The Challenge
On June 12, 2020, the CDC published a report on household cleaning and disinfection practices during COVID-19, concluding that 39% of Americans had engaged in at least one high-risk cleaning behavior in the previous month. That included 19% who used bleach on food, 18% who applied cleaning products to their skin, and 4% who reported drinking or gargling diluted bleach or another disinfectant.
The CDC's report drew on the same type of online market research panels used across the survey industry. Those panels are also where inattentive, distracted, or dishonest respondents are hardest to screen out, and the CDC's numbers had not been checked for that kind of noise before they were published.
The Solution
CloudResearch researchers Leib Litman, PhD, and Zohn Rosen, PhD, ran a new survey using the same type of panels the CDC had relied on, matching its sample on size and demographics. This time, Sentry's behavioral assessment and fraud detection methods were applied to flag careless, inattentive, and dishonest respondents before their answers counted toward the results.
Rebuilding the CDC's Sample with Screening in Place
The reanalysis surveyed more than 2,500 people whose demographics closely matched the CDC's original respondent pool, then applied Sentry to identify and remove low-quality responses before calculating the final rates.
Isolating Where the Risky Numbers Came From
Respondents flagged by Sentry reported the riskiest behaviors, like drinking or gargling a disinfectant, far more often than respondents who passed the screening. That gap showed how much of the CDC's original numbers depended on exactly the kind of low-quality respondents Sentry is built to catch.
Results
Once flagged respondents were removed, the share of people reporting any high-risk cleaning behavior fell from the CDC's reported 39% to 20%. The riskiest specific behavior, gargling with a disinfectant, dropped to just 0.9% of respondents, with 0.5% specifically reporting bleach. Among the respondents Sentry flagged as low-quality, more than 31% reported one of the riskiest behaviors; among the rest, only about 1.5% did.
That gap between flagged and unflagged respondents accounts for nearly all of the CDC's original finding. When CloudResearch added the flagged respondents back into the sample, the results converged almost exactly with what the CDC had published, pointing to low-quality respondents, not a genuine shift in public behavior, as the source of the alarming numbers.
Litman and Rosen published the full reanalysis as No, Americans are not gargling bleach, and the findings were covered in the announcement of the counter-report.
What's Next
CloudResearch continues to argue for a fit-for-purpose approach to data quality: screening should be most stringent when researchers are measuring rare or low-frequency behaviors, since even a small share of careless or dishonest respondents can distort those numbers enough to shape media coverage and public health messaging.
About the Centers for Disease Control and Prevention (CDC)
The Centers for Disease Control and Prevention (CDC) is the United States' national public health agency, operating under the U.S. Department of Health and Human Services. Headquartered in Atlanta, Georgia, the CDC tracks disease outbreaks, conducts and funds public health research, and issues guidance on protecting the public from health threats.