ABSTRACT

"Question format and anchoring effects on the elicitation of house value and
checking account value"

Michael Hurd

In the Health and Retirement Study, wave 3, about 10,000 subjects were
randomly assigned to two types of treatments in the elicitation of the
bracket interval for house value and for checking account value.   The first
treatment was assignment to one of three question formats: "[would the value
be] greater than $$$;  greater than or about $$$;  less than or more than
$$$.  The second treatment was assignment to one of three entry points or
anchors into a bracketing sequence.  The question format influenced the
distribution of outcomes:  both of the unbalanced formats produced
distributions that were shifted to the right of the distribution of the
balanced format.  The anchoring effects depended on the knowledge of the
respondent.  Financial respondents (the spouse most knowledgeable about
financial matters) showed little anchoring effects;  nonfinancial
respondents showed economically meaningful anchoring.  These results imply
that the elicitation of economic information by bracketing or the use of
brackets to reduce item nonresponse can lead to bias in population estimates
of mean and median both through question format effects and anchoring effects.