Cleaning Matters! Lessons Learned from Comparing Response Behaviors and Willingness to Pay in Probability- and Non-Probability Samples

Nonmarket valuation surveys are designed to ask the who, what, when, where and why for a population of interest to understand preferences for environmental goods. Recent declines in survey response rates and high costs associated with traditional survey modes (e.g., mail), along with recent advances in online sampling have led to increased use of non probability sample frames. This raises an important question for stated preference surveys about potential differences in willingness to pay (WTP) based on data collected by probability versus nonprobability samples. We develop a layered, sequential approach to test whether data processing and adjustments to estimation strategies can lead to similar welfare distributions for nonmarket attributes. Using a survey on the protection of safe recreation hours at ocean beaches, we find that our proposed process decreases the variance of marginal WTP for the non-probability sample and produces WTP distributions that overlap with the probability sample for our key attribute of interest.