COMPARING WATER QUALITY VALUATION ACROSS PROBABILITY AND NON-PROBABILITY SAMPLES

This thesis compares the results of a stated preference survey administered to three samples: one non-probability sample and two non-probability samples. The probability sample is an address-based sample from the USPS postal delivery file, while the two non-probability samples are from the opt-in panels, MTurk and Qualtrics. The survey used a single binary referendum contingent valuation question with respondents voting on a water quality change at a cost to their household. To understand differences in economic values across samples, we compared results of logit models that relate the referendum vote to cost and each water quality index. Several tests reveal differences across samples. First, almost all parameters were significantly different across samples except for water clarity. Second, we compared marginal willingness to pay (MWTP). However, many of the MWTP estimates for individual water quality indices were not significantly different across the three sources. Third, we calculated total WTP (TWTP) for a range of non-marginal changes. The MTurk values were always significantly greater than the address sample at the 1% level, and the Qualtrics values were significantly greater than the address sample for changes up to about a 20% improvement. In summary, we find that the non-probability methods generate different valuation results than the probability-based sample, especially in terms of TWTP.