Estimating Response to Follow-ups in Mail Surveys

A simple method for estimating the effectiveness of successivefollow-ups in mail surveys is presented.     

The Presidents and the Polls

Since Franklin Delano Roosevelt, presidents have used opinionpolls to aid them in being elected and as input to policy formulation.Several examples of such effective use are presented. Pollsmay also be misused by presidents simply to enhance presidentialpopularity. Such actions are generally ineffective either forincreasing presidential popularity or for making wise policydecisions.     

The Use of Bounded Recall Procedures in Single Interviews

This note describes procedures for bounding recall in a singleinterview. The methods tested seem to be as effective as themore traditional panel procedures in reducing telescoping. Nomajor administration problems are observed.     

Question Threat and Response Bias

Perceived normative threat influences responses to questionsin surveys. Respondents who report that questions about an activitywould make most people very uneasy are less likely to reportever engaging in that activity than are persons who report lessuneasiness. If respondents do not admit to participating inan activity, perceived threat appears to have acted as a gatekeeperto prevent further questions. Since perceived threat is associatedwith underreporting, some simple adjustment methods may be usedto improve behavioral estimates.     

On Finding and Interviewing the Needles in the Haystack: The Use of Multiplicity Sampling

A multiplicity sample of a relatively rare population—Vietnameraveterans—provides insight into the following fieldissues: yield, location problems, coverage bias, and the effectof inclusion rule, i.e., eligible kin nominators. Here, thelatter included parents, siblings, aunts and uncles. The resultantyield was double that of a conventional sample but was muchhigher for black and Mexican-American veterans than for whites.Location problems (on which there was little prior knowledge)were less serious than anticipated, requiring persistence butnot extravagant expenditures to solve. Undercoverage bias wasreduced by "nonselective" screening; this raised location cost,however. Relative yield, an indicator of selection bias, variedby kin category. Parents showed a higher relative yield thandid siblings, while aunts and uncles were strikingly low innominations relative to their numbers.     

The Network Polls: A Critical Review

The network polls perform an important service reasonably competently,although they all have major weaknesses in questionnaire testingand in analyses of results. They are an excellent source ofmaterial for secondary analyses.