If You Provide It, Will They Read It? The Effect of Information on Choices

This paper investigates the effect of information on respondent's choices in an internet survey for measuring the value of water quality improvements in Deckers Creek (DC) watershed in Monongalia and Peterson Counties of West Virginia, USA. A multiattribute, choice experiment and multinomial logit (MNL) models are used in estimating the marginal utilities of restoring the three attributes of DC: aquatic life, swimming safety, and scenic quality. Response times serve as proxy variables regarding whether respondents read or did not read all the information provided in the survey. Response times fell quickly, but then tapered off as they progressed through the various sections of the survey. Results show that the estimated coefficients of subsamples, read and did not read all the information, were statistically different from each other. Based on log likelihood tests of MNL models, two subsamples of the survey population (read and did not read all information) were found to be from different populations. Estimates of marginal utilities reveal that respondents value aquatic life restoration the highest, followed by scenic quality restoration. Average compensating variation estimates for full restoration of the aquatic life and scenic quality attributes are $9 and $ 6 per month per household, respectively, when the subsamples are pooled. However, the individual subsamples resulted in $5 per month for aquatic life and $3 per month for scenic quality for respondents that read the information, while respondents that did not read the information resulted in statistically higher estimates of $16 and $12, respectively. While respondents' motives for not reading the resource information provided is uncertain, results show their values for watershed restoration are substantially higher than respondents that read the information.