Abstract:
|
We view responding to a survey question as decision-making: R gains/loses utility by how s/he is seen by the interviewer/analysts. Econometric and cognitive theories suggest processes Rs use to decide how to answer factual quantitative questions. Econometricians Arrow & Hurwicz (1972) and Cohen & Jaffray (1980) suggest people who are completely ignorant about a quantity answer with extreme values of offered intervals. In contrast, cognitive psychologists Schwartz, Hippler, Deutch, & Strack (1985) theorize Rs see pre-set intervals as hints of normative values, guiding them to choose middle values. We test the opposing predictions on data collected under the Respondent-Generated Interval protocol (RGI, Press 2002).
In RGI, Rs are asked to recall a quantitative fact and give lower and upper bounds for its possible value. Our data, from a record-check survey conducted at the U.S. Census Bureau on recall of salary /wages and income/dividends, include Rs' confidence ratings of recall accuracy (Press & Marquis, 2001). We test whether less confident Rs place the recalled quantity near the bounds they supply (as Arrow/ Hurwicz suggest), or center it (as Schwartz et al., suggest).
|