Abstract :
Several studies document that iterative question formats used in contingent valuation studies produce anomalies in respondent behavior that appear to threaten the validity of welfare estimates. By decomposing iterative question formats into their ascending and descending sequences, we show that these anomalies occur only in ascending sequences. We describe the conditions under which comparable patterns of behavior are likely to be found in other iterative question formats that begin with a discrete willingness-to-pay question. We then develop a unified explanation of these anomalies using our model of framing based on prospect theory. We provide the first head-to-head test of rival explanations of these behavioral patterns by developing refutable hypotheses for the strategic behavior, yea-saying, anchoring, and cost-expectations models. Finally, based on our own statistically robust model, we show how these anomalies can be eliminated without loss of the statistical efficiency of most iterative question formats.
