3. Mai 2016
"Make some Noise: Information Garbling and the Elicitation of Sensitive Data"
Fear of sanctions makes people reluctant to reveal sensitive information about themselves and/or others. Important phenomena such as corruption, domestic violence, or mental health problems are therefore notoriously difficult to measure. The randomized response method (Warner 1965) has been proposed as a solution for this problem. The basic idea is that the data collector elicits garbled information that grants the respondent plausible deniability. However, since the the respondent himself can always undo the garbling, threats can effectively distort information collection. Using a simple laboratory experiment, we confirm that the randomized response method works well only in the short term. In the long term, the information is not better than directly elicited information. We propose an alternative method that relies on "hard-wired garbling" and makes it impossible for the respondent to undo the garbling. While the implementation of hard-wired noise comes at an efficiency cost, we show that, in the long term, this method allows collecting more reliable data.