Open AccessWhose data can we trust: How meta-predictions can be used to uncover credible respondents in survey data
Author(s) -
Sonja Radas,
Dražen Prelec
Publication year - 2019
Publication title -
plos one
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.99
H-Index - 332
ISSN - 1932-6203
DOI - 10.1371/journal.pone.0225432
Subject(s) - reliability (semiconductor) , survey data collection , quality (philosophy) , data quality , identification (biology) , data science , meta analysis , psychology , computer science , test (biology) , marketing , statistics , medicine , mathematics , paleontology , metric (unit) , power (physics) , physics , philosophy , botany , epistemology , quantum mechanics , business , biology
Many areas of economics use subjective data, although it had been known to present problems regarding its reliability. To improve data quality, researchers may use scoring rules that reward respondents so that it is most profitable for them to tell the truth. However, if the subjects are not well informed about the topic or if they do not pay sufficient attention, they will produce data that could not be dependably used for decision-making even though subjects gave their honest answer. In this paper we show how meta-predictions (respondents’ predictions about choices of others) can be used for identification of respondents who produce dependable data. We use purchase intention survey, a popular method to elicit early adoption forecasts for a new concept, as a test bed for our approach. We present results from three online experiments, demonstrating that corrected purchase intentions are closer to the real outcomes.