Using paradata to improve data quality: A case study on post-wildfire restoration

Data quality is a crucial problem in applied research, particularly for that based on online surveys. This study addresses data quality in discrete choice experiments (DCE) by analysing participants' time to answer each choice task. Our methodology for evaluating data quality is implemented in a case study addressing wildfire restoration, providing a practical application of our approach. Using the paradata, we find that nearly 30% of respondents complete the DCE too quickly, spending less than one second in two or more choice tasks and without reading the attributes description. Individuals most likely to rush through the DCE are those under 40 with a medium-low level of education. Interestingly, the selection of the status quo option is not related to speeding through the tasks. After accounting for these data quality issues, we assess preferences for restoring recently burned areas. Among the different post-fire restoration actions, the most valued by respondents corresponded to the protection of environmental services at risk. Our estimates are more consistent when controlling for data quality, with the willingness to pay (WTP) estimates becoming more efficient and tending to decrease.