is a statistically significant difference, we may wish to consider the use of one or more of these in post-stratification weighting. We can also classify the telephone number based on the percent African-American and Hispanic population in the associated exchange, and the average income based on Census estimates. This would identify potential bias in the achieved sample by race, ethnicity and income of area, but it would not be appropriate for case-level weighting.

We can attempt to collect information from all contacts for additional non-response analysis. However, most refusals occur before the survey introduction is completed, so there is limited opportunity to collect this information. Additionally, those who refuse after the introduction generally refuse to provide even limited information about themselves. Finally, since we will have information on respondent age, gender, earnings type and location, the most useful additional information to identify non-response bias would be race, ethnicity, education and income. Since these are among the sensitive questions for most respondents, we are unlikely to get many responses. Hence, we would recommend limiting the measures of potential non-response bias to those that we can obtain for all cases.

Social Security Statement Survey

Part B - Page 2 of 2

April 17, 2008