Does the way a question is asked affect the answer given? This thesis looks at this question from two perspectives.

First, using contingent valuation data for water quality in the Monongahela River in Western Pennsylvania, an analysis is performed to test for differences in responses across four different formats of the willingness-to-pay question. This analysis shows that the question format has some systematic relationship with the valuation responses given. Specifically, iterative bidding games with high starting points lead to higher willingness-to-pay values than other formats. This result appears to arise from information relayed to the respondent from the question format. That is, since the bidding begins at a high value the respondent interprets that information as a cue to what the value of the commodity should be.

The four different question formats are then re-interpreted in a referendum context. Referendum campaigns can be characterized by the amount of information provided to the voters and the degree of focus of that information. For example, some referendum campaigns give the voters a great deal of data, and that data all gives a consistent message. Other campaigns provide very little information for the voters to use in their decision. By transforming the willingness-to-pay data into a discrete-choice format, referendum models are simulated. Then, an analysis shows how the characteristics of the information content affect the vote. Also, the value of the median voter is calculated. This median shows the value at which the referendum just passes. This analysis shows that the campaign associated with the high starting point bidding game yields that highest percentage of yes votes and the highest median voter value.