Data quality of platforms and panels for online behavioral research

Data quality of platforms and panels for online behavioral research

29 September 2021 | Eyal Peer¹ · David Rothschild² · Andrew Gordon³ · Zak Evernden³ · Ekaterina Damer³
This study examines data quality across online behavioral research platforms and panels, including Amazon Mechanical Turk (MTurk), CloudResearch, Prolific, Qualtrics, and Dynata. Researchers identified key data quality aspects: attention, comprehension, honesty, and reliability. Two studies (N ~ 4000) were conducted, with and without data quality filters. Study 1 (without filters) found Prolific provided the highest data quality on all measures. Study 2 (with filters) showed high data quality on CloudResearch and Prolific, while MTurk had alarmingly low data quality even with filters. Reputation (approval ratings) did not predict data quality, but frequency of use and purpose (e.g., main income source) did. Participants using MTurk as their main income source but spending few hours per week had the lowest data quality. A framework for future data quality research is proposed, considering the evolving nature of platforms and panels. Key data quality aspects for behavioral research include attention (reading instructions), comprehension (understanding instructions), honesty (truthful responses), and reliability (consistent responses). Study 1 found significant differences in these aspects across platforms. Prolific had the highest attention and comprehension rates, while MTurk and Dynata had the lowest. Prolific participants were more honest, and reliability was higher on Prolific and Qualtrics. Study 2, with prescreening filters, confirmed these findings, with Prolific and CR showing high data quality, while MTurk had lower quality. Usage patterns, such as frequency and purpose, predicted data quality. Participants using MTurk as their main income source but spending few hours per week had lower data quality. Prolific participants who used the site as their main income source had higher data quality than those who did not. Data quality was also influenced by payment rates and the use of prescreening filters. Overall, Prolific provided the highest data quality across all measures, while MTurk and Dynata had lower quality. Middleman services like Qualtrics and Dynata showed comparable data quality to MTurk but were less efficient. Study 2 confirmed these findings, with Prolific and CR showing high data quality, while MTurk had lower quality. The study highlights the importance of data quality in online research and the need for future research to consider factors like payment rates and prescreening filters.This study examines data quality across online behavioral research platforms and panels, including Amazon Mechanical Turk (MTurk), CloudResearch, Prolific, Qualtrics, and Dynata. Researchers identified key data quality aspects: attention, comprehension, honesty, and reliability. Two studies (N ~ 4000) were conducted, with and without data quality filters. Study 1 (without filters) found Prolific provided the highest data quality on all measures. Study 2 (with filters) showed high data quality on CloudResearch and Prolific, while MTurk had alarmingly low data quality even with filters. Reputation (approval ratings) did not predict data quality, but frequency of use and purpose (e.g., main income source) did. Participants using MTurk as their main income source but spending few hours per week had the lowest data quality. A framework for future data quality research is proposed, considering the evolving nature of platforms and panels. Key data quality aspects for behavioral research include attention (reading instructions), comprehension (understanding instructions), honesty (truthful responses), and reliability (consistent responses). Study 1 found significant differences in these aspects across platforms. Prolific had the highest attention and comprehension rates, while MTurk and Dynata had the lowest. Prolific participants were more honest, and reliability was higher on Prolific and Qualtrics. Study 2, with prescreening filters, confirmed these findings, with Prolific and CR showing high data quality, while MTurk had lower quality. Usage patterns, such as frequency and purpose, predicted data quality. Participants using MTurk as their main income source but spending few hours per week had lower data quality. Prolific participants who used the site as their main income source had higher data quality than those who did not. Data quality was also influenced by payment rates and the use of prescreening filters. Overall, Prolific provided the highest data quality across all measures, while MTurk and Dynata had lower quality. Middleman services like Qualtrics and Dynata showed comparable data quality to MTurk but were less efficient. Study 2 confirmed these findings, with Prolific and CR showing high data quality, while MTurk had lower quality. The study highlights the importance of data quality in online research and the need for future research to consider factors like payment rates and prescreening filters.
Reach us at info@study.space
Understanding Data quality of platforms and panels for online behavioral research